Compare commits
121 Commits
remove_rul
...
main
| Author | SHA1 | Date | |
|---|---|---|---|
|
a425aaae44 | ||
|
7f462a85d5 | ||
|
|
ee54e6d33c | ||
|
|
c8bac2f25d | ||
|
|
9d568e2e16 | ||
|
|
5752b22217 | ||
|
|
66273bee9d | ||
|
|
5b1eba118f | ||
|
|
3b76b884c3 | ||
|
|
13e7117426 | ||
|
|
c4b411ce7a | ||
|
|
4bfd6e4bdf | ||
|
|
dde09419fa | ||
|
|
35ad1be7fb | ||
|
|
31cfe2dccc | ||
|
|
da71bb7e76 | ||
|
|
5ff6666aec | ||
|
|
acd4385c80 | ||
|
|
23454a9f63 | ||
|
|
177cfa08fc | ||
|
|
781a47b2db | ||
|
|
79fbf8028b | ||
|
|
9a9330bdc0 | ||
|
|
20e79b7530 | ||
|
|
dc4f93c1a3 | ||
|
|
7642e91b93 | ||
|
|
e83423c179 | ||
|
|
f5fe732847 | ||
|
|
eb89247281 | ||
|
|
d0f751ec23 | ||
|
|
055e5c9548 | ||
|
|
932ead2ce7 | ||
|
|
01866c3064 | ||
|
|
d83dd67095 | ||
|
|
5b6c2245fa | ||
|
|
33bce4a44f | ||
|
|
58f5aff348 | ||
|
|
d1b74453a8 | ||
|
|
bebc604fc6 | ||
|
|
54e69cb421 | ||
|
|
b815f4749d | ||
|
|
dcb2b8c8c4 | ||
|
|
99be3e6cb5 | ||
|
|
11257ee600 | ||
|
|
831dcb624c | ||
|
|
80c6a1e344 | ||
|
02237a1f28 | ||
|
|
bcac10ca74 | ||
|
|
9a3d0b51d8 | ||
|
|
df38407fb3 | ||
|
|
0914919499 | ||
|
|
918b6b9cb4 | ||
|
|
55f3424118 | ||
|
|
98e5ecd887 | ||
|
|
58b438cdb5 | ||
|
|
c4af598b75 | ||
|
c78fe2d9fa | ||
|
|
bae6e28b51 | ||
|
8f38e9508f | ||
|
de8a87dea1 | ||
|
|
48fad9f387 | ||
|
|
4ec5f08646 | ||
|
|
a1b66f0cd4 | ||
|
|
34bbe47d75 | ||
|
0ef850e96c | ||
|
|
c91a819e77 | ||
|
|
71462a9f93 | ||
|
|
c5f1daf8f0 | ||
|
|
952c3e85d9 | ||
|
|
f6b72897cd | ||
|
|
8b6314fde3 | ||
|
|
4d42fb664d | ||
|
|
9457c25ced | ||
|
|
ca686afc3e | ||
|
|
4b01214225 | ||
|
|
0e63a86fe5 | ||
|
|
4e8b2be044 | ||
|
|
df12d96f9c | ||
|
|
fcb5de6793 | ||
|
|
661662caec | ||
|
|
2a681ce1eb | ||
|
|
52e4516e04 | ||
|
|
95085c4e72 | ||
|
|
55d3c9d723 | ||
|
|
618ab3778b | ||
|
|
89d9bdb5e2 | ||
|
|
291f680c16 | ||
|
|
3658769c7a | ||
|
|
1be9bc8d0a | ||
|
|
81d63a4383 | ||
|
|
333ba3a3fd | ||
|
|
7aa091cbf8 | ||
|
|
d309a5bc50 | ||
|
|
346dd4968e | ||
|
|
f5c9fa0593 | ||
|
|
d5e8df856d | ||
|
|
2d85e7e120 | ||
|
|
1a4a1ad885 | ||
|
|
c1ff364c29 | ||
|
|
bd0ef0e2cc | ||
|
|
0216163885 | ||
|
|
c42718649f | ||
|
|
650df8217a | ||
|
|
f7946ff713 | ||
|
|
b312fc37fc | ||
|
|
ad96f09600 | ||
|
|
090f1ef91a | ||
|
|
b2957d90f0 | ||
|
|
f8aea814c5 | ||
|
|
91c19f07d3 | ||
|
|
315b203082 | ||
|
|
caf4eda1be | ||
|
|
21ba3ebe8c | ||
|
|
f0a934a393 | ||
|
|
941420b417 | ||
|
|
1ea10cdbfa | ||
|
|
b99d816057 | ||
|
|
f89a6816a8 | ||
|
|
890137e7b3 | ||
|
|
75395ba196 | ||
|
|
7e3145e2eb |
19
.github/configs/cluster-config.yaml
vendored
Normal file
19
.github/configs/cluster-config.yaml
vendored
Normal file
@ -0,0 +1,19 @@
|
||||
apiVersion: kind.x-k8s.io/v1alpha4
|
||||
kind: Cluster
|
||||
nodes:
|
||||
- role: control-plane
|
||||
kubeadmConfigPatches:
|
||||
- |
|
||||
kind: ClusterConfiguration
|
||||
controllerManager:
|
||||
extraArgs:
|
||||
bind-address: 0.0.0.0
|
||||
secure-port: "10257"
|
||||
scheduler:
|
||||
extraArgs:
|
||||
bind-address: 0.0.0.0
|
||||
secure-port: "10259"
|
||||
- |
|
||||
kind: KubeProxyConfiguration
|
||||
metricsBindAddress: 0.0.0.0:10249
|
||||
- role: worker
|
||||
@ -19,6 +19,9 @@ jobs:
|
||||
updateVersions:
|
||||
name: Update the subcharts
|
||||
runs-on: "ubuntu-latest"
|
||||
permissions:
|
||||
contents: write
|
||||
id-token: write
|
||||
steps:
|
||||
- name: Checkout
|
||||
uses: actions/checkout@v2
|
||||
@ -66,6 +69,20 @@ jobs:
|
||||
echo "changed=true" >> "${GITHUB_OUTPUT}"
|
||||
fi
|
||||
|
||||
- id: get-secrets
|
||||
uses: grafana/shared-workflows/actions/get-vault-secrets@main
|
||||
with:
|
||||
# Secrets placed in the ci/repo/grafana/<repo>/<path> path in Vault
|
||||
repo_secrets: |
|
||||
APP_ID=github-app:app-id
|
||||
PRIVATE_KEY=github-app:private-key
|
||||
|
||||
- uses: actions/create-github-app-token@v1
|
||||
id: app-token
|
||||
with:
|
||||
app-id: ${{ env.APP_ID }}
|
||||
private-key: ${{ env.PRIVATE_KEY }}
|
||||
|
||||
- name: Create pull request
|
||||
if: steps.update-loki.outputs.changed == 'true' || steps.update-grafana-alloy.outputs.changed == 'true' || steps.update-mimir-distributed.outputs.changed == 'true' || steps.update-tempo-distributed.outputs.changed == 'true' || steps.update-minio.outputs.changed == 'true'
|
||||
uses: peter-evans/create-pull-request@v5
|
||||
@ -79,10 +96,15 @@ jobs:
|
||||
labels: dependencies
|
||||
branch: chore/update-dependencies
|
||||
delete-branch: true
|
||||
team-reviewers: "@grafana/loki-squad"
|
||||
token: ${{ steps.app-token.outputs.token }}
|
||||
|
||||
updateGrafana:
|
||||
name: Update the Grafana version
|
||||
runs-on: "ubuntu-latest"
|
||||
permissions:
|
||||
contents: write
|
||||
id-token: write
|
||||
steps:
|
||||
- name: Checkout
|
||||
uses: actions/checkout@v2
|
||||
@ -98,6 +120,20 @@ jobs:
|
||||
echo "changed=true" >> "${GITHUB_OUTPUT}"
|
||||
fi
|
||||
|
||||
- id: get-secrets
|
||||
uses: grafana/shared-workflows/actions/get-vault-secrets@main
|
||||
with:
|
||||
# Secrets placed in the ci/repo/grafana/<repo>/<path> path in Vault
|
||||
repo_secrets: |
|
||||
APP_ID=github-app:app-id
|
||||
PRIVATE_KEY=github-app:private-key
|
||||
|
||||
- uses: actions/create-github-app-token@v1
|
||||
id: app-token
|
||||
with:
|
||||
app-id: ${{ env.APP_ID }}
|
||||
private-key: ${{ env.PRIVATE_KEY }}
|
||||
|
||||
- name: Create pull request
|
||||
if: steps.update-grafana.outputs.changed == 'true'
|
||||
uses: peter-evans/create-pull-request@v5
|
||||
@ -111,3 +147,5 @@ jobs:
|
||||
labels: dependencies
|
||||
branch: chore/update-minio
|
||||
delete-branch: true
|
||||
team-reviewers: "@grafana/loki-squad"
|
||||
token: ${{ steps.app-token.outputs.token }}
|
||||
|
||||
11
.github/workflows/helm-ci.yml
vendored
11
.github/workflows/helm-ci.yml
vendored
@ -1,6 +1,7 @@
|
||||
---
|
||||
name: helm-ci
|
||||
on:
|
||||
workflow_dispatch:
|
||||
pull_request:
|
||||
paths:
|
||||
- "charts/meta-monitoring/**"
|
||||
@ -24,7 +25,7 @@ jobs:
|
||||
# runs-on: ubuntu-latest
|
||||
# steps:
|
||||
# - name: Checkout
|
||||
# uses: actions/checkout@v3
|
||||
# uses: actions/checkout@v4
|
||||
# with:
|
||||
# fetch-depth: 0
|
||||
|
||||
@ -38,10 +39,10 @@ jobs:
|
||||
# - name: Set up Python
|
||||
# uses: actions/setup-python@v4
|
||||
# with:
|
||||
# python-version: 3.7
|
||||
# python-version: 3.9
|
||||
|
||||
# - name: Set up chart-testing
|
||||
# uses: helm/chart-testing-action@v2.4.0
|
||||
# uses: helm/chart-testing-action@v2
|
||||
|
||||
# - name: Run chart-testing (list-changed)
|
||||
# id: list-changed
|
||||
@ -55,10 +56,10 @@ jobs:
|
||||
# run: ct lint --config "${CT_CONFIGFILE}" --check-version-increment=false
|
||||
|
||||
# - name: Create kind cluster
|
||||
# uses: helm/kind-action@v1.8.0
|
||||
# uses: helm/kind-action@v1
|
||||
# if: steps.list-changed.outputs.changed == 'true'
|
||||
# with:
|
||||
# config: tools/kind.config
|
||||
# config: "${{ github.workspace }}/.github/configs/cluster-config.yaml"
|
||||
|
||||
# - name: Run chart-testing (install)
|
||||
# run: |
|
||||
|
||||
16
README.md
16
README.md
@ -1,8 +1,6 @@
|
||||
# meta-monitoring-chart
|
||||
|
||||
This is a meta-monitoring chart for Loki.
|
||||
|
||||
Note that this is pre-production software at the moment.
|
||||
This is a meta-monitoring chart for Loki, specifically Loki installed via the Loki helm chart.
|
||||
|
||||
## Local and cloud modes
|
||||
|
||||
@ -11,19 +9,15 @@ to small Loki, Mimir and Tempo installations running in the meta-monitoring name
|
||||
|
||||
|
||||
|
||||
To enable local mode set `local.<logs|metrics|traces>.enabled` to true.
|
||||
|
||||
In the cloud mode the logs, metrics and/or traces are sent to Grafana Cloud.
|
||||
|
||||
|
||||
|
||||
To enable cloud mode set `cloud.<logs|metrics|traces>.enabled` to true. The `endpoint`, `username` and `password` settings for your Grafana Cloud logs, metrics and traces instances have to be filled in as well.
|
||||
|
||||
Both modes can be enabled at the same time. Cloud mode is preferred.
|
||||
|
||||
## Installation
|
||||
|
||||
For more instructions including how to update the chart go to the [installation](docs/installation.md) page.
|
||||
For more instructions including how to install the chart go to the [installation](docs/installation.md) page.
|
||||
|
||||
## Supported features
|
||||
|
||||
@ -33,8 +27,7 @@ For more instructions including how to update the chart go to the [installation]
|
||||
- Specify PII regexes that are applied to logs before they are sent to Loki (cloud or local). The capture group in the regex is replaced with *****.
|
||||
- a Grafana instance is installed (when local mode is used) with the relevant datasources installed. The following dashboards are installed:
|
||||
- logs dashboards
|
||||
- agent dashboards
|
||||
- Retention is set to 24 hours
|
||||
- Alloy dashboards
|
||||
|
||||
Most of these features are enabled by default. See the values.yaml file for how to enable/disable them.
|
||||
|
||||
@ -42,8 +35,7 @@ Most of these features are enabled by default. See the values.yaml file for how
|
||||
|
||||
- This has not been tested on Openshift yet.
|
||||
- The underlying Loki, Mimir and Tempo are at the default size installed by the Helm chart. This might need changing when monitoring bigger Loki, Mimir or Tempo installations.
|
||||
- MinIO is used as storage at the moment with a limited retention. At the moment this chart cannot be used for monitoring over longer periods.
|
||||
- Agent self monitoring is not done at the moment.
|
||||
- MinIO is used as storage for the local mode at the moment with a limited retention. At the moment this chart cannot be used for monitoring over longer periods.
|
||||
|
||||
## Developer help topics
|
||||
|
||||
|
||||
@ -1,18 +1,18 @@
|
||||
dependencies:
|
||||
- name: loki
|
||||
repository: https://grafana.github.io/helm-charts
|
||||
version: 6.5.0
|
||||
version: 6.29.0
|
||||
- name: alloy
|
||||
repository: https://grafana.github.io/helm-charts
|
||||
version: 0.1.1
|
||||
version: 0.12.5
|
||||
- name: mimir-distributed
|
||||
repository: https://grafana.github.io/helm-charts
|
||||
version: 5.3.0
|
||||
version: 5.6.0
|
||||
- name: tempo-distributed
|
||||
repository: https://grafana.github.io/helm-charts
|
||||
version: 1.9.9
|
||||
version: 1.33.0
|
||||
- name: minio
|
||||
repository: https://charts.min.io
|
||||
version: 5.2.0
|
||||
digest: sha256:5328702b5f6b0487aba8f7bc77d6abfcd5e094569e9205cd725971e3e31255dd
|
||||
generated: "2024-05-08T07:03:21.797461955Z"
|
||||
version: 5.4.0
|
||||
digest: sha256:5225a03d9003384639f5d43b1971126371269347f16f221b7aed377ab85d71be
|
||||
generated: "2025-03-27T07:03:11.17404081Z"
|
||||
|
||||
@ -13,7 +13,7 @@ type: application
|
||||
# This is the chart version. This version number should be incremented each time you make changes
|
||||
# to the chart and its templates, including the app version.
|
||||
# Versions are expected to follow Semantic Versioning (https://semver.org/)
|
||||
version: 0.0.2
|
||||
version: 1.3.0
|
||||
# This is the version number of the application being deployed. This version number should be
|
||||
# incremented each time you make changes to the application. Versions are not expected to
|
||||
# follow Semantic Versioning. They should reflect the version the application is using.
|
||||
@ -22,20 +22,20 @@ appVersion: "0.0.1"
|
||||
dependencies:
|
||||
- name: loki
|
||||
repository: https://grafana.github.io/helm-charts
|
||||
version: 6.5.0
|
||||
version: 6.29.0
|
||||
condition: local.logs.enabled
|
||||
- name: alloy
|
||||
repository: https://grafana.github.io/helm-charts
|
||||
version: 0.1.1
|
||||
version: 0.12.5
|
||||
- name: mimir-distributed
|
||||
repository: https://grafana.github.io/helm-charts
|
||||
version: 5.3.0
|
||||
version: 5.6.0
|
||||
condition: local.metrics.enabled
|
||||
- name: tempo-distributed
|
||||
repository: https://grafana.github.io/helm-charts
|
||||
version: 1.9.9
|
||||
version: 1.33.0
|
||||
condition: local.traces.enabled
|
||||
- name: minio
|
||||
repository: https://charts.min.io
|
||||
version: 5.2.0
|
||||
version: 5.4.0
|
||||
condition: local.minio.enabled
|
||||
|
||||
Binary file not shown.
BIN
charts/meta-monitoring/charts/alloy-0.12.5.tgz
Normal file
BIN
charts/meta-monitoring/charts/alloy-0.12.5.tgz
Normal file
Binary file not shown.
BIN
charts/meta-monitoring/charts/loki-6.29.0.tgz
Normal file
BIN
charts/meta-monitoring/charts/loki-6.29.0.tgz
Normal file
Binary file not shown.
Binary file not shown.
Binary file not shown.
BIN
charts/meta-monitoring/charts/mimir-distributed-5.6.0.tgz
Normal file
BIN
charts/meta-monitoring/charts/mimir-distributed-5.6.0.tgz
Normal file
Binary file not shown.
Binary file not shown.
BIN
charts/meta-monitoring/charts/minio-5.4.0.tgz
Normal file
BIN
charts/meta-monitoring/charts/minio-5.4.0.tgz
Normal file
Binary file not shown.
BIN
charts/meta-monitoring/charts/tempo-distributed-1.33.0.tgz
Normal file
BIN
charts/meta-monitoring/charts/tempo-distributed-1.33.0.tgz
Normal file
Binary file not shown.
Binary file not shown.
