Add thresholds for memory alerts

Signed-off-by: Vitaly Zhuravlev <v-zhuravlev@users.noreply.github.com>
This commit is contained in:
Vitaly Zhuravlev 2023-04-06 00:21:50 +08:00
parent 2111e70ac7
commit 77ae769179
2 changed files with 13 additions and 5 deletions

View File

@ -344,7 +344,7 @@
{
alert: 'NodeMemoryMajorPagesFaults',
expr: |||
rate(node_vmstat_pgmajfault{%(nodeExporterSelector)s}[5m]) > 500
rate(node_vmstat_pgmajfault{%(nodeExporterSelector)s}[5m]) > %(memoryMajorPagesFaultsWarningThreshold)s
||| % $._config,
'for': '15m',
labels: {
@ -353,15 +353,15 @@
annotations: {
summary: 'Memory major page faults are occurring at very high rate.',
description: |||
Memory major pages are occurring at very high rate at {{ $labels.instance }}, 500 major page faults per second for the last 15 minutes, is currently at {{ printf "%.2f" $value }}.
Memory major pages are occurring at very high rate at {{ $labels.instance }}, %(memoryMajorPagesFaultsWarningThreshold)s major page faults per second for the last 15 minutes, is currently at {{ printf "%.2f" $value }}.
Please check that there is enough memory available at this instance.
|||,
||| % $._config,
},
},
{
alert: 'NodeMemoryHighUtilization',
expr: |||
100 - (node_memory_MemAvailable_bytes{%(nodeExporterSelector)s} / node_memory_MemTotal_bytes{%(nodeExporterSelector)s} * 100) > 90
100 - (node_memory_MemAvailable_bytes{%(nodeExporterSelector)s} / node_memory_MemTotal_bytes{%(nodeExporterSelector)s} * 100) > %(memoryHighUtilizationThreshold)s
||| % $._config,
'for': '15m',
labels: {
@ -370,7 +370,7 @@
annotations: {
summary: 'Host is running out of memory.',
description: |||
Memory is filling up at {{ $labels.instance }}, has been above 90% for the last 15 minutes, is currently at {{ printf "%.2f" $value }}%.
Memory is filling up at {{ $labels.instance }}, has been above %(memoryHighUtilizationThreshold)s% for the last 15 minutes, is currently at {{ printf "%.2f" $value }}%.
|||,
},
},

View File

@ -60,6 +60,14 @@
fsSpaceAvailableWarningThreshold: 5,
fsSpaceAvailableCriticalThreshold: 3,
// Memory utilzation (%) level on which to trigger the
// 'NodeMemoryHighUtilization' alert.
memoryHighUtilizationThreshold: 90,
// Threshold for the rate of memory major page faults to trigger
// 'NodeMemoryMajorPagesFaults' alert.
memoryMajorPagesFaultsWarningThreshold: 500,
rateInterval: '5m',
// Opt-in for multi-cluster support.
showMultiCluster: false,