Switch critical severity to error for CI/CD alerts

parent a232b94f
...@@ -80,7 +80,7 @@ groups: ...@@ -80,7 +80,7 @@ groups:
for: 5m for: 5m
labels: labels:
channel: ci-cd channel: ci-cd
severity: critical severity: error
annotations: annotations:
title: "{{ $labels.job }} runners are using 95% of concurrent limit for more than 5 minutes." title: "{{ $labels.job }} runners are using 95% of concurrent limit for more than 5 minutes."
description: 'Hey <!subteam^S940BK2TV|cicdops>! This may suggest problems with our autoscaled machines fleet OR description: 'Hey <!subteam^S940BK2TV|cicdops>! This may suggest problems with our autoscaled machines fleet OR
...@@ -91,7 +91,7 @@ groups: ...@@ -91,7 +91,7 @@ groups:
for: 5m for: 5m
labels: labels:
channel: ci-cd channel: ci-cd
severity: critical severity: error
annotations: annotations:
title: Runners manager is down on {{ $labels.instance }} title: Runners manager is down on {{ $labels.instance }}
description: 'Hey <!subteam^S940BK2TV|cicdops>! This impacts CI execution builds, consider tweeting: !tweet ''Builds description: 'Hey <!subteam^S940BK2TV|cicdops>! This impacts CI execution builds, consider tweeting: !tweet ''Builds
...@@ -118,7 +118,7 @@ groups: ...@@ -118,7 +118,7 @@ groups:
for: 5m for: 5m
labels: labels:
channel: ci-cd channel: ci-cd
severity: critical severity: error
annotations: annotations:
title: Runners cache service {{ $labels.instance }} on {{ $labels.fqdn }} has been down for more than 5 minutes. title: Runners cache service {{ $labels.instance }} on {{ $labels.fqdn }} has been down for more than 5 minutes.
description: 'Hey <!subteam^S940BK2TV|cicdops>! This impacts CI execution builds, consider tweeting: !tweet ''CI description: 'Hey <!subteam^S940BK2TV|cicdops>! This impacts CI execution builds, consider tweeting: !tweet ''CI
...@@ -131,7 +131,7 @@ groups: ...@@ -131,7 +131,7 @@ groups:
for: 5m for: 5m
labels: labels:
channel: ci-cd channel: ci-cd
severity: critical severity: error
annotations: annotations:
title: Runners cache nginx service on {{ $labels.fqdn }} has been down for more than 5 minutes. title: Runners cache nginx service on {{ $labels.fqdn }} has been down for more than 5 minutes.
description: 'Hey <!subteam^S940BK2TV|cicdops>! This impacts CI execution builds, consider tweeting: !tweet ''CI description: 'Hey <!subteam^S940BK2TV|cicdops>! This impacts CI execution builds, consider tweeting: !tweet ''CI
...@@ -144,7 +144,7 @@ groups: ...@@ -144,7 +144,7 @@ groups:
for: 20m for: 20m
labels: labels:
channel: ci-cd channel: ci-cd
severity: critical severity: error
annotations: annotations:
title: Number of established connections for {{ $labels.instance }} is too high title: Number of established connections for {{ $labels.instance }} is too high
description: 'Hey <!subteam^S940BK2TV|cicdops>! This impacts CI execution builds, consider tweeting: !tweet ''CI description: 'Hey <!subteam^S940BK2TV|cicdops>! This impacts CI execution builds, consider tweeting: !tweet ''CI
...@@ -199,7 +199,7 @@ groups: ...@@ -199,7 +199,7 @@ groups:
for: 5m for: 5m
labels: labels:
channel: ci-cd channel: ci-cd
severity: critical severity: error
annotations: annotations:
title: GCP Quota usage of {{ $labels.quota }} is near limit title: GCP Quota usage of {{ $labels.quota }} is near limit
description: | description: |
...@@ -232,7 +232,7 @@ groups: ...@@ -232,7 +232,7 @@ groups:
for: 5m for: 5m
labels: labels:
channel: ci-cd channel: ci-cd
severity: critical severity: error
annotations: annotations:
title: Job queue duration performance for '0 - 10s' is to low title: Job queue duration performance for '0 - 10s' is to low
description: | description: |
...@@ -252,7 +252,7 @@ groups: ...@@ -252,7 +252,7 @@ groups:
for: 5m for: 5m
labels: labels:
channel: ci-cd channel: ci-cd
severity: critical severity: error
annotations: annotations:
title: 90% of request queued on Workhorse is longer than 30s title: 90% of request queued on Workhorse is longer than 30s
description: | description: |
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment