diff --git a/roles/prometheus/files/node.rules.yml b/roles/prometheus/files/node.rules.yml index 61d9d5bb917a07c858b7bca552d72aa32b6acc35..2f13af3633383eb2d2df4acfccde306e4f5870f7 100644 --- a/roles/prometheus/files/node.rules.yml +++ b/roles/prometheus/files/node.rules.yml @@ -3,7 +3,7 @@ groups: interval: 60s rules: - alert: HostHighCpuLoad - expr: 100 - (avg by(instance) (irate(node_cpu_seconds_total{mode="idle",instance!~"build.archlinux.org",instance!~"repro3.pkgbuild.com",instance!~"repro2.pkgbuild.com",instance!~"runner1.archlinux.org",instance!~"runner3.archlinux.org"}[10m])) * 100) > 90 + expr: 100 - (avg by(instance) (irate(node_cpu_seconds_total{mode="idle",instance!~"build.archlinux.org",instance!~"repro.*pkgbuild.com",instance!~"runner.*archlinux.org"}[10m])) * 100) > 90 for: 10m labels: severity: warning @@ -21,7 +21,7 @@ groups: description: "Swap is filling up (>80%)\n VALUE = {{ $value }}\n LABELS: {{ $labels }}" - alert: HostOutOfMemory - expr: node_memory_MemAvailable_bytes{instance!~"build.archlinux.org",instance!~"repro3.pkgbuild.com",instance!~"repro2.pkgbuild.com"} / node_memory_MemTotal_bytes * 100 < 10 + expr: node_memory_MemAvailable_bytes{instance!~"build.archlinux.org",instance!~"repro.*pkgbuild.com"} / node_memory_MemTotal_bytes * 100 < 10 for: 5m labels: severity: warning @@ -30,7 +30,7 @@ groups: description: "Node memory is filling up (< 10% left)\n VALUE = {{ $value }}\n LABELS: {{ $labels }}" - alert: BuildHostOutOfMemory - expr: node_memory_MemAvailable_bytes{instance="build.archlinux.org",instance="repro3.pkgbuild.com",instance="repro2.pkgbuild.com"} / node_memory_MemTotal_bytes * 100 < 10 + expr: node_memory_MemAvailable_bytes{instance="build.archlinux.org",instance="repro.*pkgbuild.com"} / node_memory_MemTotal_bytes * 100 < 10 for: 12h labels: severity: warning