Просмотр исходного кода

health: enable go.d data collection job status alert (#18989)

Ilya Mashchenko 4 месяцев назад
Родитель
Сommit
85988eb177
1 измененных файлов с 9 добавлено и 11 удалено
  1. 9 11
      src/health/health.d/go.d.plugin.conf

+ 9 - 11
src/health/health.d/go.d.plugin.conf

@@ -1,17 +1,15 @@
 # make sure go.d.plugin data collection job is running
 
-   template: go.d_job_last_collected_secs
-         on: netdata.go_plugin_execution_time
+   template: plugin_data_collection_status
+         on: netdata.plugin_data_collection_status
       class: Errors
        type: Netdata
   component: go.d.plugin
-host labels: _hostname=!*
-       calc: $now - $last_collected_t
-      units: seconds ago
+     lookup: average -30s unaligned of failed
+      units: status
       every: 10s
-       warn: $this > (($status >= $WARNING)  ? ($update_every) : ( 5 * $update_every))
-       crit: $this > (($status == $CRITICAL) ? ($update_every) : (60 * $update_every))
-      delay: down 5m multiplier 1.5 max 1h
-    summary: Go.d plugin last collection
-       info: Number of seconds since the last successful data collection
-         to: webmaster
+       warn: $this == 1
+      delay: down 1m
+    summary: Data collection failure (${label:_collect_plugin}/${label:_collect_module}/${label:_collect_job})
+       info: Data collection failure (${label:_collect_plugin}/${label:_collect_module}/${label:_collect_job})
+         to: silent