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
@ -1,786 +0,0 @@
|
||||
{
|
||||
"annotations": {
|
||||
"list": [ ]
|
||||
},
|
||||
"editable": true,
|
||||
"gnetId": null,
|
||||
"graphTooltip": 0,
|
||||
"hideControls": false,
|
||||
"links": [ ],
|
||||
"refresh": "30s",
|
||||
"rows": [
|
||||
{
|
||||
"collapse": false,
|
||||
"height": "250px",
|
||||
"panels": [
|
||||
{
|
||||
"aliasColors": { },
|
||||
"bars": false,
|
||||
"dashLength": 10,
|
||||
"dashes": false,
|
||||
"datasource": "$datasource",
|
||||
"fill": 1,
|
||||
"id": 1,
|
||||
"legend": {
|
||||
"avg": false,
|
||||
"current": false,
|
||||
"max": false,
|
||||
"min": false,
|
||||
"show": true,
|
||||
"total": false,
|
||||
"values": false
|
||||
},
|
||||
"lines": true,
|
||||
"linewidth": 1,
|
||||
"links": [ ],
|
||||
"nullPointMode": "null as zero",
|
||||
"percentage": false,
|
||||
"pointradius": 5,
|
||||
"points": false,
|
||||
"renderer": "flot",
|
||||
"seriesOverrides": [ ],
|
||||
"spaceLength": 10,
|
||||
"span": 12,
|
||||
"stack": false,
|
||||
"steppedLine": false,
|
||||
"styles": [
|
||||
{
|
||||
"alias": "Time",
|
||||
"dateFormat": "YYYY-MM-DD HH:mm:ss",
|
||||
"pattern": "Time",
|
||||
"type": "hidden"
|
||||
},
|
||||
{
|
||||
"alias": "Count",
|
||||
"colorMode": null,
|
||||
"colors": [ ],
|
||||
"dateFormat": "YYYY-MM-DD HH:mm:ss",
|
||||
"decimals": 2,
|
||||
"link": false,
|
||||
"linkTargetBlank": false,
|
||||
"linkTooltip": "Drill down",
|
||||
"linkUrl": "",
|
||||
"pattern": "Value #A",
|
||||
"thresholds": [ ],
|
||||
"type": "hidden",
|
||||
"unit": "short"
|
||||
},
|
||||
{
|
||||
"alias": "Uptime",
|
||||
"colorMode": null,
|
||||
"colors": [ ],
|
||||
"dateFormat": "YYYY-MM-DD HH:mm:ss",
|
||||
"decimals": 2,
|
||||
"link": false,
|
||||
"linkTargetBlank": false,
|
||||
"linkTooltip": "Drill down",
|
||||
"linkUrl": "",
|
||||
"pattern": "Value #B",
|
||||
"thresholds": [ ],
|
||||
"type": "number",
|
||||
"unit": "short"
|
||||
},
|
||||
{
|
||||
"alias": "Container",
|
||||
"colorMode": null,
|
||||
"colors": [ ],
|
||||
"dateFormat": "YYYY-MM-DD HH:mm:ss",
|
||||
"decimals": 2,
|
||||
"link": false,
|
||||
"linkTargetBlank": false,
|
||||
"linkTooltip": "Drill down",
|
||||
"linkUrl": "",
|
||||
"pattern": "container",
|
||||
"thresholds": [ ],
|
||||
"type": "number",
|
||||
"unit": "short"
|
||||
},
|
||||
{
|
||||
"alias": "Pod",
|
||||
"colorMode": null,
|
||||
"colors": [ ],
|
||||
"dateFormat": "YYYY-MM-DD HH:mm:ss",
|
||||
"decimals": 2,
|
||||
"link": false,
|
||||
"linkTargetBlank": false,
|
||||
"linkTooltip": "Drill down",
|
||||
"linkUrl": "",
|
||||
"pattern": "pod",
|
||||
"thresholds": [ ],
|
||||
"type": "number",
|
||||
"unit": "short"
|
||||
},
|
||||
{
|
||||
"alias": "Version",
|
||||
"colorMode": null,
|
||||
"colors": [ ],
|
||||
"dateFormat": "YYYY-MM-DD HH:mm:ss",
|
||||
"decimals": 2,
|
||||
"link": false,
|
||||
"linkTargetBlank": false,
|
||||
"linkTooltip": "Drill down",
|
||||
"linkUrl": "",
|
||||
"pattern": "version",
|
||||
"thresholds": [ ],
|
||||
"type": "number",
|
||||
"unit": "short"
|
||||
},
|
||||
{
|
||||
"alias": "",
|
||||
"colorMode": null,
|
||||
"colors": [ ],
|
||||
"dateFormat": "YYYY-MM-DD HH:mm:ss",
|
||||
"decimals": 2,
|
||||
"pattern": "/.*/",
|
||||
"thresholds": [ ],
|
||||
"type": "string",
|
||||
"unit": "short"
|
||||
}
|
||||
],
|
||||
"targets": [
|
||||
{
|
||||
"expr": "count by (pod, container, version) (agent_build_info{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"$container\"})",
|
||||
"format": "table",
|
||||
"instant": true,
|
||||
"intervalFactor": 2,
|
||||
"legendFormat": "",
|
||||
"refId": "A",
|
||||
"step": 10
|
||||
},
|
||||
{
|
||||
"expr": "max by (pod, container) (time() - process_start_time_seconds{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"$container\"})",
|
||||
"format": "table",
|
||||
"instant": true,
|
||||
"intervalFactor": 2,
|
||||
"legendFormat": "",
|
||||
"refId": "B",
|
||||
"step": 10
|
||||
}
|
||||
],
|
||||
"thresholds": [ ],
|
||||
"timeFrom": null,
|
||||
"timeShift": null,
|
||||
"title": "Agent Stats",
|
||||
"tooltip": {
|
||||
"shared": true,
|
||||
"sort": 2,
|
||||
"value_type": "individual"
|
||||
},
|
||||
"transform": "table",
|
||||
"type": "table",
|
||||
"xaxis": {
|
||||
"buckets": null,
|
||||
"mode": "time",
|
||||
"name": null,
|
||||
"show": true,
|
||||
"values": [ ]
|
||||
},
|
||||
"yaxes": [
|
||||
{
|
||||
"format": "short",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": 0,
|
||||
"show": true
|
||||
},
|
||||
{
|
||||
"format": "short",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": null,
|
||||
"show": false
|
||||
}
|
||||
]
|
||||
}
|
||||
],
|
||||
"repeat": null,
|
||||
"repeatIteration": null,
|
||||
"repeatRowId": null,
|
||||
"showTitle": true,
|
||||
"title": "Agent Stats",
|
||||
"titleSize": "h6"
|
||||
},
|
||||
{
|
||||
"collapse": false,
|
||||
"height": "250px",
|
||||
"panels": [
|
||||
{
|
||||
"aliasColors": { },
|
||||
"bars": false,
|
||||
"dashLength": 10,
|
||||
"dashes": false,
|
||||
"datasource": "$datasource",
|
||||
"fill": 1,
|
||||
"id": 2,
|
||||
"legend": {
|
||||
"avg": false,
|
||||
"current": false,
|
||||
"max": false,
|
||||
"min": false,
|
||||
"show": true,
|
||||
"total": false,
|
||||
"values": false
|
||||
},
|
||||
"lines": true,
|
||||
"linewidth": 1,
|
||||
"links": [ ],
|
||||
"nullPointMode": "null as zero",
|
||||
"percentage": false,
|
||||
"pointradius": 5,
|
||||
"points": false,
|
||||
"renderer": "flot",
|
||||
"seriesOverrides": [ ],
|
||||
"spaceLength": 10,
|
||||
"span": 6,
|
||||
"stack": false,
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum(rate(prometheus_target_sync_length_seconds_sum{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"$container\"}[5m])) by (pod, scrape_job) * 1e3",
|
||||
"format": "time_series",
|
||||
"intervalFactor": 2,
|
||||
"legendFormat": "{{pod}}/{{scrape_job}}",
|
||||
"legendLink": null,
|
||||
"step": 10
|
||||
}
|
||||
],
|
||||
"thresholds": [ ],
|
||||
"timeFrom": null,
|
||||
"timeShift": null,
|
||||
"title": "Target Sync",
|
||||
"tooltip": {
|
||||
"shared": true,
|
||||
"sort": 2,
|
||||
"value_type": "individual"
|
||||
},
|
||||
"type": "graph",
|
||||
"xaxis": {
|
||||
"buckets": null,
|
||||
"mode": "time",
|
||||
"name": null,
|
||||
"show": true,
|
||||
"values": [ ]
|
||||
},
|
||||
"yaxes": [
|
||||
{
|
||||
"format": "ms",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": 0,
|
||||
"show": true
|
||||
},
|
||||
{
|
||||
"format": "short",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": null,
|
||||
"show": false
|
||||
}
|
||||
]
|
||||
},
|
||||
{
|
||||
"aliasColors": { },
|
||||
"bars": false,
|
||||
"dashLength": 10,
|
||||
"dashes": false,
|
||||
"datasource": "$datasource",
|
||||
"fill": 10,
|
||||
"id": 3,
|
||||
"legend": {
|
||||
"avg": false,
|
||||
"current": false,
|
||||
"max": false,
|
||||
"min": false,
|
||||
"show": true,
|
||||
"total": false,
|
||||
"values": false
|
||||
},
|
||||
"lines": true,
|
||||
"linewidth": 0,
|
||||
"links": [ ],
|
||||
"nullPointMode": "null as zero",
|
||||
"percentage": false,
|
||||
"pointradius": 5,
|
||||
"points": false,
|
||||
"renderer": "flot",
|
||||
"seriesOverrides": [ ],
|
||||
"spaceLength": 10,
|
||||
"span": 6,
|
||||
"stack": true,
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by (pod) (prometheus_sd_discovered_targets{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"$container\"})",
|
||||
"format": "time_series",
|
||||
"intervalFactor": 2,
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null,
|
||||
"step": 10
|
||||
}
|
||||
],
|
||||
"thresholds": [ ],
|
||||
"timeFrom": null,
|
||||
"timeShift": null,
|
||||
"title": "Targets",
|
||||
"tooltip": {
|
||||
"shared": true,
|
||||
"sort": 2,
|
||||
"value_type": "individual"
|
||||
},
|
||||
"type": "graph",
|
||||
"xaxis": {
|
||||
"buckets": null,
|
||||
"mode": "time",
|
||||
"name": null,
|
||||
"show": true,
|
||||
"values": [ ]
|
||||
},
|
||||
"yaxes": [
|
||||
{
|
||||
"format": "short",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": 0,
|
||||
"show": true
|
||||
},
|
||||
{
|
||||
"format": "short",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": null,
|
||||
"show": false
|
||||
}
|
||||
]
|
||||
}
|
||||
],
|
||||
"repeat": null,
|
||||
"repeatIteration": null,
|
||||
"repeatRowId": null,
|
||||
"showTitle": true,
|
||||
"title": "Prometheus Discovery",
|
||||
"titleSize": "h6"
|
||||
},
|
||||
{
|
||||
"collapse": false,
|
||||
"height": "250px",
|
||||
"panels": [
|
||||
{
|
||||
"aliasColors": { },
|
||||
"bars": false,
|
||||
"dashLength": 10,
|
||||
"dashes": false,
|
||||
"datasource": "$datasource",
|
||||
"fill": 1,
|
||||
"id": 4,
|
||||
"legend": {
|
||||
"avg": false,
|
||||
"current": false,
|
||||
"max": false,
|
||||
"min": false,
|
||||
"show": true,
|
||||
"total": false,
|
||||
"values": false
|
||||
},
|
||||
"lines": true,
|
||||
"linewidth": 1,
|
||||
"links": [ ],
|
||||
"nullPointMode": "null as zero",
|
||||
"percentage": false,
|
||||
"pointradius": 5,
|
||||
"points": false,
|
||||
"renderer": "flot",
|
||||
"seriesOverrides": [ ],
|
||||
"spaceLength": 10,
|
||||
"span": 4,
|
||||
"stack": false,
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "rate(prometheus_target_interval_length_seconds_sum{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"$container\"}[5m])\n/\nrate(prometheus_target_interval_length_seconds_count{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"$container\"}[5m])\n* 1e3\n",
|
||||
"format": "time_series",
|
||||
"intervalFactor": 2,
|
||||
"legendFormat": "{{pod}} {{interval}} configured",
|
||||
"legendLink": null,
|
||||
"step": 10
|
||||
}
|
||||
],
|
||||
"thresholds": [ ],
|
||||
"timeFrom": null,
|
||||
"timeShift": null,
|
||||
"title": "Average Scrape Interval Duration",
|
||||
"tooltip": {
|
||||
"shared": true,
|
||||
"sort": 2,
|
||||
"value_type": "individual"
|
||||
},
|
||||
"type": "graph",
|
||||
"xaxis": {
|
||||
"buckets": null,
|
||||
"mode": "time",
|
||||
"name": null,
|
||||
"show": true,
|
||||
"values": [ ]
|
||||
},
|
||||
"yaxes": [
|
||||
{
|
||||
"format": "ms",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": 0,
|
||||
"show": true
|
||||
},
|
||||
{
|
||||
"format": "short",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": null,
|
||||
"show": false
|
||||
}
|
||||
]
|
||||
},
|
||||
{
|
||||
"aliasColors": { },
|
||||
"bars": false,
|
||||
"dashLength": 10,
|
||||
"dashes": false,
|
||||
"datasource": "$datasource",
|
||||
"fill": 10,
|
||||
"id": 5,
|
||||
"legend": {
|
||||
"avg": false,
|
||||
"current": false,
|
||||
"max": false,
|
||||
"min": false,
|
||||
"show": true,
|
||||
"total": false,
|
||||
"values": false
|
||||
},
|
||||
"lines": true,
|
||||
"linewidth": 0,
|
||||
"links": [ ],
|
||||
"nullPointMode": "null as zero",
|
||||
"percentage": false,
|
||||
"pointradius": 5,
|
||||
"points": false,
|
||||
"renderer": "flot",
|
||||
"seriesOverrides": [ ],
|
||||
"spaceLength": 10,
|
||||
"span": 4,
|
||||
"stack": true,
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by (job) (rate(prometheus_target_scrapes_exceeded_sample_limit_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"$container\"}[1m]))",
|
||||
"format": "time_series",
|
||||
"intervalFactor": 2,
|
||||
"legendFormat": "exceeded sample limit: {{job}}",
|
||||
"legendLink": null,
|
||||
"step": 10
|
||||
},
|
||||
{
|
||||
"expr": "sum by (job) (rate(prometheus_target_scrapes_sample_duplicate_timestamp_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"$container\"}[1m]))",
|
||||
"format": "time_series",
|
||||
"intervalFactor": 2,
|
||||
"legendFormat": "duplicate timestamp: {{job}}",
|
||||
"legendLink": null,
|
||||
"step": 10
|
||||
},
|
||||
{
|
||||
"expr": "sum by (job) (rate(prometheus_target_scrapes_sample_out_of_bounds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"$container\"}[1m]))",
|
||||
"format": "time_series",
|
||||
"intervalFactor": 2,
|
||||
"legendFormat": "out of bounds: {{job}}",
|
||||
"legendLink": null,
|
||||
"step": 10
|
||||
},
|
||||
{
|
||||
"expr": "sum by (job) (rate(prometheus_target_scrapes_sample_out_of_order_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"$container\"}[1m]))",
|
||||
"format": "time_series",
|
||||
"intervalFactor": 2,
|
||||
"legendFormat": "out of order: {{job}}",
|
||||
"legendLink": null,
|
||||
"step": 10
|
||||
}
|
||||
],
|
||||
"thresholds": [ ],
|
||||
"timeFrom": null,
|
||||
"timeShift": null,
|
||||
"title": "Scrape failures",
|
||||
"tooltip": {
|
||||
"shared": true,
|
||||
"sort": 2,
|
||||
"value_type": "individual"
|
||||
},
|
||||
"type": "graph",
|
||||
"xaxis": {
|
||||
"buckets": null,
|
||||
"mode": "time",
|
||||
"name": null,
|
||||
"show": true,
|
||||
"values": [ ]
|
||||
},
|
||||
"yaxes": [
|
||||
{
|
||||
"format": "short",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": 0,
|
||||
"show": true
|
||||
},
|
||||
{
|
||||
"format": "short",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": null,
|
||||
"show": false
|
||||
}
|
||||
]
|
||||
},
|
||||
{
|
||||
"aliasColors": { },
|
||||
"bars": false,
|
||||
"dashLength": 10,
|
||||
"dashes": false,
|
||||
"datasource": "$datasource",
|
||||
"fill": 10,
|
||||
"id": 6,
|
||||
"legend": {
|
||||
"avg": false,
|
||||
"current": false,
|
||||
"max": false,
|
||||
"min": false,
|
||||
"show": true,
|
||||
"total": false,
|
||||
"values": false
|
||||
},
|
||||
"lines": true,
|
||||
"linewidth": 0,
|
||||
"links": [ ],
|
||||
"nullPointMode": "null as zero",
|
||||
"percentage": false,
|
||||
"pointradius": 5,
|
||||
"points": false,
|
||||
"renderer": "flot",
|
||||
"seriesOverrides": [ ],
|
||||
"spaceLength": 10,
|
||||
"span": 4,
|
||||
"stack": true,
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by (job, instance_group_name) (rate(agent_wal_samples_appended_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"$container\"}[5m]))",
|
||||
"format": "time_series",
|
||||
"intervalFactor": 2,
|
||||
"legendFormat": "{{job}} {{instance_group_name}}",
|
||||
"legendLink": null,
|
||||
"step": 10
|
||||
}
|
||||
],
|
||||
"thresholds": [ ],
|
||||
"timeFrom": null,
|
||||
"timeShift": null,
|
||||
"title": "Appended Samples",
|
||||
"tooltip": {
|
||||
"shared": true,
|
||||
"sort": 2,
|
||||
"value_type": "individual"
|
||||
},
|
||||
"type": "graph",
|
||||
"xaxis": {
|
||||
"buckets": null,
|
||||
"mode": "time",
|
||||
"name": null,
|
||||
"show": true,
|
||||
"values": [ ]
|
||||
},
|
||||
"yaxes": [
|
||||
{
|
||||
"format": "short",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": 0,
|
||||
"show": true
|
||||
},
|
||||
{
|
||||
"format": "short",
|
||||
"label": null,
|
||||
"logBase": 1,
|
||||
"max": null,
|
||||
"min": null,
|
||||
"show": false
|
||||
}
|
||||
]
|
||||
}
|
||||
],
|
||||
"repeat": null,
|
||||
"repeatIteration": null,
|
||||
"repeatRowId": null,
|
||||
"showTitle": true,
|
||||
"title": "Prometheus Retrieval",
|
||||
"titleSize": "h6"
|
||||
}
|
||||
],
|
||||
"schemaVersion": 14,
|
||||
"style": "dark",
|
||||
"tags": [
|
||||
"grafana-agent-mixin"
|
||||
],
|
||||
"templating": {
|
||||
"list": [
|
||||
{
|
||||
"current": {
|
||||
"text": "default",
|
||||
"value": "default"
|
||||
},
|
||||
"hide": 0,
|
||||
"label": "Data Source",
|
||||
"name": "datasource",
|
||||
"options": [ ],
|
||||
"query": "prometheus",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"type": "datasource"
|
||||
},
|
||||
{
|
||||
"allValue": ".+",
|
||||
"current": {
|
||||
"selected": true,
|
||||
"text": "All",
|
||||
"value": "$__all"
|
||||
},
|
||||
"datasource": "$datasource",
|
||||
"hide": 0,
|
||||
"includeAll": true,
|
||||
"label": "cluster",
|
||||
"multi": true,
|
||||
"name": "cluster",
|
||||
"options": [ ],
|
||||
"query": "label_values(agent_build_info, cluster)",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"sort": 2,
|
||||
"tagValuesQuery": "",
|
||||
"tags": [ ],
|
||||
"tagsQuery": "",
|
||||
"type": "query",
|
||||
"useTags": false
|
||||
},
|
||||
{
|
||||
"allValue": ".+",
|
||||
"current": {
|
||||
"selected": true,
|
||||
"text": "All",
|
||||
"value": "$__all"
|
||||
},
|
||||
"datasource": "$datasource",
|
||||
"hide": 0,
|
||||
"includeAll": true,
|
||||
"label": "namespace",
|
||||
"multi": true,
|
||||
"name": "namespace",
|
||||
"options": [ ],
|
||||
"query": "label_values(agent_build_info, namespace)",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"sort": 2,
|
||||
"tagValuesQuery": "",
|
||||
"tags": [ ],
|
||||
"tagsQuery": "",
|
||||
"type": "query",
|
||||
"useTags": false
|
||||
},
|
||||
{
|
||||
"allValue": ".+",
|
||||
"current": {
|
||||
"selected": true,
|
||||
"text": "All",
|
||||
"value": "$__all"
|
||||
},
|
||||
"datasource": "$datasource",
|
||||
"hide": 0,
|
||||
"includeAll": true,
|
||||
"label": "container",
|
||||
"multi": true,
|
||||
"name": "container",
|
||||
"options": [ ],
|
||||
"query": "label_values(agent_build_info, container)",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"sort": 2,
|
||||
"tagValuesQuery": "",
|
||||
"tags": [ ],
|
||||
"tagsQuery": "",
|
||||
"type": "query",
|
||||
"useTags": false
|
||||
},
|
||||
{
|
||||
"allValue": "grafana-agent-.*",
|
||||
"current": {
|
||||
"selected": true,
|
||||
"text": "All",
|
||||
"value": "$__all"
|
||||
},
|
||||
"datasource": "$datasource",
|
||||
"hide": 0,
|
||||
"includeAll": true,
|
||||
"label": "pod",
|
||||
"multi": true,
|
||||
"name": "pod",
|
||||
"options": [ ],
|
||||
"query": "label_values(agent_build_info{container=~\"$container\"}, pod)",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"sort": 2,
|
||||
"tagValuesQuery": "",
|
||||
"tags": [ ],
|
||||
"tagsQuery": "",
|
||||
"type": "query",
|
||||
"useTags": false
|
||||
}
|
||||
]
|
||||
},
|
||||
"time": {
|
||||
"from": "now-1h",
|
||||
"to": "now"
|
||||
},
|
||||
"timepicker": {
|
||||
"refresh_intervals": [
|
||||
"5s",
|
||||
"10s",
|
||||
"30s",
|
||||
"1m",
|
||||
"5m",
|
||||
"15m",
|
||||
"30m",
|
||||
"1h",
|
||||
"2h",
|
||||
"1d"
|
||||
],
|
||||
"time_options": [
|
||||
"5m",
|
||||
"15m",
|
||||
"1h",
|
||||
"6h",
|
||||
"12h",
|
||||
"24h",
|
||||
"2d",
|
||||
"7d",
|
||||
"30d"
|
||||
]
|
||||
},
|
||||
"timezone": "",
|
||||
"title": "Agent",
|
||||
"uid": "",
|
||||
"version": 0
|
||||
}
|
||||
1171
charts/meta-monitoring/src/dashboards/alloy-cluster-node.json
Normal file
1171
charts/meta-monitoring/src/dashboards/alloy-cluster-node.json
Normal file
File diff suppressed because it is too large
Load Diff
@ -0,0 +1,540 @@
|
||||
{
|
||||
"annotations": {
|
||||
"list": [
|
||||
{
|
||||
"builtIn": 1,
|
||||
"datasource": {
|
||||
"type": "grafana",
|
||||
"uid": "-- Grafana --"
|
||||
},
|
||||
"enable": true,
|
||||
"hide": true,
|
||||
"iconColor": "rgba(0, 211, 255, 1)",
|
||||
"name": "Annotations & Alerts",
|
||||
"type": "dashboard"
|
||||
},
|
||||
{
|
||||
"datasource": "$loki_datasource",
|
||||
"enable": true,
|
||||
"expr": "{cluster=\"$cluster\", container=\"kube-diff-logger\"} | json | namespace_extracted=\"alloy\" | name_extracted=~\"alloy.*\"",
|
||||
"iconColor": "rgba(0, 211, 255, 1)",
|
||||
"instant": false,
|
||||
"name": "Deployments",
|
||||
"titleFormat": "{{cluster}}/{{namespace}}"
|
||||
}
|
||||
]
|
||||
},
|
||||
"editable": true,
|
||||
"fiscalYearStartMonth": 0,
|
||||
"graphTooltip": 1,
|
||||
"id": 27,
|
||||
"links": [
|
||||
{
|
||||
"icon": "doc",
|
||||
"targetBlank": true,
|
||||
"title": "Documentation",
|
||||
"tooltip": "Clustering documentation",
|
||||
"type": "link",
|
||||
"url": "https://grafana.com/docs/alloy/latest/reference/cli/run/#clustered-mode"
|
||||
},
|
||||
{
|
||||
"asDropdown": true,
|
||||
"icon": "external link",
|
||||
"includeVars": true,
|
||||
"keepTime": true,
|
||||
"tags": [
|
||||
"alloy-mixin"
|
||||
],
|
||||
"targetBlank": false,
|
||||
"title": "Dashboards",
|
||||
"type": "dashboards"
|
||||
}
|
||||
],
|
||||
"panels": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 9,
|
||||
"w": 8,
|
||||
"x": 0,
|
||||
"y": 0
|
||||
},
|
||||
"id": 1,
|
||||
"options": {
|
||||
"colorMode": "value",
|
||||
"graphMode": "area",
|
||||
"justifyMode": "auto",
|
||||
"orientation": "auto",
|
||||
"reduceOptions": {
|
||||
"calcs": [
|
||||
"lastNotNull"
|
||||
],
|
||||
"fields": "",
|
||||
"values": false
|
||||
},
|
||||
"showPercentChange": false,
|
||||
"textMode": "auto",
|
||||
"wideLayout": true
|
||||
},
|
||||
"pluginVersion": "10.4.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "count(cluster_node_info{cluster=\"$cluster\", namespace=\"$namespace\"})",
|
||||
"instant": true,
|
||||
"legendFormat": "__auto",
|
||||
"range": false,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Nodes",
|
||||
"type": "stat"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Nodes info.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"align": "auto",
|
||||
"cellOptions": {
|
||||
"type": "auto"
|
||||
},
|
||||
"inspect": false
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
}
|
||||
},
|
||||
"overrides": [
|
||||
{
|
||||
"matcher": {
|
||||
"id": "byName",
|
||||
"options": "Dashboard"
|
||||
},
|
||||
"properties": [
|
||||
{
|
||||
"id": "mappings",
|
||||
"value": [
|
||||
{
|
||||
"options": {
|
||||
"1": {
|
||||
"index": 0,
|
||||
"text": "Link"
|
||||
}
|
||||
},
|
||||
"type": "value"
|
||||
}
|
||||
]
|
||||
},
|
||||
{
|
||||
"id": "links",
|
||||
"value": [
|
||||
{
|
||||
"targetBlank": false,
|
||||
"title": "Detail dashboard for node",
|
||||
"url": "/d/4047e755d822da63c8158cde32ae4dce/alloy-cluster-node?var-instance=${__data.fields.instance}&var-datasource=${datasource}&var-loki_datasource=${loki_datasource}&var-cluster=${cluster}&var-namespace=${namespace}"
|
||||
}
|
||||
]
|
||||
}
|
||||
]
|
||||
}
|
||||
]
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 9,
|
||||
"w": 16,
|
||||
"x": 8,
|
||||
"y": 0
|
||||
},
|
||||
"id": 2,
|
||||
"options": {
|
||||
"cellHeight": "sm",
|
||||
"footer": {
|
||||
"countRows": false,
|
||||
"fields": "",
|
||||
"reducer": [
|
||||
"sum"
|
||||
],
|
||||
"show": false
|
||||
},
|
||||
"showHeader": true
|
||||
},
|
||||
"pluginVersion": "10.4.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "cluster_node_info{cluster=\"$cluster\", namespace=\"$namespace\"}",
|
||||
"format": "table",
|
||||
"instant": true,
|
||||
"legendFormat": "__auto",
|
||||
"range": false,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Node table",
|
||||
"transformations": [
|
||||
{
|
||||
"id": "organize",
|
||||
"options": {
|
||||
"excludeByName": {
|
||||
"Time": true,
|
||||
"Value": false,
|
||||
"__name__": true,
|
||||
"cluster": true,
|
||||
"namespace": true,
|
||||
"state": false
|
||||
},
|
||||
"indexByName": {},
|
||||
"renameByName": {
|
||||
"Value": "Dashboard",
|
||||
"instance": "",
|
||||
"state": ""
|
||||
}
|
||||
}
|
||||
}
|
||||
],
|
||||
"type": "table"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Whether the cluster state has converged.\n\nIt is normal for the cluster state to be diverged briefly as gossip events propagate. It is not normal for the cluster state to be diverged for a long period of time.\n\nThis will show one of the following:\n\n* Converged: Nodes are aware of all other nodes, with the correct states.\n* Not converged: A subset of nodes aren't aware of their peers, or don't have an updated view of peer states.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"mappings": [
|
||||
{
|
||||
"options": {
|
||||
"1": {
|
||||
"color": "red",
|
||||
"index": 1,
|
||||
"text": "Not converged"
|
||||
}
|
||||
},
|
||||
"type": "value"
|
||||
},
|
||||
{
|
||||
"options": {
|
||||
"match": "null",
|
||||
"result": {
|
||||
"color": "green",
|
||||
"index": 0,
|
||||
"text": "Converged"
|
||||
}
|
||||
},
|
||||
"type": "special"
|
||||
}
|
||||
],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "suffix:nodes"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 9,
|
||||
"w": 8,
|
||||
"x": 0,
|
||||
"y": 9
|
||||
},
|
||||
"id": 3,
|
||||
"options": {
|
||||
"colorMode": "background",
|
||||
"graphMode": "none",
|
||||
"justifyMode": "auto",
|
||||
"orientation": "auto",
|
||||
"reduceOptions": {
|
||||
"calcs": [
|
||||
"lastNotNull"
|
||||
],
|
||||
"fields": "",
|
||||
"values": false
|
||||
},
|
||||
"showPercentChange": false,
|
||||
"textMode": "auto",
|
||||
"wideLayout": true
|
||||
},
|
||||
"pluginVersion": "10.4.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "clamp((\n sum(stddev by (state) (cluster_node_peers{cluster=\"$cluster\", namespace=\"$namespace\"}) != 0) or\n (sum(abs(sum without (state) (cluster_node_peers{cluster=\"$cluster\", namespace=\"$namespace\"})) - scalar(count(cluster_node_info{cluster=\"$cluster\", namespace=\"$namespace\"})) != 0))\n ),\n 1, 1\n)\n",
|
||||
"format": "time_series",
|
||||
"instant": true,
|
||||
"legendFormat": "__auto",
|
||||
"range": false,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Convergance state",
|
||||
"type": "stat"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "continuous-GrYlRd"
|
||||
},
|
||||
"custom": {
|
||||
"fillOpacity": 80,
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineWidth": 0,
|
||||
"spanNulls": true
|
||||
},
|
||||
"mappings": [
|
||||
{
|
||||
"options": {
|
||||
"0": {
|
||||
"color": "green",
|
||||
"text": "Yes"
|
||||
}
|
||||
},
|
||||
"type": "value"
|
||||
},
|
||||
{
|
||||
"options": {
|
||||
"1": {
|
||||
"color": "red",
|
||||
"text": "No"
|
||||
}
|
||||
},
|
||||
"type": "value"
|
||||
}
|
||||
],
|
||||
"max": 1,
|
||||
"noValue": 0,
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 9,
|
||||
"w": 16,
|
||||
"x": 8,
|
||||
"y": 9
|
||||
},
|
||||
"id": 4,
|
||||
"options": {
|
||||
"alignValue": "left",
|
||||
"legend": {
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"mergeValues": true,
|
||||
"rowHeight": 0.9,
|
||||
"showValue": "auto",
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "ceil(clamp((\n sum(stddev by (state) (cluster_node_peers{cluster=\"$cluster\", namespace=\"$namespace\"})) or\n (sum(abs(sum without (state) (cluster_node_peers{cluster=\"$cluster\", namespace=\"$namespace\"})) - scalar(count(cluster_node_info{cluster=\"$cluster\", namespace=\"$namespace\"}))))\n ),\n 0, 1\n))\n",
|
||||
"instant": false,
|
||||
"legendFormat": "Converged",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Convergance state timeline",
|
||||
"type": "state-timeline"
|
||||
}
|
||||
],
|
||||
"refresh": "10s",
|
||||
"schemaVersion": 39,
|
||||
"tags": [
|
||||
"alloy-mixin"
|
||||
],
|
||||
"templating": {
|
||||
"list": [
|
||||
{
|
||||
"current": {
|
||||
"selected": false,
|
||||
"text": "Mimir",
|
||||
"value": "mimir_ds"
|
||||
},
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "Data Source",
|
||||
"multi": false,
|
||||
"name": "datasource",
|
||||
"options": [],
|
||||
"query": "prometheus",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "datasource"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"selected": false,
|
||||
"text": "Loki",
|
||||
"value": "loki_ds"
|
||||
},
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "Loki Data Source",
|
||||
"multi": false,
|
||||
"name": "loki_datasource",
|
||||
"options": [],
|
||||
"query": "loki",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "datasource"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"isNone": true,
|
||||
"selected": false,
|
||||
"text": "None",
|
||||
"value": ""
|
||||
},
|
||||
"datasource": {
|
||||
"uid": "${datasource}"
|
||||
},
|
||||
"definition": "",
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "cluster",
|
||||
"multi": false,
|
||||
"name": "cluster",
|
||||
"options": [],
|
||||
"query": {
|
||||
"query": "label_values(alloy_component_controller_running_components, cluster)\n",
|
||||
"refId": "cluster"
|
||||
},
|
||||
"refresh": 2,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "query"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"isNone": true,
|
||||
"selected": false,
|
||||
"text": "None",
|
||||
"value": ""
|
||||
},
|
||||
"datasource": {
|
||||
"uid": "${datasource}"
|
||||
},
|
||||
"definition": "",
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "namespace",
|
||||
"multi": false,
|
||||
"name": "namespace",
|
||||
"options": [],
|
||||
"query": {
|
||||
"query": "label_values(alloy_component_controller_running_components{cluster=\"$cluster\"}, namespace)\n",
|
||||
"refId": "namespace"
|
||||
},
|
||||
"refresh": 2,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "query"
|
||||
}
|
||||
]
|
||||
},
|
||||
"time": {
|
||||
"from": "now-1h",
|
||||
"to": "now"
|
||||
},
|
||||
"timepicker": {
|
||||
"refresh_intervals": [
|
||||
"5s",
|
||||
"10s",
|
||||
"30s",
|
||||
"1m",
|
||||
"5m",
|
||||
"15m",
|
||||
"30m",
|
||||
"1h",
|
||||
"2h",
|
||||
"1d"
|
||||
],
|
||||
"time_options": [
|
||||
"5m",
|
||||
"15m",
|
||||
"1h",
|
||||
"6h",
|
||||
"12h",
|
||||
"24h",
|
||||
"2d",
|
||||
"7d",
|
||||
"30d",
|
||||
"90d"
|
||||
]
|
||||
},
|
||||
"timezone": "",
|
||||
"title": "Alloy / Cluster Overview",
|
||||
"uid": "",
|
||||
"version": 0,
|
||||
"weekStart": ""
|
||||
}
|
||||
970
charts/meta-monitoring/src/dashboards/alloy-controller.json
Normal file
970
charts/meta-monitoring/src/dashboards/alloy-controller.json
Normal file
@ -0,0 +1,970 @@
|
||||
{
|
||||
"annotations": {
|
||||
"list": [
|
||||
{
|
||||
"builtIn": 1,
|
||||
"datasource": {
|
||||
"type": "grafana",
|
||||
"uid": "-- Grafana --"
|
||||
},
|
||||
"enable": true,
|
||||
"hide": true,
|
||||
"iconColor": "rgba(0, 211, 255, 1)",
|
||||
"name": "Annotations & Alerts",
|
||||
"type": "dashboard"
|
||||
},
|
||||
{
|
||||
"datasource": "$loki_datasource",
|
||||
"enable": true,
|
||||
"expr": "{cluster=\"$cluster\", container=\"kube-diff-logger\"} | json | namespace_extracted=\"alloy\" | name_extracted=~\"alloy.*\"",
|
||||
"iconColor": "rgba(0, 211, 255, 1)",
|
||||
"instant": false,
|
||||
"name": "Deployments",
|
||||
"titleFormat": "{{cluster}}/{{namespace}}"
|
||||
}
|
||||
]
|
||||
},
|
||||
"editable": true,
|
||||
"fiscalYearStartMonth": 0,
|
||||
"graphTooltip": 1,
|
||||
"id": 28,
|
||||
"links": [
|
||||
{
|
||||
"icon": "doc",
|
||||
"targetBlank": true,
|
||||
"title": "Documentation",
|
||||
"tooltip": "Component controller documentation",
|
||||
"type": "link",
|
||||
"url": "https://grafana.com/docs/alloy/latest/concepts/component_controller/"
|
||||
},
|
||||
{
|
||||
"asDropdown": true,
|
||||
"icon": "external link",
|
||||
"includeVars": true,
|
||||
"keepTime": true,
|
||||
"tags": [
|
||||
"alloy-mixin"
|
||||
],
|
||||
"targetBlank": false,
|
||||
"title": "Dashboards",
|
||||
"type": "dashboards"
|
||||
}
|
||||
],
|
||||
"panels": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "The number of Alloy instances whose metrics are being sent and reported.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "instances"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 4,
|
||||
"w": 10,
|
||||
"x": 0,
|
||||
"y": 0
|
||||
},
|
||||
"id": 1,
|
||||
"options": {
|
||||
"colorMode": "none",
|
||||
"graphMode": "none",
|
||||
"justifyMode": "auto",
|
||||
"orientation": "auto",
|
||||
"reduceOptions": {
|
||||
"calcs": [
|
||||
"lastNotNull"
|
||||
],
|
||||
"fields": "",
|
||||
"values": false
|
||||
},
|
||||
"showPercentChange": false,
|
||||
"textMode": "auto",
|
||||
"wideLayout": true
|
||||
},
|
||||
"pluginVersion": "10.4.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "count(alloy_component_controller_evaluating{cluster=\"$cluster\", namespace=\"$namespace\"})",
|
||||
"instant": false,
|
||||
"legendFormat": "__auto",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Running instances",
|
||||
"type": "stat"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Breakdown of components by health across all running instances.\n\n* Healthy: components have been evaluated completely and are reporting themselves as healthy.\n* Unhealthy: Components either could not be evaluated or are reporting themselves as unhealthy.\n* Unknown: A component has been created but has not yet been started.\n* Exited: A component has exited. It will not return to the running state.\n\nMore information on a component's health state can be retrieved using\nthe Alloy UI.\n\nNote that components may be in a degraded state even if they report\nthemselves as healthy. Use component-specific dashboards and alerts\nto observe detailed information about the behavior of a component.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"mappings": [],
|
||||
"min": 0,
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
}
|
||||
]
|
||||
}
|
||||
},
|
||||
"overrides": [
|
||||
{
|
||||
"matcher": {
|
||||
"id": "byName",
|
||||
"options": "Unhealthy"
|
||||
},
|
||||
"properties": [
|
||||
{
|
||||
"id": "thresholds",
|
||||
"value": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 1
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
]
|
||||
},
|
||||
{
|
||||
"matcher": {
|
||||
"id": "byName",
|
||||
"options": "Unknown"
|
||||
},
|
||||
"properties": [
|
||||
{
|
||||
"id": "thresholds",
|
||||
"value": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "blue",
|
||||
"value": 1
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
]
|
||||
},
|
||||
{
|
||||
"matcher": {
|
||||
"id": "byName",
|
||||
"options": "Exited"
|
||||
},
|
||||
"properties": [
|
||||
{
|
||||
"id": "thresholds",
|
||||
"value": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "orange",
|
||||
"value": 1
|
||||
}
|
||||
]
|
||||
}
|
||||
}
|
||||
]
|
||||
}
|
||||
]
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 12,
|
||||
"w": 14,
|
||||
"x": 10,
|
||||
"y": 0
|
||||
},
|
||||
"id": 4,
|
||||
"options": {
|
||||
"displayMode": "gradient",
|
||||
"maxVizHeight": 300,
|
||||
"minVizHeight": 16,
|
||||
"minVizWidth": 8,
|
||||
"namePlacement": "auto",
|
||||
"orientation": "vertical",
|
||||
"reduceOptions": {
|
||||
"calcs": [
|
||||
"lastNotNull"
|
||||
],
|
||||
"fields": "",
|
||||
"values": false
|
||||
},
|
||||
"showUnfilled": true,
|
||||
"sizing": "auto",
|
||||
"valueMode": "color"
|
||||
},
|
||||
"pluginVersion": "10.4.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum(alloy_component_controller_running_components{cluster=\"$cluster\", namespace=\"$namespace\", health_type=\"healthy\"}) or vector(0)",
|
||||
"instant": true,
|
||||
"legendFormat": "Healthy",
|
||||
"range": false,
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum(alloy_component_controller_running_components{cluster=\"$cluster\", namespace=\"$namespace\", health_type=\"unhealthy\"}) or vector(0)",
|
||||
"instant": true,
|
||||
"legendFormat": "Unhealthy",
|
||||
"range": false,
|
||||
"refId": "B"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum(alloy_component_controller_running_components{cluster=\"$cluster\", namespace=\"$namespace\", health_type=\"unknown\"}) or vector(0)",
|
||||
"instant": true,
|
||||
"legendFormat": "Unknown",
|
||||
"range": false,
|
||||
"refId": "C"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum(alloy_component_controller_running_components{cluster=\"$cluster\", namespace=\"$namespace\", health_type=\"exited\"}) or vector(0)",
|
||||
"instant": true,
|
||||
"legendFormat": "Exited",
|
||||
"range": false,
|
||||
"refId": "D"
|
||||
}
|
||||
],
|
||||
"title": "Components by health",
|
||||
"type": "bargauge"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "The number of running components across all running instances.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "components"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 4,
|
||||
"w": 10,
|
||||
"x": 0,
|
||||
"y": 4
|
||||
},
|
||||
"id": 2,
|
||||
"options": {
|
||||
"colorMode": "none",
|
||||
"graphMode": "none",
|
||||
"justifyMode": "auto",
|
||||
"orientation": "auto",
|
||||
"reduceOptions": {
|
||||
"calcs": [
|
||||
"lastNotNull"
|
||||
],
|
||||
"fields": "",
|
||||
"values": false
|
||||
},
|
||||
"showPercentChange": false,
|
||||
"textMode": "auto",
|
||||
"wideLayout": true
|
||||
},
|
||||
"pluginVersion": "10.4.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum(alloy_component_controller_running_components{cluster=\"$cluster\", namespace=\"$namespace\"})",
|
||||
"instant": false,
|
||||
"legendFormat": "__auto",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Running components",
|
||||
"type": "stat"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "The percentage of components which are in a healthy state.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"mappings": [],
|
||||
"max": 1,
|
||||
"min": 0,
|
||||
"noValue": "No components",
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "percentunit"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 4,
|
||||
"w": 10,
|
||||
"x": 0,
|
||||
"y": 8
|
||||
},
|
||||
"id": 3,
|
||||
"options": {
|
||||
"colorMode": "value",
|
||||
"graphMode": "area",
|
||||
"justifyMode": "auto",
|
||||
"orientation": "auto",
|
||||
"reduceOptions": {
|
||||
"calcs": [
|
||||
"lastNotNull"
|
||||
],
|
||||
"fields": "",
|
||||
"values": false
|
||||
},
|
||||
"showPercentChange": false,
|
||||
"text": {
|
||||
"valueSize": 80
|
||||
},
|
||||
"textMode": "auto",
|
||||
"wideLayout": true
|
||||
},
|
||||
"pluginVersion": "10.4.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum(alloy_component_controller_running_components{cluster=\"$cluster\", namespace=\"$namespace\",health_type=\"healthy\"}) /\nsum(alloy_component_controller_running_components{cluster=\"$cluster\", namespace=\"$namespace\"})\n",
|
||||
"instant": false,
|
||||
"legendFormat": "__auto",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Overall component health",
|
||||
"type": "stat"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "The frequency at which components get updated.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "points",
|
||||
"fillOpacity": 0,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 3,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "none"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "ops"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 0,
|
||||
"y": 12
|
||||
},
|
||||
"id": 5,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "multi",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum by (instance) (rate(alloy_component_evaluation_seconds_count{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))",
|
||||
"instant": false,
|
||||
"legendFormat": "__auto",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Component evaluation rate",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "The percentiles for how long it takes to complete component evaluations.\n\nComponent evaluations must complete for components to have the latest\narguments. The longer the evaluations take, the slower it will be to\nreconcile the state of components.\n\nIf evaluation is taking too long, consider sharding your components to\ndeal with smaller amounts of data and reuse data as much as possible.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 0,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "none"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "s"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 8,
|
||||
"y": 12
|
||||
},
|
||||
"id": 6,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "histogram_quantile(0.99, sum(rate(alloy_component_evaluation_seconds{cluster=\"$cluster\",namespace=\"$namespace\"}[$__rate_interval])))\nor\nhistogram_quantile(0.99, sum by (le) (rate(alloy_component_evaluation_seconds_bucket{cluster=\"$cluster\",namespace=\"$namespace\"}[$__rate_interval])))\n",
|
||||
"instant": false,
|
||||
"legendFormat": "99th percentile",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "histogram_quantile(0.50, sum(rate(alloy_component_evaluation_seconds{cluster=\"$cluster\",namespace=\"$namespace\"}[$__rate_interval])))\nor\nhistogram_quantile(0.50, sum by (le) (rate(alloy_component_evaluation_seconds_bucket{cluster=\"$cluster\",namespace=\"$namespace\"}[$__rate_interval])))\n",
|
||||
"instant": false,
|
||||
"legendFormat": "50th percentile",
|
||||
"range": true,
|
||||
"refId": "B"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "(\n histogram_sum(sum(rate(alloy_component_evaluation_seconds{cluster=\"$cluster\",namespace=\"$namespace\"}[$__rate_interval]))) /\n histogram_count(sum(rate(alloy_component_evaluation_seconds{cluster=\"$cluster\",namespace=\"$namespace\"}[$__rate_interval])))\n)\nor\n(\n sum(rate(alloy_component_evaluation_seconds_sum{cluster=\"$cluster\",namespace=\"$namespace\"}[$__rate_interval])) /\n sum(rate(alloy_component_evaluation_seconds_count{cluster=\"$cluster\",namespace=\"$namespace\"}[$__rate_interval]))\n)\n",
|
||||
"instant": false,
|
||||
"legendFormat": "Average",
|
||||
"range": true,
|
||||
"refId": "C"
|
||||
}
|
||||
],
|
||||
"title": "Component evaluation time",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "The percentage of time spent evaluating 'slow' components - components that took longer than 1 minute to evaluate.\n\nIdeally, no component should take more than 1 minute to evaluate. The components displayed in this chart\nmay be a sign of a problem with the pipeline.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 0,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "none"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "percentunit"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 16,
|
||||
"y": 12
|
||||
},
|
||||
"id": 7,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum by (component_path, component_id) (rate(alloy_component_evaluation_slow_seconds{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))\n/ scalar(sum(rate(alloy_component_evaluation_seconds_sum{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])))\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{component path}} {{component_id}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Slow components evaluation times",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Detailed histogram view of how long component evaluations take.\n\nThe goal is to design your config so that evaluations take as little\ntime as possible; under 100ms is a good goal.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
}
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 0,
|
||||
"y": 22
|
||||
},
|
||||
"id": 8,
|
||||
"maxDataPoints": 30,
|
||||
"options": {
|
||||
"calculate": false,
|
||||
"cellGap": 0,
|
||||
"color": {
|
||||
"exponent": 0.5,
|
||||
"fill": "dark-orange",
|
||||
"mode": "scheme",
|
||||
"reverse": false,
|
||||
"scale": "exponential",
|
||||
"scheme": "Spectral",
|
||||
"steps": 64
|
||||
},
|
||||
"exemplars": {
|
||||
"color": "rgba(255,0,255,0.7)"
|
||||
},
|
||||
"filterValues": {
|
||||
"le": 0.1
|
||||
},
|
||||
"legend": {
|
||||
"show": true
|
||||
},
|
||||
"rowsFrame": {
|
||||
"layout": "auto"
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"showColorScale": false,
|
||||
"yHistogram": true
|
||||
},
|
||||
"yAxis": {
|
||||
"axisPlacement": "left",
|
||||
"reverse": false,
|
||||
"unit": "s"
|
||||
}
|
||||
},
|
||||
"pluginVersion": "10.4.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum(increase(alloy_component_evaluation_seconds{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))\nor ignoring (le)\nsum by (le) (increase(alloy_component_evaluation_seconds_bucket{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))\n",
|
||||
"format": "heatmap",
|
||||
"instant": false,
|
||||
"legendFormat": "{{le}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Component evaluation histogram",
|
||||
"type": "heatmap"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Detailed histogram of how long components wait to be evaluated after their dependency is updated.\n\nThe goal is to design your config so that most of the time components do not\nqueue for long; under 10ms is a good goal.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
}
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 8,
|
||||
"y": 22
|
||||
},
|
||||
"id": 9,
|
||||
"maxDataPoints": 30,
|
||||
"options": {
|
||||
"calculate": false,
|
||||
"cellGap": 0,
|
||||
"color": {
|
||||
"exponent": 0.5,
|
||||
"fill": "dark-orange",
|
||||
"mode": "scheme",
|
||||
"reverse": false,
|
||||
"scale": "exponential",
|
||||
"scheme": "Spectral",
|
||||
"steps": 64
|
||||
},
|
||||
"exemplars": {
|
||||
"color": "rgba(255,0,255,0.7)"
|
||||
},
|
||||
"filterValues": {
|
||||
"le": 0.1
|
||||
},
|
||||
"legend": {
|
||||
"show": true
|
||||
},
|
||||
"rowsFrame": {
|
||||
"layout": "auto"
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"showColorScale": false,
|
||||
"yHistogram": true
|
||||
},
|
||||
"yAxis": {
|
||||
"axisPlacement": "left",
|
||||
"reverse": false,
|
||||
"unit": "s"
|
||||
}
|
||||
},
|
||||
"pluginVersion": "10.4.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum(increase(alloy_component_dependencies_wait_seconds{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))\nor ignoring (le)\nsum by (le) (increase(alloy_component_dependencies_wait_seconds_bucket{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))\n",
|
||||
"format": "heatmap",
|
||||
"instant": false,
|
||||
"legendFormat": "{{le}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Component dependency wait histogram",
|
||||
"type": "heatmap"
|
||||
}
|
||||
],
|
||||
"refresh": "10s",
|
||||
"schemaVersion": 39,
|
||||
"tags": [
|
||||
"alloy-mixin"
|
||||
],
|
||||
"templating": {
|
||||
"list": [
|
||||
{
|
||||
"current": {
|
||||
"selected": false,
|
||||
"text": "Mimir",
|
||||
"value": "mimir_ds"
|
||||
},
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "Data Source",
|
||||
"multi": false,
|
||||
"name": "datasource",
|
||||
"options": [],
|
||||
"query": "prometheus",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "datasource"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"selected": false,
|
||||
"text": "Loki",
|
||||
"value": "loki_ds"
|
||||
},
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "Loki Data Source",
|
||||
"multi": false,
|
||||
"name": "loki_datasource",
|
||||
"options": [],
|
||||
"query": "loki",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "datasource"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"isNone": true,
|
||||
"selected": false,
|
||||
"text": "None",
|
||||
"value": ""
|
||||
},
|
||||
"datasource": {
|
||||
"uid": "${datasource}"
|
||||
},
|
||||
"definition": "",
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "cluster",
|
||||
"multi": false,
|
||||
"name": "cluster",
|
||||
"options": [],
|
||||
"query": {
|
||||
"query": "label_values(alloy_component_controller_running_components, cluster)\n",
|
||||
"refId": "cluster"
|
||||
},
|
||||
"refresh": 2,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "query"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"isNone": true,
|
||||
"selected": false,
|
||||
"text": "None",
|
||||
"value": ""
|
||||
},
|
||||
"datasource": {
|
||||
"uid": "${datasource}"
|
||||
},
|
||||
"definition": "",
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "namespace",
|
||||
"multi": false,
|
||||
"name": "namespace",
|
||||
"options": [],
|
||||
"query": {
|
||||
"query": "label_values(alloy_component_controller_running_components{cluster=\"$cluster\"}, namespace)\n",
|
||||
"refId": "namespace"
|
||||
},
|
||||
"refresh": 2,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "query"
|
||||
}
|
||||
]
|
||||
},
|
||||
"time": {
|
||||
"from": "now-1h",
|
||||
"to": "now"
|
||||
},
|
||||
"timepicker": {
|
||||
"refresh_intervals": [
|
||||
"5s",
|
||||
"10s",
|
||||
"30s",
|
||||
"1m",
|
||||
"5m",
|
||||
"15m",
|
||||
"30m",
|
||||
"1h",
|
||||
"2h",
|
||||
"1d"
|
||||
],
|
||||
"time_options": [
|
||||
"5m",
|
||||
"15m",
|
||||
"1h",
|
||||
"6h",
|
||||
"12h",
|
||||
"24h",
|
||||
"2d",
|
||||
"7d",
|
||||
"30d",
|
||||
"90d"
|
||||
]
|
||||
},
|
||||
"timezone": "",
|
||||
"title": "Alloy / Controller",
|
||||
"uid": "bf9f456aad7108b2c808dbd9973e386f",
|
||||
"version": 0,
|
||||
"weekStart": ""
|
||||
}
|
||||
923
charts/meta-monitoring/src/dashboards/alloy-opentelemetry.json
Normal file
923
charts/meta-monitoring/src/dashboards/alloy-opentelemetry.json
Normal file
@ -0,0 +1,923 @@
|
||||
{
|
||||
"annotations": {
|
||||
"list": [
|
||||
{
|
||||
"builtIn": 1,
|
||||
"datasource": {
|
||||
"type": "grafana",
|
||||
"uid": "-- Grafana --"
|
||||
},
|
||||
"enable": true,
|
||||
"hide": true,
|
||||
"iconColor": "rgba(0, 211, 255, 1)",
|
||||
"name": "Annotations & Alerts",
|
||||
"type": "dashboard"
|
||||
}
|
||||
]
|
||||
},
|
||||
"editable": true,
|
||||
"fiscalYearStartMonth": 0,
|
||||
"graphTooltip": 1,
|
||||
"id": 25,
|
||||
"links": [
|
||||
{
|
||||
"asDropdown": true,
|
||||
"icon": "external link",
|
||||
"includeVars": true,
|
||||
"keepTime": true,
|
||||
"tags": [
|
||||
"alloy-mixin"
|
||||
],
|
||||
"targetBlank": false,
|
||||
"title": "Dashboards",
|
||||
"type": "dashboards"
|
||||
}
|
||||
],
|
||||
"panels": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"gridPos": {
|
||||
"h": 1,
|
||||
"w": 24,
|
||||
"x": 0,
|
||||
"y": 0
|
||||
},
|
||||
"id": 1,
|
||||
