#!/bin/sh
#
# ocf:pacemaker:HealthCPU resource agent
#
# Copyright 2004-2023 the Pacemaker project contributors
#
# The version control history for this file may have further details.
#
# This source code is licensed under the GNU General Public License version 2
# (GPLv2) WITHOUT ANY WARRANTY.
#

#
# Measures CPUs idling and writes #health-cpu status into the CIB
#
################################
#
# TODO: Enter default values
#       Error handling in getting uptime
#
##################################

#######################################################################
# Initialization:

: ${OCF_FUNCTIONS:="${OCF_ROOT}/resource.d/heartbeat/.ocf-shellfuncs"}
. "${OCF_FUNCTIONS}"
: ${__OCF_ACTION:="$1"}

#######################################################################

meta_data() {
    cat <<END
<?xml version="1.0"?>
<resource-agent name="HealthCPU" version="@VERSION@">
<version>1.1</version>

<longdesc lang="en">
System health agent that measures the CPU idling and updates the #health-cpu attribute.
</longdesc>
<shortdesc lang="en">System health CPU usage</shortdesc>

<parameters>
<parameter name="state" unique-group="state">
<longdesc lang="en">
Location to store the resource state in.
</longdesc>
<shortdesc lang="en">State file</shortdesc>
<content type="string" default="${HA_VARRUN%%/}/health-cpu-${OCF_RESOURCE_INSTANCE}.state" />
</parameter>

<parameter name="yellow_limit" reloadable="1">
<longdesc lang="en">
Lower (!) limit of idle percentage to switch the health attribute to yellow. I.e.
the #health-cpu will go yellow if the %idle of the CPU falls below 50%.
</longdesc>
<shortdesc lang="en">Lower limit for yellow health attribute</shortdesc>
<content type="string" default="50"/>
</parameter>

<parameter name="red_limit" reloadable="1">
<longdesc lang="en">
Lower (!) limit of idle percentage to switch the health attribute to red. I.e.
the #health-cpu will go red if the %idle of the CPU falls below 10%.
</longdesc>
<shortdesc lang="en">Lower limit for red health attribute</shortdesc>
<content type="string" default="10"/>
</parameter>

<parameter name="dampening" reloadable="1">
<longdesc lang="en">
The time to wait (dampening) in seconds for further changes before writing
</longdesc>
<shortdesc lang="en">The time to wait (dampening) in seconds for further changes 
before writing</shortdesc>
<content type="string" default="30s"/>
</parameter>

</parameters>

<actions>
<action name="start"        timeout="10s" />
<action name="stop"         timeout="10s" />
<action name="monitor"      timeout="10s" interval="10s" start-delay="0s" />
<action name="reload-agent" timeout="20s" />
<action name="meta-data"    timeout="5s" />
<action name="validate-all"   timeout="10s" depth="0" />
</actions>
</resource-agent>
END
}

#######################################################################

healthcpu_usage() {
    cat <<END
usage: $0 {start|stop|monitor|reload-agent|validate-all|meta-data}

Expects to have a fully populated OCF RA-compliant environment set.
END
}

healthcpu_start() {
    healthcpu_monitor
    if [ $? -eq $OCF_SUCCESS ]; then
        return $OCF_SUCCESS
    fi
    touch "${OCF_RESKEY_state}"
}

healthcpu_stop() {
    healthcpu_monitor
    if [ $? -eq $OCF_SUCCESS ]; then
        rm "${OCF_RESKEY_state}"
    fi
    return $OCF_SUCCESS
}

healthcpu_monitor() {
    if [ -f "${OCF_RESKEY_state}" ]; then

        IDLE=$(top -b -n2 | grep Cpu | tail -1 | awk -F",|.[0-9][ %]id" '{ print $4 }')
        # echo "System idle: " $IDLE
        # echo "$OCF_RESKEY_red_limit"
        # echo $OCF_RESKEY_yellow_limit

        if [ $IDLE -lt ${OCF_RESKEY_red_limit} ] ; then
            # echo "System state RED!"
            attrd_updater -n "#health-cpu" -B "red" -d "${OCF_RESKEY_dampening}"
            return $OCF_SUCCESS
        fi

        if [ $IDLE -lt ${OCF_RESKEY_yellow_limit} ] ; then
            # echo "System state yellow."
            attrd_updater -n "#health-cpu" -B "yellow" -d "${OCF_RESKEY_dampening}"
        else
            # echo "System state green."
            attrd_updater -n "#health-cpu" -B "green" -d "${OCF_RESKEY_dampening}"
        fi

        return $OCF_SUCCESS
    fi

    return $OCF_NOT_RUNNING
}

healthcpu_reload_agent() {
    return $OCF_SUCCESS
}

healthcpu_validate() {
    # If specified, is yellow_limit a non-negative integer?
    case "$OCF_RESKEY_yellow_limit" in
        0|[1-9]|[1-9][0-9]|100) ;;
        *) return $OCF_ERR_CONFIGURED ;;
    esac

    # If specified, is red_limit a non-negative integer?
    case "$OCF_RESKEY_red_limit" in
        0|[1-9]|[1-9][0-9]|100) ;;
        *) return $OCF_ERR_CONFIGURED ;;
    esac

    # Host-specific checks
    if [ "$OCF_CHECK_LEVEL" = "10" ]; then

        # Is the state directory writable?
        state_dir=$(dirname "$OCF_RESKEY_state")
        [ -d "$state_dir" ] && [ -w "$state_dir" ] && [ -x "$state_dir" ]
        if [ $? -ne 0 ]; then
            ocf_log err "Invalid location for 'state': $state_dir is not writable"
            return $OCF_ERR_ARGS
        fi
    fi

    return $OCF_SUCCESS
}

: ${OCF_RESKEY_CRM_meta_interval:=0}
: ${OCF_RESKEY_CRM_meta_globally_unique:="false"}

if [ -z "$OCF_RESKEY_state" ]; then
    state="${HA_VARRUN%%/}/HealthCPU-${OCF_RESOURCE_INSTANCE}.state"
    if [ "${OCF_RESKEY_CRM_meta_globally_unique}" = "false" ]; then
        # Strip off the trailing clone marker
        OCF_RESKEY_state=$(echo $state | sed s/:[0-9][0-9]*\.state/.state/)
    else
        OCF_RESKEY_state="$state"
    fi
fi

if [ -z "${OCF_RESKEY_red_limit}" ] ; then
    OCF_RESKEY_red_limit=10
fi

if [ -z "${OCF_RESKEY_yellow_limit}" ] ; then
    OCF_RESKEY_yellow_limit=50
fi
if [ -z "${OCF_RESKEY_dampening}" ]; then
    OCF_RESKEY_dampening="30s"
fi

case "$__OCF_ACTION" in
meta-data)      meta_data
                exit $OCF_SUCCESS
                ;;
start)          healthcpu_start;;
stop)           healthcpu_stop;;
monitor)        healthcpu_validate && healthcpu_monitor;;
reload-agent)   healthcpu_reload_agent;;
validate-all)   healthcpu_validate;;
usage|help)     healthcpu_usage
                exit $OCF_SUCCESS
                ;;
*)              healthcpu_usage
                exit $OCF_ERR_UNIMPLEMENTED
                ;;
esac
rc=$?
ocf_log debug "${OCF_RESOURCE_INSTANCE} $__OCF_ACTION : $rc"
exit $rc

# vim: set filetype=sh expandtab tabstop=4 softtabstop=4 shiftwidth=4 textwidth=80: