kudos/check_health
kudos/check_health
2
1
Fork 0
Code Issues 7 Pull Requests Projects Releases 1 Wiki Activity

make _WAIT_TIME & _CRON_LOG_FILE configurable

Browse Source
This commit is contained in:
Patrick Van der Veken 2021-03-29 17:50:18 +02:00
parent a00d68afde
commit 11a888638d
3 changed files with 118 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
build/hpux/hc_hpux_platform/hc_hpux_platform.psf
View File

@ -105,6 +105,7 @@ This is the OS/platform plugin package"
directory ../../../etc/opt/hc/=/etc/opt/hc
file -m 644 check_hpux_autofs.conf.dist
file -m 644 check_hpux_cron_status.conf.dist
file -m 644 check_hpux_drd_status.conf.dist
file -m 644 check_hpux_file_age.conf.dist
file -m 644 check_hpux_file_change.conf.dist

25
etc/opt/hc/check_hpux_cron_status.conf.dist Normal file
View File

@ -0,0 +1,25 @@
#******************************************************************************
# @(#) check_hpux_cron_status.conf
#******************************************************************************
# This is a configuration file for the check_hpux_cron_status HC plugin.
# All lines starting with a '#' are comment lines.
# [default: indicates hardcoded script values if no value is defined here]
#******************************************************************************
# specify whether to also log passed health checks
# (warning: this may rapidly grow the HC log)
# [default: no]
log_healthy="no"
# specify the wait time (in seconds) between registering and checking a cron task
# [default: 10]
wait_time=10
# specify the path the cron log file
# [default: /var/adm/cron/log]
cron_log="/var/adm/cron/log"
#******************************************************************************
# End of FILE
#******************************************************************************