"title": "Receivers for traces [otelcol.receiver]",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Number of spans successfully pushed into the pipeline.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 20,
|
||||
"gradientMode": "hue",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "normal"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 0,
|
||||
"y": 1
|
||||
},
|
||||
"id": 2,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "rate(receiver_accepted_spans_ratio_total{cluster=\"$cluster\", namespace=\"$namespace\", instance=~\"$instance\"}[$__rate_interval])\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{ pod }} / {{ transport }}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Accepted spans",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Number of spans that could not be pushed into the pipeline.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 20,
|
||||
"gradientMode": "hue",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "normal"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 8,
|
||||
"y": 1
|
||||
},
|
||||
"id": 3,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "rate(receiver_refused_spans_ratio_total{cluster=\"$cluster\", namespace=\"$namespace\", instance=~\"$instance\"}[$__rate_interval])\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{ pod }} / {{ transport }}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Refused spans",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "The duration of inbound RPCs.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
}
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 16,
|
||||
"y": 1
|
||||
},
|
||||
"id": 4,
|
||||
"maxDataPoints": 30,
|
||||
"options": {
|
||||
"calculate": false,
|
||||
"cellGap": 1,
|
||||
"color": {
|
||||
"exponent": 0.5,
|
||||
"fill": "dark-orange",
|
||||
"mode": "scheme",
|
||||
"reverse": false,
|
||||
"scale": "exponential",
|
||||
"scheme": "Oranges",
|
||||
"steps": 65
|
||||
},
|
||||
"exemplars": {
|
||||
"color": "rgba(255,0,255,0.7)"
|
||||
},
|
||||
"filterValues": {
|
||||
"le": 1e-9
|
||||
},
|
||||
"legend": {
|
||||
"show": true
|
||||
},
|
||||
"rowsFrame": {
|
||||
"layout": "auto"
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"showColorScale": false,
|
||||
"yHistogram": true
|
||||
},
|
||||
"yAxis": {
|
||||
"axisPlacement": "left",
|
||||
"reverse": false,
|
||||
"unit": "ms"
|
||||
}
|
||||
},
|
||||
"pluginVersion": "10.4.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum by (le) (increase(rpc_server_duration_milliseconds_bucket{cluster=\"$cluster\", namespace=\"$namespace\", instance=~\"$instance\", rpc_service=\"opentelemetry.proto.collector.trace.v1.TraceService\"}[$__rate_interval]))",
|
||||
"format": "heatmap",
|
||||
"instant": false,
|
||||
"legendFormat": "{{le}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "RPC server duration",
|
||||
"type": "heatmap"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"gridPos": {
|
||||
"h": 1,
|
||||
"w": 24,
|
||||
"x": 0,
|
||||
"y": 11
|
||||
},
|
||||
"id": 5,
|
||||
"title": "Batching of logs, metrics, and traces [otelcol.processor.batch]",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Number of spans, metric datapoints, or log lines in a batch\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"custom": {
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
}
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 0,
|
||||
"y": 12
|
||||
},
|
||||
"id": 6,
|
||||
"maxDataPoints": 30,
|
||||
"options": {
|
||||
"calculate": false,
|
||||
"cellGap": 1,
|
||||
"color": {
|
||||
"exponent": 0.5,
|
||||
"fill": "dark-orange",
|
||||
"mode": "scheme",
|
||||
"reverse": false,
|
||||
"scale": "exponential",
|
||||
"scheme": "Oranges",
|
||||
"steps": 65
|
||||
},
|
||||
"exemplars": {
|
||||
"color": "rgba(255,0,255,0.7)"
|
||||
},
|
||||
"filterValues": {
|
||||
"le": 1e-9
|
||||
},
|
||||
"legend": {
|
||||
"show": true
|
||||
},
|
||||
"rowsFrame": {
|
||||
"layout": "auto"
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"showColorScale": false,
|
||||
"yHistogram": true
|
||||
},
|
||||
"yAxis": {
|
||||
"axisPlacement": "left",
|
||||
"reverse": false,
|
||||
"unit": "short"
|
||||
}
|
||||
},
|
||||
"pluginVersion": "10.4.2",
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "sum by (le) (increase(processor_batch_batch_send_size_ratio_bucket{cluster=\"$cluster\", namespace=\"$namespace\", instance=~\"$instance\"}[$__rate_interval]))",
|
||||
"format": "heatmap",
|
||||
"instant": false,
|
||||
"legendFormat": "{{le}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Number of units in the batch",
|
||||
"type": "heatmap"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Number of distinct metadata value combinations being processed\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 0,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "none"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 8,
|
||||
"y": 12
|
||||
},
|
||||
"id": 7,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "processor_batch_metadata_cardinality_ratio{cluster=\"$cluster\", namespace=\"$namespace\", instance=~\"$instance\"}\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{ pod }}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Distinct metadata values",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Number of times the batch was sent due to a timeout trigger\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 0,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "none"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 16,
|
||||
"y": 12
|
||||
},
|
||||
"id": 8,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "rate(processor_batch_timeout_trigger_send_ratio_total{cluster=\"$cluster\", namespace=\"$namespace\", instance=~\"$instance\"}[$__rate_interval])\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{ pod }}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Timeout trigger",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"gridPos": {
|
||||
"h": 1,
|
||||
"w": 24,
|
||||
"x": 0,
|
||||
"y": 22
|
||||
},
|
||||
"id": 9,
|
||||
"title": "Exporters for traces [otelcol.exporter]",
|
||||
"type": "row"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Number of spans successfully sent to destination.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 20,
|
||||
"gradientMode": "hue",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "normal"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 0,
|
||||
"y": 23
|
||||
},
|
||||
"id": 10,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "rate(exporter_sent_spans_ratio_total{cluster=\"$cluster\", namespace=\"$namespace\", instance=~\"$instance\"}[$__rate_interval])\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{ pod }}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Exported sent spans",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Number of spans in failed attempts to send to destination.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 20,
|
||||
"gradientMode": "hue",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "normal"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
}
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 10,
|
||||
"w": 8,
|
||||
"x": 8,
|
||||
"y": 23
|
||||
},
|
||||
"id": 11,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "rate(exporter_send_failed_spans_ratio_total{cluster=\"$cluster\", namespace=\"$namespace\", instance=~\"$instance\"}[$__rate_interval])\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{ pod }}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Exported failed spans",
|
||||
"type": "timeseries"
|
||||
}
|
||||
],
|
||||
"refresh": "10s",
|
||||
"schemaVersion": 39,
|
||||
"tags": [
|
||||
"alloy-mixin"
|
||||
],
|
||||
"templating": {
|
||||
"list": [
|
||||
{
|
||||
"current": {
|
||||
"selected": false,
|
||||
"text": "Mimir",
|
||||
"value": "mimir_ds"
|
||||
},
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "Data Source",
|
||||
"multi": false,
|
||||
"name": "datasource",
|
||||
"options": [],
|
||||
"query": "prometheus",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "datasource"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"selected": false,
|
||||
"text": "Loki",
|
||||
"value": "loki_ds"
|
||||
},
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "Loki Data Source",
|
||||
"multi": false,
|
||||
"name": "loki_datasource",
|
||||
"options": [],
|
||||
"query": "loki",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "datasource"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"isNone": true,
|
||||
"selected": false,
|
||||
"text": "None",
|
||||
"value": ""
|
||||
},
|
||||
"datasource": {
|
||||
"uid": "${datasource}"
|
||||
},
|
||||
"definition": "",
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "cluster",
|
||||
"multi": false,
|
||||
"name": "cluster",
|
||||
"options": [],
|
||||
"query": {
|
||||
"query": "label_values(alloy_component_controller_running_components, cluster)\n",
|
||||
"refId": "cluster"
|
||||
},
|
||||
"refresh": 2,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "query"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"isNone": true,
|
||||
"selected": false,
|
||||
"text": "None",
|
||||
"value": ""
|
||||
},
|
||||
"datasource": {
|
||||
"uid": "${datasource}"
|
||||
},
|
||||
"definition": "",
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "namespace",
|
||||
"multi": false,
|
||||
"name": "namespace",
|
||||
"options": [],
|
||||
"query": {
|
||||
"query": "label_values(alloy_component_controller_running_components{cluster=\"$cluster\"}, namespace)\n",
|
||||
"refId": "namespace"
|
||||
},
|
||||
"refresh": 2,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "query"
|
||||
},
|
||||
{
|
||||
"allValue": ".*",
|
||||
"current": {
|
||||
"selected": false,
|
||||
"text": "All",
|
||||
"value": "$__all"
|
||||
},
|
||||
"datasource": {
|
||||
"uid": "${datasource}"
|
||||
},
|
||||
"definition": "",
|
||||
"hide": 0,
|
||||
"includeAll": true,
|
||||
"label": "instance",
|
||||
"multi": true,
|
||||
"name": "instance",
|
||||
"options": [],
|
||||
"query": {
|
||||
"query": "label_values(alloy_component_controller_running_components{cluster=\"$cluster\", namespace=\"$namespace\"}, instance)\n",
|
||||
"refId": "instance"
|
||||
},
|
||||
"refresh": 2,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "query"
|
||||
}
|
||||
]
|
||||
},
|
||||
"time": {
|
||||
"from": "now-1h",
|
||||
"to": "now"
|
||||
},
|
||||
"timepicker": {
|
||||
"refresh_intervals": [
|
||||
"5s",
|
||||
"10s",
|
||||
"30s",
|
||||
"1m",
|
||||
"5m",
|
||||
"15m",
|
||||
"30m",
|
||||
"1h",
|
||||
"2h",
|
||||
"1d"
|
||||
],
|
||||
"time_options": [
|
||||
"5m",
|
||||
"15m",
|
||||
"1h",
|
||||
"6h",
|
||||
"12h",
|
||||
"24h",
|
||||
"2d",
|
||||
"7d",
|
||||
"30d",
|
||||
"90d"
|
||||
]
|
||||
},
|
||||
"timezone": "",
|
||||
"title": "Alloy / OpenTelemetry",
|
||||
"uid": "9b6d37c8603e19e8922133984faad93d",
|
||||
"version": 0,
|
||||
"weekStart": ""
|
||||
}
|
||||
1337
charts/meta-monitoring/src/dashboards/alloy-prometheus.json
Normal file
1337
charts/meta-monitoring/src/dashboards/alloy-prometheus.json
Normal file
File diff suppressed because it is too large
Load Diff
840
charts/meta-monitoring/src/dashboards/alloy-resources.json
Normal file
840
charts/meta-monitoring/src/dashboards/alloy-resources.json
Normal file
@ -0,0 +1,840 @@
|
||||
{
|
||||
"annotations": {
|
||||
"list": [
|
||||
{
|
||||
"builtIn": 1,
|
||||
"datasource": {
|
||||
"type": "grafana",
|
||||
"uid": "-- Grafana --"
|
||||
},
|
||||
"enable": true,
|
||||
"hide": true,
|
||||
"iconColor": "rgba(0, 211, 255, 1)",
|
||||
"name": "Annotations & Alerts",
|
||||
"type": "dashboard"
|
||||
},
|
||||
{
|
||||
"datasource": "$loki_datasource",
|
||||
"enable": true,
|
||||
"expr": "{cluster=\"$cluster\", container=\"kube-diff-logger\"} | json | namespace_extracted=\"alloy\" | name_extracted=~\"alloy.*\"",
|
||||
"iconColor": "rgba(0, 211, 255, 1)",
|
||||
"instant": false,
|
||||
"name": "Deployments",
|
||||
"titleFormat": "{{cluster}}/{{namespace}}"
|
||||
}
|
||||
]
|
||||
},
|
||||
"editable": true,
|
||||
"fiscalYearStartMonth": 0,
|
||||
"graphTooltip": 1,
|
||||
"id": 26,
|
||||
"links": [
|
||||
{
|
||||
"asDropdown": true,
|
||||
"icon": "external link",
|
||||
"includeVars": true,
|
||||
"keepTime": true,
|
||||
"tags": [
|
||||
"alloy-mixin"
|
||||
],
|
||||
"targetBlank": false,
|
||||
"title": "Dashboards",
|
||||
"type": "dashboards"
|
||||
}
|
||||
],
|
||||
"panels": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "CPU usage of the Alloy process relative to 1 CPU core.\n\nFor example, 100% means using one entire CPU core.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 0,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "none"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "percentunit"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 0,
|
||||
"y": 0
|
||||
},
|
||||
"id": 1,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "rate(alloy_resources_process_cpu_seconds_total{cluster=\"$cluster\",namespace=\"$namespace\",instance=~\"$instance\"}[$__rate_interval])",
|
||||
"instant": false,
|
||||
"legendFormat": "{{instance}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "CPU usage",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Resident memory size of the Alloy process.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 0,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "none"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "decbytes"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 12,
|
||||
"y": 0
|
||||
},
|
||||
"id": 2,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "alloy_resources_process_resident_memory_bytes{cluster=\"$cluster\",namespace=\"$namespace\",instance=~\"$instance\"}",
|
||||
"instant": false,
|
||||
"legendFormat": "{{instance}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Memory (RSS)",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Rate at which the Alloy process performs garbage collections.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "points",
|
||||
"fillOpacity": 0,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 3,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "none"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "ops"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 8,
|
||||
"x": 0,
|
||||
"y": 8
|
||||
},
|
||||
"id": 3,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "rate(go_gc_duration_seconds_count{cluster=\"$cluster\",namespace=\"$namespace\",instance=~\"$instance\"}[5m])\nand on(instance)\nalloy_build_info{cluster=\"$cluster\",namespace=\"$namespace\",instance=~\"$instance\"}\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{instance}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Garbage collections",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Number of goroutines which are running in parallel. An infinitely\ngrowing number of these indicates a goroutine leak.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 0,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "none"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "none"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 8,
|
||||
"x": 8,
|
||||
"y": 8
|
||||
},
|
||||
"id": 4,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "go_goroutines{cluster=\"$cluster\",namespace=\"$namespace\",instance=~\"$instance\"}\nand on(instance)\nalloy_build_info{cluster=\"$cluster\",namespace=\"$namespace\",instance=~\"$instance\"}\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{instance}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Goroutines",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Heap memory currently in use by the Alloy process.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 0,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "none"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "decbytes"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 8,
|
||||
"x": 16,
|
||||
"y": 8
|
||||
},
|
||||
