summaryrefslogtreecommitdiffstats
path: root/health/health.d/vcsa.conf
diff options
context:
space:
mode:
Diffstat (limited to 'health/health.d/vcsa.conf')
-rw-r--r--health/health.d/vcsa.conf122
1 files changed, 122 insertions, 0 deletions
diff --git a/health/health.d/vcsa.conf b/health/health.d/vcsa.conf
new file mode 100644
index 000000000..7bb98a9ba
--- /dev/null
+++ b/health/health.d/vcsa.conf
@@ -0,0 +1,122 @@
+
+# make sure vcsa is running and responding
+
+template: vcsa_last_collected_secs
+ on: vcsa.system_health
+ calc: $now - $last_collected_t
+ units: seconds ago
+ every: 10s
+ warn: $this > (($status >= $WARNING) ? ($update_every) : ( 5 * $update_every))
+ crit: $this > (($status == $CRITICAL) ? ($update_every) : (60 * $update_every))
+ delay: down 5m multiplier 1.5 max 1h
+ info: number of seconds since the last successful data collection
+ to: sysadmin
+
+# Overall system health:
+# - 0: all components are healthy.
+# - 1: one or more components might become overloaded soon.
+# - 2: one or more components in the appliance might be degraded.
+# - 3: one or more components might be in an unusable status and the appliance might become unresponsive soon.
+# - 4: no health data is available.
+
+template: vcsa_system_health
+ on: vcsa.system_health
+ lookup: max -10s unaligned of system
+ units: status
+ every: 10s
+ warn: ($this == 1) || ($this == 2)
+ crit: $this == 3
+ delay: down 1m multiplier 1.5 max 1h
+ info: overall system health status
+ to: sysadmin
+
+# Components health:
+# - 0: healthy.
+# - 1: healthy, but may have some problems.
+# - 2: degraded, and may have serious problems.
+# - 3: unavailable, or will stop functioning soon.
+# - 4: no health data is available.
+
+template: vcsa_swap_health
+ on: vcsa.components_health
+ lookup: max -10s unaligned of swap
+ units: status
+ every: 10s
+ warn: $this == 1
+ crit: ($this == 2) || ($this == 3)
+ delay: down 1m multiplier 1.5 max 1h
+ info: swap health status
+ to: sysadmin
+
+template: vcsa_storage_health
+ on: vcsa.components_health
+ lookup: max -10s unaligned of storage
+ units: status
+ every: 10s
+ warn: $this == 1
+ crit: ($this == 2) || ($this == 3)
+ delay: down 1m multiplier 1.5 max 1h
+ info: storage health status
+ to: sysadmin
+
+template: vcsa_mem_health
+ on: vcsa.components_health
+ lookup: max -10s unaligned of mem
+ units: status
+ every: 10s
+ warn: $this == 1
+ crit: ($this == 2) || ($this == 3)
+ delay: down 1m multiplier 1.5 max 1h
+ info: mem health status
+ to: sysadmin
+
+template: vcsa_load_health
+ on: vcsa.components_health
+ lookup: max -10s unaligned of load
+ units: status
+ every: 10s
+ warn: $this == 1
+ crit: ($this == 2) || ($this == 3)
+ delay: down 1m multiplier 1.5 max 1h
+ info: load health status
+ to: sysadmin
+
+template: vcsa_database_storage_health
+ on: vcsa.components_health
+ lookup: max -10s unaligned of database_storage
+ units: status
+ every: 10s
+ warn: $this == 1
+ crit: ($this == 2) || ($this == 3)
+ delay: down 1m multiplier 1.5 max 1h
+ info: database storage health status
+ to: sysadmin
+
+template: vcsa_applmgmt_health
+ on: vcsa.components_health
+ lookup: max -10s unaligned of applmgmt
+ units: status
+ every: 10s
+ warn: $this == 1
+ crit: ($this == 2) || ($this == 3)
+ delay: down 1m multiplier 1.5 max 1h
+ info: appl mgmt health status
+ to: sysadmin
+
+
+# Software updates health:
+# - 0: no updates available.
+# - 2: non-security updates are available.
+# - 3: security updates are available.
+# - 4: an error retrieving information on software updates.
+
+template: vcsa_software_updates_health
+ on: vcsa.software_updates_health
+ lookup: max -10s unaligned of software_packages
+ units: status
+ every: 10s
+ warn: $this == 4
+ crit: $this == 3
+ delay: down 1m multiplier 1.5 max 1h
+ info: software packages health status
+ to: sysadmin