diff options
| author | Ilya Mashchenko <ilya@netdata.cloud> | 2021-03-23 10:44:02 +0300 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2021-03-23 10:44:02 +0300 |
| commit | 6fe27ffe0109d0c9bbeddaa893715cb389fb4234 (patch) | |
| tree | d40b9748d1fa171bd49deba5cd7c9ef2a9b7b516 | |
| parent | 877021800d3810c154e9c4eec8a7b6cab8855267 (diff) | |
health: apply megacli alarms for all adapters/physical disks (#10834)
| -rw-r--r-- | health/health.d/megacli.conf | 54 |
1 files changed, 31 insertions, 23 deletions
diff --git a/health/health.d/megacli.conf b/health/health.d/megacli.conf index 73f106530a..434eee3b41 100644 --- a/health/health.d/megacli.conf +++ b/health/health.d/megacli.conf @@ -1,12 +1,40 @@ + +## Adapters (controllers) + template: megacli_adapter_state on: megacli.adapter_degraded - units: is degraded - lookup: sum -10s + lookup: max -10s foreach * + units: boolean every: 10s crit: $this > 0 - info: adapter state + delay: down 5m multiplier 2 max 10m + info: adapter state is degraded + to: sysadmin + +## Physical Disks + +template: megacli_pd_predictive_failures + on: megacli.pd_predictive_failure + lookup: sum -10s foreach * + units: predictive failures + every: 10s + warn: $this > 0 + delay: up 1m down 5m multiplier 2 max 10m + info: physical drive predictive failures + to: sysadmin + +template: megacli_pd_media_errors + on: megacli.pd_media_error + lookup: sum -10s foreach * + units: media errors + every: 10s + warn: $this > 0 + delay: up 1m down 5m multiplier 2 max 10m + info: physical drive media errors to: sysadmin +## Battery Backup Units (BBU) + template: megacli_bbu_relative_charge on: megacli.bbu_relative_charge units: percent @@ -26,23 +54,3 @@ template: megacli_bbu_cycle_count crit: $this >= 500 info: BBU cycle count to: sysadmin - -template: megacli_pd_media_errors - on: megacli.pd_media_error - units: media errors - lookup: sum -10s - every: 10s - warn: $this > 0 - delay: down 1m multiplier 2 max 10m - info: physical drive media errors - to: sysadmin - -template: megacli_pd_predictive_failures - on: megacli.pd_predictive_failure - units: predictive failures - lookup: sum -10s - every: 10s - warn: $this > 0 - delay: down 1m multiplier 2 max 10m - info: physical drive predictive failures - to: sysadmin |