"id": 5,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "go_memstats_heap_inuse_bytes{cluster=\"$cluster\",namespace=\"$namespace\",instance=~\"$instance\"}\nand on(instance)\nalloy_build_info{cluster=\"$cluster\",namespace=\"$namespace\",instance=~\"$instance\"}\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{instance}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Memory (heap inuse)",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Rate of data received across all network interfaces for the machine\nAlloy is running on.\n\nData shown here is across all running processes and not exclusive to\nthe running Alloy process.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 30,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "normal"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "Bps"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 0,
|
||||
"y": 16
|
||||
},
|
||||
"id": 6,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "rate(alloy_resources_machine_rx_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\",instance=~\"$instance\"}[$__rate_interval])\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{instance}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Network receive bandwidth",
|
||||
"type": "timeseries"
|
||||
},
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"description": "Rate of data sent across all network interfaces for the machine\nAlloy is running on.\n\nData shown here is across all running processes and not exclusive to\nthe running Alloy process.\n",
|
||||
"fieldConfig": {
|
||||
"defaults": {
|
||||
"color": {
|
||||
"mode": "palette-classic"
|
||||
},
|
||||
"custom": {
|
||||
"axisBorderShow": false,
|
||||
"axisCenteredZero": false,
|
||||
"axisColorMode": "text",
|
||||
"axisLabel": "",
|
||||
"axisPlacement": "auto",
|
||||
"barAlignment": 0,
|
||||
"drawStyle": "line",
|
||||
"fillOpacity": 30,
|
||||
"gradientMode": "none",
|
||||
"hideFrom": {
|
||||
"legend": false,
|
||||
"tooltip": false,
|
||||
"viz": false
|
||||
},
|
||||
"insertNulls": false,
|
||||
"lineInterpolation": "linear",
|
||||
"lineWidth": 1,
|
||||
"pointSize": 5,
|
||||
"scaleDistribution": {
|
||||
"type": "linear"
|
||||
},
|
||||
"showPoints": "auto",
|
||||
"spanNulls": false,
|
||||
"stacking": {
|
||||
"group": "A",
|
||||
"mode": "normal"
|
||||
},
|
||||
"thresholdsStyle": {
|
||||
"mode": "off"
|
||||
}
|
||||
},
|
||||
"mappings": [],
|
||||
"thresholds": {
|
||||
"mode": "absolute",
|
||||
"steps": [
|
||||
{
|
||||
"color": "green",
|
||||
"value": null
|
||||
},
|
||||
{
|
||||
"color": "red",
|
||||
"value": 80
|
||||
}
|
||||
]
|
||||
},
|
||||
"unit": "Bps"
|
||||
},
|
||||
"overrides": []
|
||||
},
|
||||
"gridPos": {
|
||||
"h": 8,
|
||||
"w": 12,
|
||||
"x": 12,
|
||||
"y": 16
|
||||
},
|
||||
"id": 7,
|
||||
"options": {
|
||||
"legend": {
|
||||
"calcs": [],
|
||||
"displayMode": "list",
|
||||
"placement": "bottom",
|
||||
"showLegend": true
|
||||
},
|
||||
"tooltip": {
|
||||
"mode": "single",
|
||||
"sort": "none"
|
||||
}
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"datasource": "${datasource}",
|
||||
"expr": "rate(alloy_resources_machine_tx_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\",instance=~\"$instance\"}[$__rate_interval])\n",
|
||||
"instant": false,
|
||||
"legendFormat": "{{instance}}",
|
||||
"range": true,
|
||||
"refId": "A"
|
||||
}
|
||||
],
|
||||
"title": "Network send bandwidth",
|
||||
"type": "timeseries"
|
||||
}
|
||||
],
|
||||
"refresh": "10s",
|
||||
"schemaVersion": 39,
|
||||
"tags": [
|
||||
"alloy-mixin"
|
||||
],
|
||||
"templating": {
|
||||
"list": [
|
||||
{
|
||||
"current": {
|
||||
"selected": false,
|
||||
"text": "Mimir",
|
||||
"value": "mimir_ds"
|
||||
},
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "Data Source",
|
||||
"multi": false,
|
||||
"name": "datasource",
|
||||
"options": [],
|
||||
"query": "prometheus",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "datasource"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"selected": false,
|
||||
"text": "Loki",
|
||||
"value": "loki_ds"
|
||||
},
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "Loki Data Source",
|
||||
"multi": false,
|
||||
"name": "loki_datasource",
|
||||
"options": [],
|
||||
"query": "loki",
|
||||
"refresh": 1,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "datasource"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"isNone": true,
|
||||
"selected": false,
|
||||
"text": "None",
|
||||
"value": ""
|
||||
},
|
||||
"datasource": {
|
||||
"uid": "${datasource}"
|
||||
},
|
||||
"definition": "",
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "cluster",
|
||||
"multi": false,
|
||||
"name": "cluster",
|
||||
"options": [],
|
||||
"query": {
|
||||
"query": "label_values(alloy_component_controller_running_components, cluster)\n",
|
||||
"refId": "cluster"
|
||||
},
|
||||
"refresh": 2,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "query"
|
||||
},
|
||||
{
|
||||
"current": {
|
||||
"isNone": true,
|
||||
"selected": false,
|
||||
"text": "None",
|
||||
"value": ""
|
||||
},
|
||||
"datasource": {
|
||||
"uid": "${datasource}"
|
||||
},
|
||||
"definition": "",
|
||||
"hide": 0,
|
||||
"includeAll": false,
|
||||
"label": "namespace",
|
||||
"multi": false,
|
||||
"name": "namespace",
|
||||
"options": [],
|
||||
"query": {
|
||||
"query": "label_values(alloy_component_controller_running_components{cluster=\"$cluster\"}, namespace)\n",
|
||||
"refId": "namespace"
|
||||
},
|
||||
"refresh": 2,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "query"
|
||||
},
|
||||
{
|
||||
"allValue": ".*",
|
||||
"current": {
|
||||
"selected": false,
|
||||
"text": "All",
|
||||
"value": "$__all"
|
||||
},
|
||||
"datasource": {
|
||||
"uid": "${datasource}"
|
||||
},
|
||||
"definition": "",
|
||||
"hide": 0,
|
||||
"includeAll": true,
|
||||
"label": "instance",
|
||||
"multi": true,
|
||||
"name": "instance",
|
||||
"options": [],
|
||||
"query": {
|
||||
"query": "label_values(alloy_component_controller_running_components{cluster=\"$cluster\", namespace=\"$namespace\"}, instance)\n",
|
||||
"refId": "instance"
|
||||
},
|
||||
"refresh": 2,
|
||||
"regex": "",
|
||||
"skipUrlSync": false,
|
||||
"sort": 2,
|
||||
"type": "query"
|
||||
}
|
||||
]
|
||||
},
|
||||
"time": {
|
||||
"from": "now-1h",
|
||||
"to": "now"
|
||||
},
|
||||
"timepicker": {
|
||||
"refresh_intervals": [
|
||||
"5s",
|
||||
"10s",
|
||||
"30s",
|
||||
"1m",
|
||||
"5m",
|
||||
"15m",
|
||||
"30m",
|
||||
"1h",
|
||||
"2h",
|
||||
"1d"
|
||||
],
|
||||
"time_options": [
|
||||
"5m",
|
||||
"15m",
|
||||
"1h",
|
||||
"6h",
|
||||
"12h",
|
||||
"24h",
|
||||
"2d",
|
||||
"7d",
|
||||
"30d",
|
||||
"90d"
|
||||
]
|
||||
},
|
||||
"timezone": "",
|
||||
"title": "Alloy / Resources",
|
||||
"uid": "d6a8574c31f3d7cb8f1345ec84d15a67",
|
||||
"version": 0,
|
||||
"weekStart": ""
|
||||
}
|
||||
@ -1824,7 +1824,7 @@
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", pod=~\"(.*distributor.*|(loki|enterprise-logs)-write)\"}[$__rate_interval]))",
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", pod=~\"(.*distributor.*|(loki|enterprise-logs)-write.*|$namespace-[0-9]+)\"}[$__rate_interval]))",
|
||||
"intervalFactor": 3,
|
||||
"legendFormat": "{{pod}}",
|
||||
"refId": "A"
|
||||
@ -1921,7 +1921,7 @@
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "go_memstats_heap_inuse_bytes{cluster=\"$cluster\", namespace=\"$namespace\", job=~\"(.*/distributor|(loki|enterprise-logs)-write|.*/loki)\"}",
|
||||
"expr": "go_memstats_heap_inuse_bytes{cluster=\"$cluster\", namespace=\"$namespace\", job=~\"(.*/.*distributor|$namespace/(loki|enterprise-logs)-write|.*/loki|$namespace/loki-single-binary)\"}",
|
||||
"instant": false,
|
||||
"intervalFactor": 3,
|
||||
"legendFormat": "{{pod}}",
|
||||
@ -2525,7 +2525,7 @@
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", pod=~\"(.*ingester.*|(loki|enterprise-logs)-write|loki-single-binary)\"}[$__rate_interval]))",
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", pod=~\"(.*ingester.*|(loki|enterprise-logs)-write.*|loki-single-binary|$namespace-[0-9]+)\"}[$__rate_interval]))",
|
||||
"intervalFactor": 3,
|
||||
"legendFormat": "{{pod}}",
|
||||
"refId": "A"
|
||||
@ -2622,7 +2622,7 @@
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "go_memstats_heap_inuse_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"(.*ingester.*|(loki|enterprise-logs)-write|loki-single-binary)\"}",
|
||||
"expr": "go_memstats_heap_inuse_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"(.*ingester.*|(loki|enterprise-logs)-write.*|loki-single-binary|$namespace-[0-9]+)\"}",
|
||||
"instant": false,
|
||||
"intervalFactor": 3,
|
||||
"legendFormat": "{{pod}}",
|
||||
@ -3308,7 +3308,7 @@
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(reason) (rate(loki_ingester_chunks_flushed_total{cluster=~\"$cluster\",job=~\"$namespace/.*ingester.*\", namespace=~\"$namespace\"}[$__rate_interval])) / ignoring(reason) group_left sum(rate(loki_ingester_chunks_flushed_total{cluster=~\"$cluster\",job=~\"$namespace/.*ingester.*\", namespace=~\"$namespace\"}[$__rate_interval]))",
|
||||
"expr": "sum by(reason) (rate(loki_ingester_chunks_flushed_total{cluster=~\"$cluster\",job=~\"($namespace)/(.*ingester.*|(loki|enterprise-logs)-write|loki-single-binary)\", namespace=~\"$namespace\"}[$__rate_interval])) / ignoring(reason) group_left sum(rate(loki_ingester_chunks_flushed_total{cluster=~\"$cluster\",job=~\"($namespace)/(.*ingester.*|(loki|enterprise-logs)-write|loki-single-binary)\", namespace=~\"$namespace\"}[$__rate_interval]))",
|
||||
"interval": "",
|
||||
"legendFormat": "{{ reason }}"
|
||||
}
|
||||
@ -3388,7 +3388,7 @@
|
||||
"reverseYBuckets": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by (le) (rate(loki_ingester_chunk_utilization_bucket{cluster=\"$cluster\", job=~\"($namespace)/(ingester|(loki|enterprise-logs)-write|loki-single-binary)\"}[$__rate_interval]))",
|
||||
"expr": "sum by (le) (rate(loki_ingester_chunk_utilization_bucket{cluster=\"$cluster\", job=~\"($namespace)/(.*ingester|(loki|enterprise-logs)-write|loki-single-binary)\"}[$__rate_interval]))",
|
||||
"format": "heatmap",
|
||||
"instant": false,
|
||||
"interval": "",
|
||||
@ -3481,7 +3481,7 @@
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", pod=~\"(.*querier.*|(loki|enterprise-logs)-read|loki-single-binary)\"}[$__rate_interval]))",
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", pod=~\"(.*querier.*|(loki|enterprise-logs)-read.*|loki-single-binary|$namespace-[0-9]+)\"}[$__rate_interval]))",
|
||||
"intervalFactor": 3,
|
||||
"legendFormat": "{{pod}}",
|
||||
"refId": "A"
|
||||
@ -3578,7 +3578,7 @@
|
||||
"steppedLine": false,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "go_memstats_heap_inuse_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"(.*querier.*|(loki|enterprise-logs)-read|.*loki-single-binary)\"}",
|
||||
"expr": "go_memstats_heap_inuse_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"(.*querier.*|(loki|enterprise-logs)-read.*|.*loki-single-binary|$namespace-[0-9]+)\"}",
|
||||
"instant": false,
|
||||
"intervalFactor": 3,
|
||||
"legendFormat": "{{pod}}",
|
||||
|
||||
@ -104,19 +104,19 @@
|
||||
"span": 4,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend\"}[$__rate_interval]))",
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"}[$__rate_interval]))",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend\", resource=\"cpu\"} > 0)",
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\", resource=\"cpu\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "request",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend\"})",
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -206,19 +206,19 @@
|
||||
"span": 4,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend\"})",
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend\", resource=\"memory\"} > 0)",
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\", resource=\"memory\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "request",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend\"} > 0)",
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -269,7 +269,7 @@
|
||||
"span": 4,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/.*query-frontend\"})",
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/(.*query-frontend|loki-read|loki-single-binary)\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
@ -371,19 +371,19 @@
|
||||
"span": 4,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler\"}[$__rate_interval]))",
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler|loki\", pod=~\"query-scheduler|loki-read-.*|$namespace-[0-9]*\"}[$__rate_interval]))",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler\", resource=\"cpu\"} > 0)",
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler|loki\", pod=~\"query-scheduler|loki-read-.*|$namespace-[0-9]*\", resource=\"cpu\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "request",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler\"})",
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler|loki\", pod=~\"query-scheduler|loki-read-.*|$namespace-[0-9]*\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler|loki\", pod=~\"query-scheduler|loki-read-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -473,19 +473,19 @@
|
||||
"span": 4,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler\"})",
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler|loki\", pod=~\"query-scheduler|loki-read-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler\", resource=\"memory\"} > 0)",
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler|loki\", pod=~\"query-scheduler|loki-read-.*|$namespace-[0-9]*\", resource=\"memory\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "request",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler\"} > 0)",
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-scheduler|loki\", pod=~\"query-scheduler|loki-read-.*|$namespace-[0-9]*\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -536,7 +536,7 @@
|
||||
"span": 4,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/.*query-scheduler\"})",
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/(.*query-scheduler|loki-read|loki-single-binary)\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
@ -638,19 +638,19 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier\"}[$__rate_interval]))",
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"}[$__rate_interval]))",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier\", resource=\"cpu\"} > 0)",
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\", resource=\"cpu\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "request",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier\"})",
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -740,19 +740,19 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier\"})",
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier\", resource=\"memory\"} > 0)",
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\", resource=\"memory\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "request",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier\"} > 0)",
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"querier|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -803,7 +803,7 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/.*querier\"})",
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/(.*querier|loki-read|loki-single-binary)\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
@ -854,7 +854,7 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(instance, pod, device) (rate(node_disk_written_bytes_total[$__rate_interval])) + ignoring(pod) group_right() (label_replace(count by(instance, pod, device) (container_fs_writes_bytes_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=\"querier\", device!~\".*sda.*\"}), \"device\", \"$1\", \"device\", \"/dev/(.*)\") * 0)\n",
|
||||
"expr": "sum by(instance, pod, device) (rate(node_disk_written_bytes_total[$__rate_interval])) + ignoring(pod) group_right() (label_replace(count by(instance, pod, device) (container_fs_writes_bytes_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\", device!~\".*sda.*\"}), \"device\", \"$1\", \"device\", \"/dev/(.*)\") * 0)\n",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}} - {{device}}",
|
||||
"legendLink": null
|
||||
@ -902,7 +902,7 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(instance, pod, device) (rate(node_disk_read_bytes_total[$__rate_interval])) + ignoring(pod) group_right() (label_replace(count by(instance, pod, device) (container_fs_writes_bytes_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=\"querier\", device!~\".*sda.*\"}), \"device\", \"$1\", \"device\", \"/dev/(.*)\") * 0)\n",
