NAME
Monitoring::Sneck::Checks::oom - checks for processes killed for running out of memory
VERSION
Version 0.0.1
SYNOPSIS
oom|perl:oom
oom_held|perl:oom -H 3600 -c 5DESCRIPTION
Counts the processes the Linux OOM killer killed since the last run, from oom_kill in /proc/vmstat, which kernels from 4.13 have. By default any is WARNING.
The count is kept in a state file, see "read_state" in Monitoring::Sneck::Checks::base, so the first run, and the first after a reboot, have nothing to compare to and are OK.
A kill is only seen by the run after it, so the next run is OK again, which a poll can easily miss. --hold keeps reporting it for that many seconds after.
Only Linux has this. Other OSes are UNKNOWN, as is a kernel without oom_kill.
OPTIONS
- -w, --warning :: Range for the kills since the last run. Default :: 0
- -c, --critical :: Range for the kills since the last run.
- -H, --hold :: Seconds to keep reporting kills for. Default :: 0
METHODS
check
Does the check.