summaryrefslogtreecommitdiffstats
path: root/health/health.d/memory.conf
diff options
context:
space:
mode:
Diffstat (limited to 'health/health.d/memory.conf')
-rw-r--r--health/health.d/memory.conf38
1 files changed, 38 insertions, 0 deletions
diff --git a/health/health.d/memory.conf b/health/health.d/memory.conf
new file mode 100644
index 000000000..4a0e6e522
--- /dev/null
+++ b/health/health.d/memory.conf
@@ -0,0 +1,38 @@
+
+# you can disable an alarm notification by setting the 'to' line to: silent
+
+ alarm: 1hour_ecc_memory_correctable
+ on: mem.ecc_ce
+ os: linux
+ hosts: *
+ lookup: sum -10m unaligned
+ units: errors
+ every: 1m
+ warn: $this > 0
+ delay: down 1h multiplier 1.5 max 1h
+ info: number of ECC correctable errors during the last hour
+ to: sysadmin
+
+ alarm: 1hour_ecc_memory_uncorrectable
+ on: mem.ecc_ue
+ os: linux
+ hosts: *
+ lookup: sum -10m unaligned
+ units: errors
+ every: 1m
+ crit: $this > 0
+ delay: down 1h multiplier 1.5 max 1h
+ info: number of ECC uncorrectable errors during the last hour
+ to: sysadmin
+
+ alarm: 1hour_memory_hw_corrupted
+ on: mem.hwcorrupt
+ os: linux
+ hosts: *
+ calc: $HardwareCorrupted
+ units: MB
+ every: 10s
+ warn: $this > 0
+ delay: down 1h multiplier 1.5 max 1h
+ info: amount of memory corrupted due to a hardware failure
+ to: sysadmin