|
||||
"expr": "sum by(instance, pod, device) (rate(node_disk_read_bytes_total[$__rate_interval])) + ignoring(pod) group_right() (label_replace(count by(instance, pod, device) (container_fs_writes_bytes_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"query-frontend|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\", device!~\".*sda.*\"}), \"device\", \"$1\", \"device\", \"/dev/(.*)\") * 0)\n",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}} - {{device}}",
|
||||
"legendLink": null
|
||||
@ -1462,19 +1462,19 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway\"}[$__rate_interval]))",
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"}[$__rate_interval]))",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway\", resource=\"cpu\"} > 0)",
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\", resource=\"cpu\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "request",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway\"})",
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -1564,19 +1564,19 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway\"})",
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway\", resource=\"memory\"} > 0)",
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\", resource=\"memory\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "request",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway\"} > 0)",
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -1627,7 +1627,7 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/.*bloom-gateway\"})",
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/(.*bloom-gateway|loki-read|loki-single-binary)\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
@ -1678,7 +1678,7 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(instance, pod, device) (rate(node_disk_written_bytes_total[$__rate_interval])) + ignoring(pod) group_right() (label_replace(count by(instance, pod, device) (container_fs_writes_bytes_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=\"bloom-gateway\", device!~\".*sda.*\"}), \"device\", \"$1\", \"device\", \"/dev/(.*)\") * 0)\n",
|
||||
"expr": "sum by(instance, pod, device) (rate(node_disk_written_bytes_total[$__rate_interval])) + ignoring(pod) group_right() (label_replace(count by(instance, pod, device) (container_fs_writes_bytes_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\", device!~\".*sda.*\"}), \"device\", \"$1\", \"device\", \"/dev/(.*)\") * 0)\n",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}} - {{device}}",
|
||||
"legendLink": null
|
||||
@ -1726,7 +1726,7 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(instance, pod, device) (rate(node_disk_read_bytes_total[$__rate_interval])) + ignoring(pod) group_right() (label_replace(count by(instance, pod, device) (container_fs_writes_bytes_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=\"bloom-gateway\", device!~\".*sda.*\"}), \"device\", \"$1\", \"device\", \"/dev/(.*)\") * 0)\n",
|
||||
"expr": "sum by(instance, pod, device) (rate(node_disk_read_bytes_total[$__rate_interval])) + ignoring(pod) group_right() (label_replace(count by(instance, pod, device) (container_fs_writes_bytes_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"bloom-gateway|loki\", pod=~\"query-frontend|loki-read-.*|$namespace-[0-9]*\", device!~\".*sda.*\"}), \"device\", \"$1\", \"device\", \"/dev/(.*)\") * 0)\n",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}} - {{device}}",
|
||||
"legendLink": null
|
||||
@ -2189,19 +2189,19 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler\"}[$__rate_interval]))",
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler|loki\", pod=~\"ruler|loki-backend-.*|$namespace-[0-9]*\"}[$__rate_interval]))",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler\", resource=\"cpu\"} > 0)",
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler|loki\", pod=~\"ruler|loki-backend-.*|$namespace-[0-9]*\", resource=\"cpu\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "request",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler\"})",
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler|loki\", pod=~\"ruler|loki-backend-.*|$namespace-[0-9]*\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler|loki\", pod=~\"ruler|loki-backend-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -2291,19 +2291,19 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler\"})",
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler|loki\", pod=~\"ruler|loki-backend-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler\", resource=\"memory\"} > 0)",
|
||||
"expr": "min(kube_pod_container_resource_requests{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler|loki\", pod=~\"ruler|loki-backend-.*|$namespace-[0-9]*\", resource=\"memory\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "request",
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler\"} > 0)",
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"ruler|loki\", pod=~\"ruler|loki-backend-.*|$namespace-[0-9]*\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -2354,7 +2354,7 @@
|
||||
},
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/ruler\"})",
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/(.*ruler|loki-backend|loki-single-binary)\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
|
||||
@ -2449,7 +2449,7 @@
|
||||
"repeatIteration": null,
|
||||
"repeatRowId": null,
|
||||
"showTitle": true,
|
||||
"title": "TSBD Index",
|
||||
"title": "TSDB Index",
|
||||
"titleSize": "h6"
|
||||
},
|
||||
{
|
||||
|
||||
@ -104,7 +104,7 @@
|
||||
"span": 4,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"distributor\"}[$__rate_interval]))",
|
||||
"expr": "sum by(pod) (rate(container_cpu_usage_seconds_total{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"distributor|loki\", pod=~\"distributor|loki-write-.*|$namespace-[0-9]*\"}[$__rate_interval]))",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
@ -116,7 +116,7 @@
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"distributor\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"distributor\"})",
|
||||
"expr": "min(container_spec_cpu_quota{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"distributor|loki\", pod=~\"distributor|loki-write-.*|$namespace-[0-9]*\"} / container_spec_cpu_period{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"distributor|loki\", pod=~\"distributor|loki-write-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -206,7 +206,7 @@
|
||||
"span": 4,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"distributor\"})",
|
||||
"expr": "max by(pod) (container_memory_working_set_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"distributor|loki\", pod=~\"distributor|loki-write-.*|$namespace-[0-9]*\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
@ -218,7 +218,7 @@
|
||||
"legendLink": null
|
||||
},
|
||||
{
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"distributor\"} > 0)",
|
||||
"expr": "min(container_spec_memory_limit_bytes{cluster=~\"$cluster\", namespace=~\"$namespace\", container=~\"distributor|loki\", pod=~\"distributor|loki-write-.*|$namespace-[0-9]*\"} > 0)",
|
||||
"format": "time_series",
|
||||
"legendFormat": "limit",
|
||||
"legendLink": null
|
||||
@ -269,7 +269,7 @@
|
||||
"span": 4,
|
||||
"targets": [
|
||||
{
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/.*distributor\"})",
|
||||
"expr": "sum by(pod) (go_memstats_heap_inuse_bytes{cluster=~\"$cluster\", job=~\"($namespace)/(.*distributor|loki-write|loki-single-binary)\"})",
|
||||
"format": "time_series",
|
||||
"legendFormat": "{{pod}}",
|
||||
"legendLink": null
|
||||
|
||||
@ -1,4 +1,5 @@
|
||||
- name: "loki_rules"
|
||||
groups:
|
||||
- name: "loki_rules"
|
||||
rules:
|
||||
- expr: "histogram_quantile(0.99, sum(rate(loki_request_duration_seconds_bucket[5m]))
|
||||
by (le, cluster, job))"
|
||||
|
||||
@ -9,6 +9,15 @@
|
||||
{{- define "agent.all_namespaces" -}}
|
||||
{{- $list := list }}
|
||||
{{- range .Values.namespacesToMonitor }}
|
||||
{{- $list = append $list (printf "\"%s\"" .) }}
|
||||
{{- end }}
|
||||
{{- $list = append $list (printf "\"%s\"" .Release.Namespace) }}
|
||||
{{- join ", " $list }}
|
||||
{{- end }}
|
||||
|
||||
{{- define "agent.all_namespaces_bar" -}}
|
||||
{{- $list := list }}
|
||||
{{- range .Values.namespacesToMonitor }}
|
||||
{{- $list = append $list (printf "%s" .) }}
|
||||
{{- end }}
|
||||
{{- $list = append $list .Release.Namespace }}
|
||||
|
||||
@ -93,7 +93,7 @@ data:
|
||||
role = "pod"
|
||||
namespaces {
|
||||
own_namespace = true
|
||||
names = [ {{ include "agent.namespaces" . }} ]
|
||||
names = [ {{ include "agent.all_namespaces" . }} ]
|
||||
}
|
||||
}
|
||||
|
||||
@ -120,9 +120,9 @@ data:
|
||||
replacement = "{{- .Values.clusterLabelValue -}}"
|
||||
}
|
||||
rule {
|
||||
source_labels = ["__meta_kubernetes_pod_container_port_number"]
|
||||
action = "drop"
|
||||
regex = "9095"
|
||||
source_labels = ["__meta_kubernetes_pod_container_port_name"]
|
||||
action = "keep"
|
||||
regex = ".*metrics.*"
|
||||
}
|
||||
}
|
||||
|
||||
@ -135,6 +135,11 @@ data:
|
||||
}
|
||||
|
||||
prometheus.relabel "filter" {
|
||||
rule {
|
||||
target_label = "cluster"
|
||||
replacement = "{{- .Values.clusterLabelValue -}}"
|
||||
}
|
||||
|
||||
rule {
|
||||
source_labels = ["__name__"]
|
||||
regex = "({{ include "agent.all_metrics" . }})"
|
||||
@ -143,140 +148,14 @@ data:
|
||||
|
||||
rule {
|
||||
source_labels = ["namespace"]
|
||||
regex = "{{ include "agent.all_namespaces" . }}"
|
||||
regex = "{{ include "agent.all_namespaces_bar" . }}"
|
||||
|
||||
action = "keep"
|
||||
}
|
||||
|
||||
forward_to = [ {{ include "agent.prometheus_write_targets" . }} ]
|
||||
}
|
||||
{{- if .Values.kubeStateMetrics.enabled }}
|
||||
|
||||
prometheus.scrape "kubeStateMetrics" {
|
||||
clustering {
|
||||
enabled = true
|
||||
}
|
||||
targets = [ { "__address__" = "{{ .Values.kubeStateMetrics.endpoint }}" } ]
|
||||
forward_to = [ prometheus.relabel.filter.receiver ]
|
||||
}
|
||||
{{- end }}
|
||||
|
||||
// cAdvisor and Kubelet metrics
|
||||
// Based on https://github.com/Chewie/loutretelecom-manifests/blob/main/manifests/addons/monitoring/config.river
|
||||
discovery.kubernetes "all_nodes" {
|
||||
role = "node"
|
||||
namespaces {
|
||||
own_namespace = true
|
||||
names = [ {{ include "agent.namespaces" . }} ]
|
||||
}
|
||||
}
|
||||
|
||||
discovery.relabel "all_nodes" {
|
||||
targets = discovery.kubernetes.all_nodes.targets
|
||||
rule {
|
||||
source_labels = ["__meta_kubernetes_node_name"]
|
||||
target_label = "node"
|
||||
}
|
||||
rule {
|
||||
source_labels = ["__meta_kubernetes_namespace"]
|
||||
target_label = "namespace"
|
||||
}
|
||||
rule {
|
||||
source_labels = ["__meta_kubernetes_pod_name"]
|
||||
target_label = "pod"
|
||||
}
|
||||
rule {
|
||||
source_labels = ["__meta_kubernetes_namespace", "__meta_kubernetes_pod_label_app_kubernetes_io_name", "__meta_kubernetes_pod_label_app_kubernetes_io_component"]
|
||||
separator = "/"
|
||||
regex = "(.*)/(.*)/(.*)"
|
||||
replacement = "${1}/${2}-${3}"
|
||||
target_label = "job"
|
||||
}
|
||||
rule {
|
||||
target_label = "cluster"
|
||||
replacement = "{{- .Values.clusterLabelValue -}}"
|
||||
}
|
||||
}
|
||||
|
||||
prometheus.scrape "cadvisor" {
|
||||
clustering {
|
||||
enabled = true
|
||||
}
|
||||
targets = discovery.relabel.all_nodes.output
|
||||
forward_to = [ prometheus.relabel.filter.receiver ]
|
||||
|
||||
metrics_path = "/metrics/cadvisor"
|
||||
scheme = "https"
|
||||
|
||||
bearer_token_file = "/var/run/secrets/kubernetes.io/serviceaccount/token"
|
||||
tls_config {
|
||||
ca_file = "/var/run/secrets/kubernetes.io/serviceaccount/ca.crt"
|
||||
}
|
||||
}
|
||||
|
||||
prometheus.scrape "kubelet" {
|
||||
clustering {
|
||||
enabled = true
|
||||
}
|
||||
targets = discovery.relabel.all_nodes.output
|
||||
forward_to = [ prometheus.relabel.filter.receiver ]
|
||||
|
||||
metrics_path = "/metrics"
|
||||
scheme = "https"
|
||||
|
||||
bearer_token_file = "/var/run/secrets/kubernetes.io/serviceaccount/token"
|
||||
tls_config {
|
||||
ca_file = "/var/run/secrets/kubernetes.io/serviceaccount/ca.crt"
|
||||
}
|
||||
}
|
||||
|
||||
prometheus.exporter.unix "promexporter" {}
|
||||
|
||||
prometheus.scrape "node_exporter" {
|
||||
clustering {
|
||||
enabled = true
|
||||
}
|
||||
targets = prometheus.exporter.unix.promexporter.targets
|
||||
forward_to = [prometheus.relabel.node_exporter.receiver]
|
||||
|
||||
job_name = "node-exporter"
|
||||
}
|
||||
|
||||
prometheus.relabel "node_exporter" {
|
||||
forward_to = [ prometheus.relabel.filter.receiver ]
|
||||
|
||||
rule {
|
||||
replacement = env("HOSTNAME")
|
||||
target_label = "nodename"
|
||||
}
|
||||
rule {
|
||||
replacement = "node-exporter"
|
||||
target_label = "job"
|
||||
}
|
||||
rule {
|
||||
source_labels = ["__meta_kubernetes_node_name"]
|
||||
target_label = "node"
|
||||
}
|
||||
rule {
|
||||
source_labels = ["__meta_kubernetes_namespace"]
|
||||
target_label = "namespace"
|
||||
}
|
||||
rule {
|
||||
source_labels = ["__meta_kubernetes_pod_name"]
|
||||
target_label = "pod"
|
||||
}
|
||||
rule {
|
||||
source_labels = ["__meta_kubernetes_namespace", "__meta_kubernetes_pod_label_app_kubernetes_io_name", "__meta_kubernetes_pod_label_app_kubernetes_io_component"]
|
||||
separator = "/"
|
||||
regex = "(.*)/(.*)/(.*)"
|
||||
replacement = "${1}/${2}-${3}"
|
||||
target_label = "job"
|
||||
}
|
||||
rule {
|
||||
target_label = "cluster"
|
||||
replacement = "{{- .Values.clusterLabelValue -}}"
|
||||
}
|
||||
}
|
||||
{{- end }}
|
||||
|
||||
{{- if or .Values.local.traces.enabled .Values.cloud.traces.enabled }}
|
||||
@ -330,7 +209,7 @@ data:
|
||||
{{- if .Values.local.logs.enabled }}
|
||||
loki.write "local" {
|
||||
endpoint {
|
||||
url = "http://{{- .Release.Namespace -}}-loki-gateway.{{- .Release.Namespace -}}.svc.cluster.local:80/loki/api/v1/push"
|
||||
url = "http://loki-write.{{- .Release.Namespace -}}.svc.cluster.local:3100/loki/api/v1/push"
|
||||
}
|
||||
}
|
||||
{{- end }}
|
||||
@ -346,7 +225,7 @@ data:
|
||||
{{- if .Values.local.traces.enabled }}
|
||||
otelcol.exporter.otlphttp "local" {
|
||||
client {
|
||||
endpoint = "http://{{- .Release.Name -}}-tempo-distributor.svc:4318"
|
||||
endpoint = "http://{{- .Release.Name -}}-tempo-distributor.{{- .Release.Namespace -}}.svc:4318"
|
||||
}
|
||||
}
|
||||
{{- end }}
|
||||
|
||||
@ -1,19 +0,0 @@
|
||||
{{- if and .Values.local.grafana.enabled .Values.dashboards.logs.enabled }}
|
||||
---
|
||||
apiVersion: v1
|
||||
kind: ConfigMap
|
||||
metadata:
|
||||
name: agent-dashboards-1
|
||||
namespace: {{ $.Release.Namespace }}
|
||||
data:
|
||||
"agent-logs-pipeline.json": |
|
||||
{{ $.Files.Get "src/dashboards/agent-logs-pipeline.json" | fromJson | toJson }}
|
||||
"agent-operational.json": |
|
||||
{{ $.Files.Get "src/dashboards/agent-operational.json" | fromJson | toJson }}
|
||||
"agent-remote-write.json": |
|
||||
{{ $.Files.Get "src/dashboards/agent-remote-write.json" | fromJson | toJson }}
|
||||
"agent-tracing-pipeline.json": |
|
||||
{{ $.Files.Get "src/dashboards/agent-tracing-pipeline.json" | fromJson | toJson }}
|
||||
"agent.json": |
|
||||
{{ $.Files.Get "src/dashboards/agent.json" | fromJson | toJson }}
|
||||
{{- end }}
|
||||
@ -0,0 +1,21 @@
|
||||
{{- if and .Values.local.grafana.enabled .Values.dashboards.logs.enabled }}
|
||||
---
|
||||
apiVersion: v1
|
||||
kind: ConfigMap
|
||||
metadata:
|
||||
name: alloy-dashboards-1
|
||||
namespace: {{ $.Release.Namespace }}
|
||||
data:
|
||||
"alloy-cluster-node.json": |
|
||||
{{ $.Files.Get "src/dashboards/alloy-cluster-node.json" | fromJson | toJson }}
|
||||
"alloy-cluster-overview.json": |
|
||||
{{ $.Files.Get "src/dashboards/alloy-cluster-overview.json" | fromJson | toJson }}
|
||||
"alloy-controller.json": |
|
||||
{{ $.Files.Get "src/dashboards/alloy-controller.json" | fromJson | toJson }}
|
||||
"alloy-opentelemetry.json": |
|
||||
{{ $.Files.Get "src/dashboards/alloy-opentelemetry.json" | fromJson | toJson }}
|
||||
"alloy-prometheus.json": |
|
||||
{{ $.Files.Get "src/dashboards/alloy-prometheus.json" | fromJson | toJson }}
|
||||
"alloy-resources.json": |
|
||||
{{ $.Files.Get "src/dashboards/alloy-resources.json" | fromJson | toJson }}
|
||||
{{- end }}
|
||||
@ -30,10 +30,10 @@ data:
|
||||
{{- end }}
|
||||
- disableDeletion: true
|
||||
