ansible-roles/nagios-nrpe/files/alerts_wrapper

#!/bin/bash
#
# Source:
# https://gitea.evolix.org/evolix/ansible-roles/src/branch/stable/nagios-nrpe
#

VERSION="21.04"
readonly VERSION
readonly base_dir="/var/lib/monitoringctl"

# Default maximum allowed disable time
disable_max_time_default="1d"
readonly disable_max_time_default

# base functions

show_version() {
    cat <<END
alerts_wrapper version ${VERSION}

Copyright 2018-2021 Evolix <info@evolix.fr>,
                    Jérémy Lecour <jlecour@evolix.fr>
                    and others.

alerts_wrapper comes with ABSOLUTELY NO WARRANTY.This is free software,
and you are welcome to redistribute it under certain conditions.
See the GNU General Public License v3.0 for details.
END
}
show_help() {
    cat <<END
alerts_wrapper wraps an NRPE command and overrides the return code.

Usage: alerts_wrapper [--maximum 1d] [--names check,other_disable_name,...]] <check command with optional arguments>
Usage: alerts_wrapper disable_name <check command with optional arguments>

Options
 -m|--max|--maximum    Max age of the disable file.
                       Can be "1d" for 1 day, "5m" for 5 minutes…
                       or more complex expressions like "1w2d10m42s"
                       If no time unit is provided, hour is assumed.
 --limit               (deprecated) Same as above (kept for backward compatibility).
 -n|--names               Disable name (recommanded: set at least the check name).
                       Special name: 'all' is already defined (and cannot be removed).
 --name                (deprecated) Disable name (kept for backward compatibility).
 -h, --help            Print this message and exit.
 -V, --version         Print version and exit.
END
}

time_in_seconds() {
    if echo "${1}" | grep -E -q '^([0-9]+[wdhms])+$'; then
        echo "${1}" | sed 's/w/ * 604800 + /g; s/d/ * 86400 + /g; s/h/ * 3600 + /g; s/m/ * 60 + /g; s/s/ + /g; s/+ $//' | xargs expr
    elif echo "${1}" | grep -E -q '^([0-9]+$)'; then
        echo "${1} * 3600" | xargs expr
    else
        return 1
    fi
}

delay_from_disable_file() {
    # $1: disabled file
    last_change=$(stat -c %Z "$1")

    disable_secs=$(grep -v -E "^\s*#" "${1}" | grep -E "[0-9]+" | head -n1 | awk '{print$1}')
    disable_max_secs=$(time_in_seconds "${disable_max_time}" || time_in_seconds "${disable_max_time_default}")

    if [ "${disable_secs}" -gt "${disable_max_secs}" ]; then
        disable_secs="${disable_max_secs}"
    fi

    disable_date=$(date --date "${disable_secs} seconds ago" +"%s")

    echo $(( last_change - disable_date ))
}

enable_checks() {
    # $1: disable name
    if [ "$(id -u)" -eq "0" ] ; then
        /usr/local/bin/alerts_switch enable "$1"
    else
        sudo /usr/local/bin/alerts_switch enable "$1"
    fi
}

main() {
    check_stdout=$(timeout 9 ${check_command})
    check_rc=$?

    if [ "${check_rc}" -eq 124 ] && [ -z "${check_stdout}" ]; then
        check_stdout="Check timeout (9 sec)"
    fi

    delay=0
    disabled="False"
    for disable_name in ${disable_names}; do
        disable_file="${base_dir}/${disable_name}_alerts_disabled"
        if [ -e "${disable_file}" ]; then
            delay=$(delay_from_disable_file "${disable_file}")

            if [ "${delay}" -le "0" ]; then
                enable_checks "${disable_name}"
            fi
        fi

        if [ -e "${disable_file}" ]; then
            disabled="True"
            formatted_last_change=$(date --date "@$(stat -c %Z "${disable_file}")" +'%c')

            delay_days="$(( delay /86400 ))"
            if [ "${delay_days}" -eq 0 ]; then delay_days=""
            else delay_days="${delay_days}d "; fi

            delay_hours="$(( (delay %86400) /3600 ))"
            if [ "${delay_hours}" -eq 0 ]; then delay_hours=""
            else delay_hours="${delay_hours}h "; fi

            delay_minutes="$(( ((delay %86400) %3600) /60 ))"
            if [ "${delay_minutes}" -eq 0 ]; then delay_minutes=""
            else delay_minutes="${delay_minutes}m "; fi

            delay_seconds="$(( ((delay %86400) %3600) %60 ))"
            if [ "${delay_seconds}" -eq 0 ]; then delay_seconds=""
            else delay_days="${delay_days}s "; fi

            echo "ALERTS DISABLED for ${disable_names} (since ${formatted_last_change}, delay: ${delay_days}${delay_hours}${delay_minutes}${delay_seconds})."
        fi
    done

    echo "${check_stdout}"

    if [ "${disabled}" == "True" ]; then
        if [ ${check_rc} = 0 ]; then
            # Nagios OK
            exit 0
        else
            # Nagios WARNING
            exit 1
        fi
    else
        exit ${check_rc}
    fi
}


if [[ "${1}" =~ -.* ]]; then
    # parse options
    # based on https://gist.github.com/deshion/10d3cb5f88a21671e17a
    while :; do
        case $1 in
            -h|-\?|--help)
                show_help
                exit 0
                ;;
            -V|--version)
                show_version
                exit 0
                ;;

            -m|--max|--maximum|--limit)
                # with value separated by space
                if [ -n "$2" ]; then
                    disable_max_time=$2
                    shift
                else
                    printf 'ERROR: "--maximum" requires a non-empty option argument.\n' >&2
                    exit 1
                fi
                ;;
            -m|--max|--maximum|--limit=?*)
                # with value speparated by =
                disable_max_time=${1#*=}
                ;;
            -m|--max|--maximum|--limit=)
                # without value
                printf 'ERROR: "--maximum" requires a non-empty option argument.\n' >&2
                exit 1
                ;;

            -n|--name|--names)
                # with value separated by space
                if [ -n "$2" ]; then
                    disable_names=$2
                    shift
                else
                    printf 'ERROR: "--name" requires a non-empty option argument.\n' >&2
                    exit 1
                fi
                ;;
            -n|--name=?*|--names=?*)
                # with value speparated by =
                disable_names=${1#*=}
                ;;
            -n|--name=|--names=)
                # without value
                printf 'ERROR: "--name" requires a non-empty option argument.\n' >&2
                exit 1
                ;;

            --)
                # End of all options.
                shift
                break
                ;;
            -?*)
                # ignore unknown options
                printf 'WARN: Unknown option : %s\n' "$1" >&2
                exit 1
                ;;
            *)
                # Default case: If no more options then break out of the loop.
                break
                ;;
        esac

        shift
    done
    # The rest is the command
    check_command="$*"
else
    # no option is passed (backward compatibility with previous version)
    # treat the first argument as disable_names and the rest as the command
    disable_names="${1}"
    shift
    check_command="$*"
fi

# Default values or errors
if [ -z "${disable_max_time}" ]; then
    disable_max_time="${disable_max_time_default}"
fi
#if [ -z "${disable_names}" ]; then
#    printf 'ERROR: You must specify a check name, with --names.\n' >&2
#    exit 1
#fi
if [ -z "${check_command}" ]; then
    printf 'ERROR: You must specify a command to execute.\n' >&2
    exit 1
fi

disable_names="all $(echo "${disable_names}" | tr ',' ' ')"
readonly disable_names
readonly check_command
readonly disable_max_time

main