summaryrefslogtreecommitdiffstats
path: root/src/health/health.d/megacli.conf
diff options
context:
space:
mode:
Diffstat (limited to 'src/health/health.d/megacli.conf')
-rw-r--r--src/health/health.d/megacli.conf77
1 files changed, 77 insertions, 0 deletions
diff --git a/src/health/health.d/megacli.conf b/src/health/health.d/megacli.conf
new file mode 100644
index 000000000..27721fa9a
--- /dev/null
+++ b/src/health/health.d/megacli.conf
@@ -0,0 +1,77 @@
+# you can disable an alarm notification by setting the 'to' line to: silent
+
+# Adapters (controllers)
+
+ template: megacli_adapter_health_state
+ on: megacli.adapter_health_state
+ class: Errors
+ type: System
+component: RAID
+ lookup: average -1m unaligned percentage of optimal
+ units: %
+ every: 10s
+ crit: $this < 100
+ delay: down 5m multiplier 2 max 10m
+ summary: MegaCLI adapter ${label:adapter_number} health
+ info: MegaCLI adapter ${label:adapter_number} is in the degraded state
+ to: sysadmin
+
+ template: megacli_phys_drive_media_errors
+ on: megacli.phys_drive_media_errors
+ class: Errors
+ type: System
+component: RAID
+ lookup: sum -10s
+ units: media errors
+ every: 10s
+ warn: $this > 0
+ delay: up 1m down 5m multiplier 2 max 10m
+ summary: MegaCLI PD adapter ${label:adapter_number} slot ${label:slot_number} media errors
+ info: MegaCLI physical drive adapter ${label:adapter_number} slot ${label:slot_number} media errors
+ to: sysadmin
+
+# Physical Drives
+
+ template: megacli_phys_drive_predictive_failures
+ on: megacli.phys_drive_predictive_failures
+ class: Errors
+ type: System
+component: RAID
+ lookup: sum -10s
+ units: failures
+ every: 10s
+ warn: $this > 0
+ delay: up 1m down 5m multiplier 2 max 10m
+ summary: MegaCLI PD adapter ${label:adapter_number} slot ${label:slot_number} predictive failures
+ info: MegaCLI physical drive (adapter ${label:adapter_number} slot ${label:slot_number}) predictive failures
+ to: sysadmin
+
+# Backup Battery Unit
+
+ template: megacli_bbu_charge
+ on: megacli.bbu_charge
+ class: Workload
+ type: System
+component: RAID
+ lookup: average -10s
+ units: percent
+ every: 10s
+ warn: $this <= (($status >= $WARNING) ? (85) : (80))
+ crit: $this <= (($status == $CRITICAL) ? (50) : (40))
+ summary: MegaCLI BBU charge
+ info: MegaCLI Backup Battery Unit (adapter ${label:adapter_number}) average charge over the last minute
+ to: sysadmin
+
+ template: megacli_bbu_recharge_cycles
+ on: megacli.bbu_recharge_cycles
+ class: Workload
+ type: System
+component: RAID
+ lookup: average -10s
+ units: cycles
+ every: 10s
+ warn: $this >= 100
+ crit: $this >= 500
+ summary: MegaCLI BBU recharge cycles
+ info: MegaCLI Backup Battery Unit (adapter ${label:adapter_number}) recharge cycles
+ to: sysadmin