108
opt/hc/lib/platform/hp-ux/check_hpux_cron_status.sh
View File

@ -19,12 +19,13 @@
# @(#) MAIN: check_hpux_cron_status
# DOES: see _show_usage()
# EXPECTS: n/a
# REQUIRES: data_comma2space(), init_hc(), log_hc()
# REQUIRES: data_is_numeric(), data_comma2space(), init_hc(), log_hc()
#
# @(#) HISTORY:
# @(#) 2018-02-08: initial version [Patrick Van der Veken]µ
# @(#) 2018-02-13: fix to avoid log check if cron is not active [Patrick Van der Veken]
# @(#) 2019-03-16: replace 'which' [Patrick Van der Veken]
# @(#) 2021-03-25: make _WAIT_TIME & _CRON_LOG_FILE configurable [Patrick Van der Veken]
# -----------------------------------------------------------------------------
# DO NOT CHANGE THIS FILE UNLESS YOU KNOW WHAT YOU ARE DOING!
#******************************************************************************
@ -33,20 +34,25 @@
function check_hpux_cron_status
{
# ------------------------- CONFIGURATION starts here -------------------------
typeset _CRON_LOG_FILE="/var/adm/cron/log"
typeset _WAIT_TIME=10
typeset _VERSION="2019-02-16" # YYYY-MM-DD
typeset _CONFIG_FILE="${CONFIG_DIR}/$0.conf"
typeset _VERSION="2021-03-25" # YYYY-MM-DD
typeset _SUPPORTED_PLATFORMS="HP-UX" # uname -s match
# ------------------------- CONFIGURATION ends here ---------------------------
# set defaults
(( ARG_DEBUG > 0 && ARG_DEBUG_LEVEL > 0 )) && set ${DEBUG_OPTS}
(( ARG_DEBUG > 0 && ARG_DEBUG_LEVEL > 0 )) && set "${DEBUG_OPTS}"
init_hc "$0" "${_SUPPORTED_PLATFORMS}" "${_VERSION}"
typeset _ARGS=$(data_comma2space "$*")
typeset _ARG=""
typeset _MSG=""
typeset _STC=0
typeset _CFG_HEALTHY=""
typeset _LOG_HEALTHY=0
typeset _CFG_CRON_LOG_FILE=""
typeset _CFG_WAIT_TIME=""
typeset _LOG_HEALTHY=0
typeset _CRON_LOG_FILE=""
typeset _WAIT_TIME=""
typeset _JOB_ID=""
typeset _AT_BIN=""
@ -55,11 +61,56 @@ for _ARG in ${_ARGS}
do
case "${_ARG}" in
help)
_show_usage $0 ${_VERSION} ${_CONFIG_FILE} && return 0
_show_usage "${0}" "${_VERSION}" "${_CONFIG_FILE}" && return 0
;;
esac
done
# handle configuration file
[[ -n "${ARG_CONFIG_FILE}" ]] && _CONFIG_FILE="${ARG_CONFIG_FILE}"
if [[ ! -r ${_CONFIG_FILE} ]]
then
warn "unable to read configuration file at ${_CONFIG_FILE}"
return 1
fi
_CFG_WAIT_TIME=$(_CONFIG_FILE="${_CONFIG_FILE}" data_get_lvalue_from_config 'wait_time')
if [[ -z "${_CFG_WAIT_TIME}" ]]
then
# default
_WAIT_TIME=10
log "setting value for parameter wait_time to its default (${_WAIT_TIME})"
else
data_is_numeric "${_CFG_WAIT_TIME}"
# shellcheck disable=SC2181
if (( $? > 0 ))
then
warn "wait time parameter is not numeric in configuration file ${_CONFIG_FILE}"
return 1
else
_WAIT_TIME=${_CFG_WAIT_TIME}
fi
fi
_CFG_CRON_LOG_FILE=$(_CONFIG_FILE="${_CONFIG_FILE}" data_get_lvalue_from_config 'cron_log')
if [[ -z "${_CFG_CRON_LOG_FILE}" ]]
then
# default
_CRON_LOG_FILE="/var/adm/cron/log"
log "setting value for parameter cron_log to its default (${_CRON_LOG_FILE})"
else
_CRON_LOG_FILE="${_CFG_CRON_LOG_FILE}"
log "setting value for parameter cron_log (${_CRON_LOG_FILE})"
fi
_CFG_HEALTHY=$(_CONFIG_FILE="${_CONFIG_FILE}" data_get_lvalue_from_config 'log_healthy')
case "${_CFG_HEALTHY}" in
yes|YES|Yes)
_LOG_HEALTHY=1
;;
*)
# do not override hc_arg
(( _LOG_HEALTHY > 0 )) || _LOG_HEALTHY=0
;;
esac
# log_healthy
(( ARG_LOG_HEALTHY > 0 )) && _LOG_HEALTHY=1
if (( _LOG_HEALTHY > 0 ))
@ -74,10 +125,20 @@ else
log "not logging/showing passed health checks"
fi
# check timeout (_WAIT_TIME must be at least 30 secs smaller than health check timeout)
if (( _WAIT_TIME > 0 ))
then
if (( (_WAIT_TIME + 30) > HC_TIME_OUT ))
then
warn "wait time value will conflict with health check timeout. Specify a (larger) --timeout value"
return 1
fi
fi
# ---- process state ----
# try the pgrep way (note: old pgreps do not support '-c')
(( ARG_DEBUG > 0 )) && debug "checking cron service via pgrep"
(( $(pgrep -u root cron 2>>${HC_STDERR_LOG} | wc -l 2>/dev/null) == 0 )) && _STC=1
(( $(pgrep -u root cron 2>>"${HC_STDERR_LOG}" | wc -l 2>/dev/null) == 0 )) && _STC=1
# evaluate results
case ${_STC} in
@ -99,18 +160,30 @@ then
fi
# ---- log state ----
_AT_BIN="$(command -v at 2>>${HC_STDERR_LOG})"
# check cron log file
if [[ ! -r ${_CRON_LOG_FILE} ]]
then
_MSG="cron log does not exist (${_CRON_LOG_FILE})"
_STC=1
if (( _LOG_HEALTHY > 0 || _STC > 0 ))
then
log_hc "$0" 1 "${_MSG}"
fi
return 0
fi
# create test event
_AT_BIN=$(command -v at 2>>"${HC_STDERR_LOG}")
if [[ -x ${_AT_BIN} && -n "${_AT_BIN}" ]]
then
# start test job
(( ARG_DEBUG > 0 )) && debug "checking cron log via {${_AT_BIN}}"
(echo "*** CHECK LOG ***" >/dev/null | ${_AT_BIN} now) >>${HC_STDOUT_LOG} 2>>${HC_STDERR_LOG}
sleep ${_WAIT_TIME}
if (( $(grep -c 'cron may not be running' ${HC_STDERR_LOG} 2>/dev/null) == 0 ))
(echo "*** CHECK LOG ***" >/dev/null | ${_AT_BIN} now) >>"${HC_STDOUT_LOG}" 2>>"${HC_STDERR_LOG}"
sleep "${_WAIT_TIME}"
if (( $(grep -c 'cron may not be running' "${HC_STDERR_LOG}" 2>/dev/null) == 0 ))
then
# find job results
_JOB_ID=$(grep -E -e '^job' ${HC_STDERR_LOG} 2>/dev/null | awk '{ print $2}' 2>/dev/null)
if [[ -n "${_JOB_ID}" ]] && (( $(grep -c "${_JOB_ID}" ${_CRON_LOG_FILE} 2>/dev/null) > 0 ))
_JOB_ID=$(grep -E -e '^job' "${HC_STDERR_LOG}" 2>/dev/null | awk '{ print $2}' 2>/dev/null)
if [[ -n "${_JOB_ID}" ]] && (( $(grep -c "${_JOB_ID}" "${_CRON_LOG_FILE}" 2>/dev/null) > 0 ))
then
_MSG="cron is logging correctly, schedule via {${_AT_BIN}} OK"
_STC=0
@ -125,7 +198,7 @@ then
else
# check cron log itself
(( ARG_DEBUG > 0 )) && debug "checking cron log via file check"
if [[ -r ${_CRON_LOG_FILE} ]] && [[ -s ${_CRON_LOG_FILE} ]]
if [[ -s ${_CRON_LOG_FILE} ]]
then
_MSG="cron is logging correctly (${_CRON_LOG_FILE})"
_STC=0
@ -148,9 +221,12 @@ function _show_usage
cat <<- EOT
NAME : $1
VERSION : $2
CONFIG : $3
CONFIG : $3 with parameters:
log_healthy=<yes|no>
wait_time=<seconds>
cron_log=<file_path>
PURPOSE : Checks whether cron (CRON) service is running and whether cron is
actually logging to ${_CRON_LOG_FILE}.
actually logging to the cron log file.
LOG HEALTHY : Supported
EOT