# Config file for mars_check.sh
#
# This file is part of MARS project: http://schoebel.github.io/mars/
#
# Copyright (C) 2015 Thomas Schoebel-Theuer
# Copyright (C) 2015 1&1 Internet AG
#
# Copying and distribution of this file, with or without modification,
# are permitted in any medium without royalty provided the copyright
# notice and this notice are preserved.  This file is offered as-is,
# without any warranty.


# For each variable $Var (as documented in "mars_check.sh --help"), the following relatives are always defined:
#
# $LastVar           the old value from the last run of mars_check.sh (whenever it was called; recommendation: 5 minutes)
# $DeltaLastVar      the difference between $Var and $LastVar
# $RateLastVar       the $DeltaLastVar normalized to the elapsed time (unit: per minutes)
#
# $MediumVar         the old value from a medium-term run of mars_check.sh ($window_medium, default 3600s)
# $DeltaMediumVar    the difference between $Var and $MediumVar
# $RateMediumVar     the $DeltaMediumVar normalized to the elapsed time (unit: per minutes)
#
# $LongtermVar       the old value from a longterm run of mars_check.sh ($window_longterm, default 24h)
# $DeltaLongtermVar  the difference between $Var and $LongtermVar
# $RateLongtermVar   the $DeltaLongtermVar normalized to the elapsed time (unit: per minutes)


# The first number in each line is the priority class.
# Lower number = higher priority = takes precence over higher class numbers
# Exception: 0 means that the check will appear unconditionally (class is irrelevant)
#
# Hint: checks can be simply disabled by commenting them out


#####################################################################################
# List of global checks

1 ModuleLoaded <= 0             CRITICAL: mars module is not loaded
2 Responsive <= 0               CRITICAL: mars_light thread is not responsive / possibly hanging
5 SpaceRest <= 4                CRITICAL: only $SpaceRest GiB left on /mars/
6 SpacePercent >= 70            CRITICAL: Used space on /mars/ is $SpacePercent %
7 SpacePercent >= 30            WARNING: Used space on /mars/ is $SpacePercent %

#####################################################################################
# List of local checks = per resource. The resource name can be substituted via $res

# all hosts
10 AliveAge >= 300               CRITICAL: resource $res: primary host ${Designated[$res]} is not reachable for $AliveAge seconds
11 Alive <= 0                    WARNING: resource $res: primary host ${Designated[$res]} is not reachable
12 Emergency >= 1                CRITICAL: resource $res is in emergency mode, too less space on /mars/
13 SplitBrain >= 1               CRITICAL: split brain on $res detected

# only secondaries
30 Sync <= 0                     WARNING: resource $res sync is switched off
31 Fetch <= 0                    WARNING: resource $res fetch is switched off
32 Replay <= 0                   WARNING: resource $res replay is switched off

40 SyncRest >= 999999            WARNING: resource $res SyncRest=${SyncRest[$res]} is too large
41 FetchRest >= 999999           WARNING: resource $res FetchRest=${FetchRest[$res]} is too large
42 ReplayRest >= 999999          WARNING: resource $res ReplayRest=${ReplayRest[$res]} is too large

50 DeltaLastSyncRest <= 99999   && SyncRest >= 1     WARNING: resource $res SyncRest=${SyncRest[$res]} sync has stopped
51 DeltaLastFetchRest <= 99999  && FetchRest >= 1    WARNING: resource $res FetchRest=${FetchRest[$res]} fetch has stopped
52 DeltaLastReplayRest <= 99999 && ReplayRest >= 1   WARNING: resource $res ReplayRest=${ReplayRest[$res]} replay has stopped