diff options
author | Costa Tsaousis <costa@tsaousis.gr> | 2017-03-08 23:54:39 +0200 |
---|---|---|
committer | GitHub <noreply@github.com> | 2017-03-08 23:54:39 +0200 |
commit | 696fe43d1855a97775c2bec17d1f2f002e32eb0e (patch) | |
tree | 956248d4473f38eec2345ae3d43627db62770775 /conf.d | |
parent | 8bec52f8768b928c3bcf57980ad1aa45f296c836 (diff) | |
parent | 4d8bb6f1af921722ffdf29342ffa3a386bca15c5 (diff) |
Merge pull request #1893 from ktsaou/master
self-cleanup removed disks and more minor fixes
Diffstat (limited to 'conf.d')
-rw-r--r-- | conf.d/Makefile.am | 1 | ||||
-rw-r--r-- | conf.d/apps_groups.conf | 2 | ||||
-rw-r--r-- | conf.d/health.d/disks.conf | 30 | ||||
-rw-r--r-- | conf.d/health.d/ipmi.conf | 20 |
4 files changed, 23 insertions, 30 deletions
diff --git a/conf.d/Makefile.am b/conf.d/Makefile.am index 655d27a737..47f6573c79 100644 --- a/conf.d/Makefile.am +++ b/conf.d/Makefile.am @@ -67,6 +67,7 @@ dist_healthconfig_DATA = \ health.d/fping.conf \ health.d/haproxy.conf \ health.d/ipfs.conf \ + health.d/ipmi.conf \ health.d/isc_dhcpd.conf \ health.d/mdstat.conf \ health.d/memcached.conf \ diff --git a/conf.d/apps_groups.conf b/conf.d/apps_groups.conf index 61d9c2c8b6..4c5171b3d0 100644 --- a/conf.d/apps_groups.conf +++ b/conf.d/apps_groups.conf @@ -73,10 +73,12 @@ netdata: netdata # netdata known plugins # plugins not defined here will be accumulated in netdata, above apps.plugin: apps.plugin +freeipmi.plugin: freeipmi.plugin charts.d.plugin: *charts.d.plugin* node.d.plugin: *node.d.plugin* python.d.plugin: *python.d.plugin* tc-qos-helper: *tc-qos-helper.sh* +fping: fping # ----------------------------------------------------------------------------- # authentication/authorization related servers diff --git a/conf.d/health.d/disks.conf b/conf.d/health.d/disks.conf index ff2d6a605a..9548f9ee01 100644 --- a/conf.d/health.d/disks.conf +++ b/conf.d/health.d/disks.conf @@ -1,34 +1,4 @@ # ----------------------------------------------------------------------------- -# make sure we collect values for each disk - -# for mount points -template: disk_space_last_collected_secs - on: disk.space -families: * - 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 of the mount point - to: sysadmin - -# for block devices -template: disk_last_collected_secs - on: disk.io -families: * - 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 of the block device - to: sysadmin - - -# ----------------------------------------------------------------------------- # low disk space # checking the latest collected values diff --git a/conf.d/health.d/ipmi.conf b/conf.d/health.d/ipmi.conf new file mode 100644 index 0000000000..c255819645 --- /dev/null +++ b/conf.d/health.d/ipmi.conf @@ -0,0 +1,20 @@ + alarm: ipmi_sensors_states + on: ipmi.sensors_states + calc: $warning + $critical + units: sensors + every: 10s + warn: $this > 0 + crit: $critical > 0 + delay: up 5m down 15m multiplier 1.5 max 1h + info: the number IPMI sensors in non-nominal state + to: sysadmin + + alarm: ipmi_events + on: ipmi.events + calc: $events + units: events + every: 10s + warn: $this > 0 + delay: up 5m down 15m multiplier 1.5 max 1h + info: the number of events in the IPMI System Event Log (SEL) + to: sysadmin |