X-Git-Url: https://arthur.barton.de/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=plugins.d%2Falarm-notify.sh;h=d0188fe3b11b5361810b62a791ece461f43d6d49;hb=HEAD;hp=28f550f306a59d1fc06759b0ad0cecc044549f4a;hpb=072e244f93439abf6e0d0ef109e606bec42350c0;p=netdata.git diff --git a/plugins.d/alarm-notify.sh b/plugins.d/alarm-notify.sh index 28f550f3..d0188fe3 100755 --- a/plugins.d/alarm-notify.sh +++ b/plugins.d/alarm-notify.sh @@ -14,12 +14,17 @@ # - severity filtering per recipient # # Supported notification methods: -# - emails -# - slack.com notifications -# - pushover.net notifications +# - emails by @ktsaou +# - slack.com notifications by @ktsaou +# - discordapp.com notifications by @lowfive +# - pushover.net notifications by @ktsaou # - pushbullet.com push notifications by Tiago Peralta @tperalta82 PR #1070 # - telegram.org notifications by @hashworks PR #1002 -# +# - twilio.com notifications by Levi Blaney @shadycuz PR #1211 +# - kafka notifications by @ktsaou #1342 +# - pagerduty.com notifications by Jim Cooley @jimcooley PR #1373 +# - messagebird.com notifications by @tech_no_logical #1453 +# - hipchat notifications by @ktsaou #1561 # ----------------------------------------------------------------------------- # testing notifications @@ -35,14 +40,26 @@ then [ -z "${recipient}" ] && recipient="sysadmin" - echo >&2 ">> SENDING TEST ALARM TO ROLE: ${recipient} <<" + id=1 + last="CLEAR" + for x in "CRITICAL" "WARNING" "CLEAR" + do + echo >&2 + echo >&2 "# SENDING TEST ${x} ALARM TO ROLE: ${recipient}" + + "${0}" "${recipient}" "$(hostname)" 1 1 "${id}" "$(date +%s)" "test_alarm" "test.chart" "test.family" "${x}" "${last}" 100 90 "${0}" 1 $((0 + id)) "units" "this is a test alarm to verify notifications work" "new value" "old value" + if [ $? -ne 0 ] + then + echo >&2 "# FAILED" + else + echo >&2 "# OK" + fi - "${0}" "${recipient}" 'test' '1' '1' '1' '1' 'test_alarm' 'test_alarm' 'test' 'CRITICAL' 'CLEAR' '1' '1' "${0}" '1' '1' 'number' 'test alarm to verify notifications work' - ret=$? + last="${x}" + id=$((id + 1)) + done - [ ${ret} -ne 0 ] && echo >&2 ">> FAILED <<" && exit ${ret} - echo >&2 ">> OK <<" - exit $? + exit 1 fi export PATH="${PATH}:/sbin:/usr/sbin:/usr/local/sbin" @@ -83,7 +100,7 @@ fatal() { debug=0 debug() { - [ $debug -eq 1 ] && log DEBUG "${@}" + [ ${debug} -eq 1 ] && log DEBUG "${@}" } # ----------------------------------------------------------------------------- @@ -122,6 +139,15 @@ duration="${15}" # the duration in seconds of the previous alarm state non_clear_duration="${16}" # the total duration in seconds this is/was non-clear units="${17}" # the units of the value info="${18}" # a short description of the alarm +value_string="${19}" # friendly value (with units) +old_value_string="${20}" # friendly old value (with units) + +# ----------------------------------------------------------------------------- +# find a suitable hostname to use, if netdata did not supply a hostname + +[ -z "${host}" ] && host="${NETDATA_HOSTNAME}" +[ -z "${host}" ] && host="${NETDATA_REGISTRY_HOSTNAME}" +[ -z "${host}" ] && host="$(hostname 2>/dev/null)" # ----------------------------------------------------------------------------- # screen statuses we don't need to send a notification @@ -129,14 +155,14 @@ info="${18}" # a short description of the alarm # don't do anything if this is not WARNING, CRITICAL or CLEAR if [ "${status}" != "WARNING" -a "${status}" != "CRITICAL" -a "${status}" != "CLEAR" ] then - info "not sending notification for ${status} on '${chart}.