editable: false
|
||||
folder: Agent
|
||||
name: agent-1
|
||||
folder: Alloy
|
||||
name: alloy-1
|
||||
options:
|
||||
path: /var/lib/grafana/dashboards/agent-1
|
||||
path: /var/lib/grafana/dashboards/alloy-1
|
||||
orgId: 1
|
||||
type: file
|
||||
{{- end }}
|
||||
@ -63,8 +63,8 @@ spec:
|
||||
- mountPath: /var/lib/grafana/dashboards/loki-2
|
||||
name: loki-dashboards-2
|
||||
{{- end }}
|
||||
- mountPath: /var/lib/grafana/dashboards/agent-1
|
||||
name: agent-dashboards-1
|
||||
- mountPath: /var/lib/grafana/dashboards/alloy-1
|
||||
name: alloy-dashboards-1
|
||||
volumes:
|
||||
- name: grafana-pv
|
||||
persistentVolumeClaim:
|
||||
@ -83,7 +83,7 @@ spec:
|
||||
configMap:
|
||||
name: loki-dashboards-2
|
||||
{{- end }}
|
||||
- name: agent-dashboards-1
|
||||
- name: alloy-dashboards-1
|
||||
configMap:
|
||||
name: agent-dashboards-1
|
||||
name: alloy-dashboards-1
|
||||
{{- end }}
|
||||
|
||||
@ -12,8 +12,6 @@ data:
|
||||
{{ $.Files.Get "src/dashboards/loki-deletion.json" | fromJson | toJson }}
|
||||
"loki-logs.json": |
|
||||
{{ $.Files.Get "src/dashboards/loki-logs.json" | fromJson | toJson }}
|
||||
"loki-mixin-recording-rules.json": |
|
||||
{{ $.Files.Get "src/dashboards/loki-mixin-recording-rules.json" | fromJson | toJson }}
|
||||
"loki-operational.json": |
|
||||
{{ $.Files.Get "src/dashboards/loki-operational.json" | fromJson | toJson }}
|
||||
{{- end }}
|
||||
@ -34,10 +34,6 @@
|
||||
{{- end -}}
|
||||
{{- end -}}
|
||||
|
||||
{{- if empty .Values.namespacesToMonitor -}}
|
||||
{{- fail "No namespaces have been specified in namespacesToMonitor" -}}
|
||||
{{- end -}}
|
||||
|
||||
{{- if empty .Values.metrics.retain -}}
|
||||
{{- fail "All metrics will be collected, please specify some in metrics.retain" -}}
|
||||
{{- end -}}
|
||||
|
||||
@ -1,8 +1,9 @@
|
||||
# Specify the namespaces to monitor here
|
||||
namespacesToMonitor:
|
||||
- loki
|
||||
# By default the chart will monitor the namespace it is installed in
|
||||
# namespacesToMonitor:
|
||||
# - loki
|
||||
# The name of the cluster where this will be installed
|
||||
clusterLabelValue: "meta-monitoring"
|
||||
clusterLabelValue: "meta"
|
||||
# Set to true to write logs, metrics or traces to Grafana Cloud
|
||||
# The secrets have to be created first
|
||||
cloud:
|
||||
@ -28,7 +29,7 @@ local:
|
||||
minio:
|
||||
enabled: false # This should be set to true if any of the previous is enabled
|
||||
grafana:
|
||||
version: 10.4.2
|
||||
version: 11.4.3
|
||||
# Gateway ingress configuration
|
||||
ingress:
|
||||
# -- Specifies whether an ingress for the gateway should be created
|
||||
@ -45,7 +46,12 @@ grafana:
|
||||
paths:
|
||||
- path: /
|
||||
# -- pathType (e.g. ImplementationSpecific, Prefix, .. etc.) might also be required by some Ingress Controllers
|
||||
# pathType: Prefix
|
||||
pathType: Prefix
|
||||
# backend:
|
||||
# service:
|
||||
# name: TODO
|
||||
# port:
|
||||
# number: TODO
|
||||
# -- TLS configuration for the gateway ingress. Hosts passed through the `tpl` function to allow templating
|
||||
#tls:
|
||||
# - secretName: grafana-tls
|
||||
@ -55,15 +61,17 @@ logs:
|
||||
# Adding regexes here will add a stage.replace block for logs. For more information see
|
||||
# https://grafana.com/docs/agent/latest/flow/reference/components/loki.process/#stagereplace-block
|
||||
piiRegexes: null # This example replaces the word after password with *****
|
||||
# - expression: "password (\\\\S+)"
|
||||
# source: "" # Empty uses the log message
|
||||
# replace: "*****""
|
||||
# The lines matching these will be kept in Loki
|
||||
# - expression: "password (\\\\S+)"
|
||||
# source: "" # Empty uses the log message
|
||||
# replace: "*****""
|
||||
# The lines matching these will be kept in Loki
|
||||
retain:
|
||||
# This shows the queries
|
||||
- executing query
|
||||
- caller=metrics.go
|
||||
# This shows any errors
|
||||
- level=error
|
||||
- level=warn
|
||||
# Log lines for delete requests
|
||||
- delete request for user added
|
||||
- Started processing delete request
|
||||
@ -75,9 +83,46 @@ logs:
|
||||
metrics:
|
||||
# The list of metrics to retain for logging dashboards
|
||||
retain:
|
||||
- agent_config_last_load_success_timestamp_seconds
|
||||
- agent_config_last_load_successful
|
||||
- agent_config_load_failures_total
|
||||
- alloy_build_info
|
||||
- alloy_config_last_load_success_timestamp_seconds
|
||||
- alloy_config_last_load_successful
|
||||
- alloy_config_load_failures_total
|
||||
- alloy_component_controller_evaluating
|
||||
- alloy_component_dependencies_wait_seconds
|
||||
- alloy_component_dependencies_wait_seconds_bucket
|
||||
- alloy_component_evaluation_seconds
|
||||
- alloy_component_evaluation_seconds_bucket
|
||||
- alloy_component_evaluation_seconds_count
|
||||
- alloy_component_evaluation_seconds_sum
|
||||
- alloy_component_evaluation_slow_seconds
|
||||
- alloy_component_controller_running_components
|
||||
- alloy_resources_machine_rx_bytes_total
|
||||
- alloy_resources_machine_tx_bytes_total
|
||||
- alloy_resources_process_cpu_seconds_total
|
||||
- alloy_resources_process_resident_memory_bytes
|
||||
- prometheus_remote_write_wal_samples_appended_total
|
||||
- prometheus_remote_write_wal_storage_active_series
|
||||
- cluster_node_info
|
||||
- cluster_node_lamport_time
|
||||
- cluster_node_update_observers
|
||||
- cluster_node_gossip_health_score
|
||||
- cluster_node_gossip_proto_version
|
||||
- cluster_node_gossip_received_events_total
|
||||
- cluster_node_peers
|
||||
- cluster_transport_rx_bytes_total
|
||||
- cluster_transport_rx_packets_total
|
||||
- cluster_transport_rx_packets_failed_total
|
||||
- cluster_transport_stream_rx_bytes_total
|
||||
- cluster_transport_stream_rx_packets_failed_total
|
||||
- cluster_transport_stream_rx_packets_total
|
||||
- cluster_transport_stream_tx_bytes_total
|
||||
- cluster_transport_stream_tx_packets_total
|
||||
- cluster_transport_stream_tx_packets_failed_total
|
||||
- cluster_transport_streams
|
||||
- cluster_transport_tx_packets_total
|
||||
- cluster_transport_tx_packets_failed_total
|
||||
- cluster_transport_rx_packet_queue_length
|
||||
- cluster_transport_tx_packet_queue_length
|
||||
- container_cpu_usage_seconds_total
|
||||
- container_fs_writes_bytes_total
|
||||
- container_memory_working_set_bytes
|
||||
@ -93,7 +138,10 @@ metrics:
|
||||
- cortex_prometheus_rule_group_last_duration_seconds
|
||||
- cortex_prometheus_rule_group_last_evaluation_timestamp_seconds
|
||||
- cortex_prometheus_rule_group_iterations_missed_total
|
||||
- exporter_send_failed_spans_ratio_total
|
||||
- exporter_sent_spans_ratio_total
|
||||
- go_gc_duration_seconds
|
||||
- go_gc_duration_seconds_count
|
||||
- go_goroutines
|
||||
- go_memstats_heap_inuse_bytes
|
||||
- kubelet_volume_stats_used_bytes
|
||||
@ -104,6 +152,7 @@ metrics:
|
||||
- kube_pod_container_resource_requests
|
||||
- kube_pod_container_status_last_terminated_reason
|
||||
- kube_pod_container_status_restarts_total
|
||||
- loki_azure_blob_request_duration_seconds_bucket
|
||||
- loki_boltdb_shipper_compact_tables_operation_duration_seconds
|
||||
- loki_boltdb_shipper_compact_tables_operation_last_successful_run_timestamp_seconds
|
||||
- loki_boltdb_shipper_retention_marker_count_total
|
||||
@ -129,10 +178,15 @@ metrics:
|
||||
- loki_compactor_deleted_lines
|
||||
- loki_compactor_oldest_pending_delete_request_age_seconds
|
||||
- loki_compactor_pending_delete_requests_count
|
||||
- loki_consul_request_duration_seconds_bucket
|
||||
- loki_discarded_samples_total
|
||||
- loki_discarded_bytes_total
|
||||
- loki_distributor_bytes_received_total
|
||||
- loki_distributor_lines_received_total
|
||||
- loki_distributor_structured_metadata_bytes_received_total
|
||||
- loki_gcs_request_duration_seconds_bucket
|
||||
- loki_gcs_request_duration_seconds_count
|
||||
- loki_index_request_duration_seconds_bucket
|
||||
- loki_index_request_duration_seconds_count
|
||||
- loki_ingester_chunk_age_seconds_bucket
|
||||
- loki_ingester_chunk_age_seconds_count
|
||||
@ -145,6 +199,7 @@ metrics:
|
||||
- loki_ingester_chunk_entries_sum
|
||||
- loki_ingester_chunk_size_bytes_bucket
|
||||
- loki_ingester_chunk_utilization_bucket
|
||||
- loki_ingester_chunk_utilization_count
|
||||
- loki_ingester_chunk_utilization_sum
|
||||
- loki_ingester_chunks_flushed_total
|
||||
- loki_ingester_flush_queue_length
|
||||
@ -162,6 +217,8 @@ metrics:
|
||||
- loki_ruler_wal_prometheus_remote_storage_samples_total
|
||||
- loki_ruler_wal_samples_appended_total
|
||||
- loki_ruler_wal_storage_created_series_total
|
||||
- loki_s3_request_duration_seconds_bucket
|
||||
- loki_s3_request_duration_seconds_count
|
||||
- loki_write_batch_retries_total
|
||||
- loki_write_dropped_bytes_total
|
||||
- loki_write_dropped_entries_total
|
||||
@ -169,7 +226,58 @@ metrics:
|
||||
- loki_write_sent_entries_total
|
||||
- node_disk_read_bytes_total
|
||||
- node_disk_written_bytes_total
|
||||
- process_start_time_seconds
|
||||
- processor_batch_batch_send_size_ratio_bucket
|
||||
- processor_batch_metadata_cardinality_ratio
|
||||
- processor_batch_timeout_trigger_send_ratio_total
|
||||
- prometheus_remote_storage_bytes_total
|
||||
- prometheus_remote_storage_enqueue_retries_total
|
||||
- prometheus_remote_storage_highest_timestamp_in_seconds
|
||||
- prometheus_remote_storage_metadata_bytes_total
|
||||
- prometheus_remote_storage_queue_highest_sent_timestamp_seconds
|
||||
- prometheus_remote_storage_samples_dropped_total
|
||||
- prometheus_remote_storage_samples_failed_total
|
||||
- prometheus_remote_storage_samples_pending
|
||||
- prometheus_remote_storage_samples_retried_total
|
||||
- prometheus_remote_storage_samples_total
|
||||
- prometheus_remote_storage_sent_batch_duration_seconds_bucket
|
||||
- prometheus_remote_storage_sent_batch_duration_seconds_count
|
||||
- prometheus_remote_storage_sent_batch_duration_seconds_sum
|
||||
- prometheus_remote_storage_shard_capacity
|
||||
- prometheus_remote_storage_shards
|
||||
- prometheus_remote_storage_shards_desired
|
||||
- prometheus_remote_storage_shards_max
|
||||
- prometheus_remote_storage_shards_min
|
||||
- prometheus_remote_storage_succeeded_samples_total
|
||||
- prometheus_remote_write_wal_samples_appended_total
|
||||
- prometheus_remote_write_wal_storage_active_series
|
||||
- prometheus_sd_discovered_targets
|
||||
- prometheus_target_interval_length_seconds_count
|
||||
- prometheus_target_interval_length_seconds_sum
|
||||
- prometheus_target_scrapes_exceeded_sample_limit_total
|
||||
- prometheus_target_scrapes_sample_duplicate_timestamp_total
|
||||
- prometheus_target_scrapes_sample_out_of_bounds_total
|
||||
- prometheus_target_scrapes_sample_out_of_order_total
|
||||
- prometheus_target_sync_length_seconds_sum
|
||||
- prometheus_wal_watcher_current_segment
|
||||
- promtail_custom_bad_words_total
|
||||
- promtail_dropped_bytes_total
|
||||
- promtail_files_active_total
|
||||
- promtail_read_bytes_total
|
||||
- promtail_read_lines_total
|
||||
- promtail_request_duration_seconds_bucket
|
||||
- promtail_sent_entries_total
|
||||
- rpc_server_duration_milliseconds_bucket
|
||||
- receiver_accepted_spans_ratio_total
|
||||
- receiver_refused_spans_ratio_total
|
||||
- scrape_duration_seconds
|
||||
- traces_exporter_sent_spans
|
||||
- traces_exporter_send_failed_spans
|
||||
- traces_loadbalancer_backend_outcome
|
||||
- traces_loadbalancer_num_backends
|
||||
- traces_receiver_accepted_spans
|
||||
- traces_receiver_refused_spans
|
||||
- up
|
||||
# Additional metrics to retain
|
||||
extraMetrics: []
|
||||
# Set enabled = true to add the default logs dashboards to the local Grafana
|
||||
|
||||
@ -1,8 +1,12 @@
|
||||
# Update the dependencies
|
||||
|
||||
The dependencies are the version of Loki, Mimir, Agent and so on that are included in this chart.
|
||||
The dependencies are the versions of Loki, Mimir, Agent and so on that are included in this chart.
|
||||
The current versions can be found in the [Chart.yaml](../charts/meta-monitoring/Chart.yaml) file.
|
||||
|
||||
A Github action runs daily to see if updated versions are available. A PR will be created.
|
||||
|
||||
The manual steps are as follows:
|
||||
|
||||
Run this in the charts/meta-monitoring directory after updating a dependency:
|
||||
|
||||
```
|
||||
|
||||
@ -4,7 +4,7 @@
|
||||
|
||||
1. Use an existing Grafana Cloud account or setup a new one. Then create an access token:
|
||||
|
||||
1. In Grafana go to Administration -> Users and Access -> Cloud access policies.
|
||||
1. In a Grafana instance on Grafana Cloud go to Administration -> Users and Access -> Cloud access policies.
|
||||
|
||||
1. Click `Create access policy`.
|
||||
|
||||
@ -39,7 +39,7 @@
|
||||
--from-literal=endpoint='https://otlp-gateway-prod-us-east-0.grafana.net/otlp'
|
||||
```
|
||||
|
||||
The logs, metrics and traces usernames are the `User / Username / Instance IDs` of the Loki, Prometheus/Mimir and OpenTelemetry instances in Grafana Cloud. From `Home` in Grafana click on `Stacks`. Then go to the `Details` pages of Loki and Prometheus/Mimir. For OpenTelemetry go to the `Configure` page.
|
||||
The logs, metrics and traces usernames are the `User / Username / Instance IDs` of the Loki, Prometheus/Mimir and OpenTelemetry instances in Grafana Cloud. From `Home` in Grafana click on `Stacks`. Then go to the `Details` pages of Loki and Prometheus/Mimir. For OpenTelemetry go to the `Configure` page. The endpoints will also have to be changed to match your settings.
|
||||
|
||||
1. Create a values.yaml file based on the [default one](../charts/meta-monitoring/values.yaml). Fill in the names of the secrets created above as needed. An example minimal values.yaml looks like this:
|
||||
|
||||
@ -91,7 +91,7 @@
|
||||
|
||||
local:
|
||||
grafana:
|
||||
enabled:true
|
||||
enabled: true
|
||||
logs:
|
||||
enabled: true
|
||||
metrics:
|
||||
@ -102,7 +102,7 @@
|
||||
enabled: true
|
||||
```
|
||||
|
||||
## Installing the chart
|
||||
## Installing, updating and deleting the chart
|
||||
|
||||
1. Add the repo
|
||||
|
||||
@ -175,7 +175,7 @@ For each of the dashboard files in charts/meta-monitoring/src/dashboards folder
|
||||
|
||||
## Configure Loki to send traces
|
||||
|
||||
1. In the Loki config enable tracing:
|
||||
1. In the Loki that is being monitored enable tracing in the config:
|
||||
|
||||
```
|
||||
loki:
|
||||
@ -187,7 +187,15 @@ For each of the dashboard files in charts/meta-monitoring/src/dashboards folder
|
||||
|
||||
1. JAEGER_ENDPOINT: http address of the mmc-alloy service installed by the meta-monitoring chart, for example "http://mmc-alloy:14268/api/traces"
|
||||
1. JAEGER_AGENT_TAGS: extra tags you would like to add to the spans, for example 'cluster="abc",namespace="def"'
|
||||
1. JAEGER_SAMPLER_TYPE: the sampling strategy, for example to sample all use 'const' with a value of 1 for the next environment variable
|
||||
1. JAEGER_SAMPLER_PARAM: 1
|
||||
1. JAEGER_SAMPLER_TYPE: the sampling strategy, we suggest setting this to `ratelimiting` so at most 1 trace is accepted per second. See these [docs](https://www.jaegertracing.io/docs/1.57/sampling/) for more options.
|
||||
1. JAEGER_SAMPLER_PARAM: 1.0
|
||||
|
||||
1. If Loki is installed in a different namespace you can create an [ExternalName service](https://kubernetes.io/docs/concepts/services-networking/service/#externalname) in Kubernetes to point to the mmc-alloy service in the meta monitoring namespace
|
||||
|
||||
## Configure external access using an Ingress in local mode
|
||||
|
||||
When using local mode by default a Kubernetes [Ingress](https://kubernetes.io/docs/concepts/services-networking/ingress/) object is created to access the Grafana instance. This will need to be adapted to your cloud provider by updating the `grafana.ingress` section of the `values.yaml` file provided to Helm. Check the documentation of your cloud provider for available options.
|
||||
|
||||
## Kube-state-metrics
|
||||
|
||||
Metrics about Kubernetes objects are scraped from [kube-state-metrics](https://github.com/kubernetes/kube-state-metrics). This needs to be installed in the cluster. The `kubeStateMetrics.endpoint` entry in values.yaml should be set to it's address (without the `/metrics` part in the URL).
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user