${name}'" + info "not sending notification for ${status} of '${host}.${chart}.${name}'" exit 1 fi # don't do anything if this is CLEAR, but it was not WARNING or CRITICAL if [ "${old_status}" != "WARNING" -a "${old_status}" != "CRITICAL" -a "${status}" = "CLEAR" ] then - info "not sending notification for ${status} on '${chart}.${name}' (last status was ${old_status})" + info "not sending notification for ${status} of '${host}.${chart}.${name}' (last status was ${old_status})" exit 1 fi @@ -156,16 +182,27 @@ sendmail= # enable / disable features SEND_SLACK="YES" +SEND_DISCORD="YES" SEND_PUSHOVER="YES" +SEND_TWILIO="YES" +SEND_HIPCHAT="YES" +SEND_MESSAGEBIRD="YES" SEND_TELEGRAM="YES" SEND_EMAIL="YES" SEND_PUSHBULLET="YES" +SEND_KAFKA="YES" +SEND_PD="YES" # slack configs SLACK_WEBHOOK_URL= DEFAULT_RECIPIENT_SLACK= declare -A role_recipients_slack=() +# discord configs +DISCORD_WEBHOOK_URL= +DEFAULT_RECIPIENT_DISCORD= +declare -A role_recipients_discord=() + # pushover configs PUSHOVER_APP_TOKEN= DEFAULT_RECIPIENT_PUSHOVER= @@ -176,11 +213,38 @@ PUSHBULLET_ACCESS_TOKEN= DEFAULT_RECIPIENT_PUSHBULLET= declare -A role_recipients_pushbullet=() +# twilio configs +TWILIO_ACCOUNT_SID= +TWILIO_ACCOUNT_TOKEN= +TWILIO_NUMBER= +DEFAULT_RECIPIENT_TWILIO= +declare -A role_recipients_twilio=() + +# hipchat configs +HIPCHAT_SERVER= +HIPCHAT_AUTH_TOKEN= +DEFAULT_RECIPIENT_HIPCHAT= +declare -A role_recipients_hipchat=() + +# messagebird configs +MESSAGEBIRD_ACCESS_KEY= +MESSAGEBIRD_NUMBER= +DEFAULT_RECIPIENT_MESSAGEBIRD= +declare -A role_recipients_messagebird=() + # telegram configs TELEGRAM_BOT_TOKEN= DEFAULT_RECIPIENT_TELEGRAM= declare -A role_recipients_telegram=() +# kafka configs +KAFKA_URL= +KAFKA_SENDER_IP= + +# pagerduty.com configs +PD_SERVICE_KEY= +declare -A role_recipients_pd=() + # email configs DEFAULT_RECIPIENT_EMAIL="root" declare -A role_recipients_email=() @@ -236,9 +300,13 @@ filter_recipient_by_criticality() { # find the recipients' addresses per method declare -A arr_slack=() +declare -A arr_discord=() declare -A arr_pushover=() declare -A arr_pushbullet=() +declare -A arr_twilio=() +declare -A arr_hipchat=() declare -A arr_telegram=() +declare -A arr_pd=() declare -A arr_email=() # netdata may call us with multiple roles, and roles may have multiple but @@ -273,6 +341,30 @@ do [ "${r}" != "disabled" ] && filter_recipient_by_criticality pushbullet "${r}" && arr_pushbullet[${r/|*/}]="1" done + # twilio + a="${role_recipients_twilio[${x}]}" + [ -z "${a}" ] && a="${DEFAULT_RECIPIENT_TWILIO}" + for r in ${a//,/ } + do + [ "${r}" != "disabled" ] && filter_recipient_by_criticality twilio "${r}" && arr_twilio[${r/|*/}]="1" + done + + # hipchat + a="${role_recipients_hipchat[${x}]}" + [ -z "${a}" ] && a="${DEFAULT_RECIPIENT_HIPCHAT}" + for r in ${a//,/ } + do + [ "${r}" != "disabled" ] && filter_recipient_by_criticality hipchat "${r}" && arr_hipchat[${r/|*/}]="1" + done + + # messagebird + a="${role_recipients_messagebird[${x}]}" + [ -z "${a}" ] && a="${DEFAULT_RECIPIENT_MESSAGEBIRD}" + for r in ${a//,/ } + do + [ "${r}" != "disabled" ] && filter_recipient_by_criticality messagebird "${r}" && arr_messagebird[${r/|*/}]="1" + done + # telegram a="${role_recipients_telegram[${x}]}" [ -z "${a}" ] && a="${DEFAULT_RECIPIENT_TELEGRAM}" @@ -288,12 +380,32 @@ do do [ "${r}" != "disabled" ] && filter_recipient_by_criticality slack "${r}" && arr_slack[${r/|*/}]="1" done + + # discord + a="${role_recipients_discord[${x}]}" + [ -z "${a}" ] && a="${DEFAULT_RECIPIENT_DISCORD}" + for r in ${a//,/ } + do + [ "${r}" != "disabled" ] && filter_recipient_by_criticality discord "${r}" && arr_discord[${r/|*/}]="1" + done + + # pagerduty.com + a="${role_recipients_pd[${x}]}" + [ -z "${a}" ] && a="${DEFAULT_RECIPIENT_PD}" + for r in ${a//,/ } + do + [ "${r}" != "disabled" ] && filter_recipient_by_criticality pd "${r}" && arr_pd[${r/|*/}]="1" + done done # build the list of slack recipients (channels) to_slack="${!arr_slack[*]}" [ -z "${to_slack}" ] && SEND_SLACK="NO" +# build the list of discord recipients (channels) +to_discord="${!arr_discord[*]}" +[ -z "${to_discord}" ] && SEND_DISCORD="NO" + # build the list of pushover recipients (user tokens) to_pushover="${!arr_pushover[*]}" [ -z "${to_pushover}" ] && SEND_PUSHOVER="NO" @@ -302,10 +414,26 @@ to_pushover="${!arr_pushover[*]}" to_pushbullet="${!arr_pushbullet[*]}" [ -z "${to_pushbullet}" ] && SEND_PUSHBULLET="NO" +# build the list of twilio recipients (phone numbers) +to_twilio="${!arr_twilio[*]}" +[ -z "${to_twilio}" ] && SEND_TWILIO="NO" + +# build the list of hipchat recipients (rooms) +to_hipchat="${!arr_hipchat[*]}" +[ -z "${to_hipchat}" ] && SEND_HIPCHAT="NO" + +# build the list of messagebird recipients (phone numbers) +to_messagebird="${!arr_messagebird[*]}" +[ -z "${to_messagebird}" ] && SEND_MESSAGEBIRD="NO" + # check array of telegram recipients (chat ids) to_telegram="${!arr_telegram[*]}" [ -z "${to_telegram}" ] && SEND_TELEGRAM="NO" +# build the list of pagerduty recipients (service keys) +to_pd="${!arr_pd[*]}" +[ -z "${to_pd}" ] && SEND_PD="NO" + # build the list of email recipients (email addresses) to_email= for x in "${!arr_email[@]}" @@ -322,27 +450,75 @@ done # check slack [ -z "${SLACK_WEBHOOK_URL}" ] && SEND_SLACK="NO" +# check discord +[ -z "${DISCORD_WEBHOOK_URL}" ] && SEND_DISCORD="NO" + # check pushover [ -z "${PUSHOVER_APP_TOKEN}" ] && SEND_PUSHOVER="NO" # check pushbullet -[ -z "${DEFAULT_RECIPIENT_PUSHBULLET}" ] && SEND_PUSHBULLET="NO" +[ -z "${PUSHBULLET_ACCESS_TOKEN}" ] && SEND_PUSHBULLET="NO" + +# check twilio +[ -z "${TWILIO_ACCOUNT_TOKEN}" -o -z "${TWILIO_ACCOUNT_SID}" -o -z "${TWILIO_NUMBER}" ] && SEND_TWILIO="NO" + +# check hipchat +[ -z "${HIPCHAT_AUTH_TOKEN}" ] && SEND_HIPCHAT="NO" + +# check messagebird +[ -z "${MESSAGEBIRD_ACCESS_KEY}" -o -z "${MESSAGEBIRD_NUMBER}" ] && SEND_MESSAGEBIRD="NO" # check telegram [ -z "${TELEGRAM_BOT_TOKEN}" ] && SEND_TELEGRAM="NO" -if [ \( "${SEND_PUSHOVER}" = "YES" -o "${SEND_SLACK}" = "YES" -o "${SEND_TELEGRAM}" = "YES" -o "${SEND_PUSHBULLET}" = "YES" \) -a -z "${curl}" ] +# check kafka +[ -z "${KAFKA_URL}" -o -z "${KAFKA_SENDER_IP}" ] && SEND_KAFKA="NO" + +# check pagerduty.com +# if we need pd-send, check for the pd-send command +# https://www.pagerduty.com/docs/guides/agent-install-guide/ +if [ "${SEND_PD}" = "YES" ] + then + pd_send="$(which pd-send 2>/dev/null || command -v pd-send 2>/dev/null)" + if [ -z "${pd_send}" ] + then + # no pd-send available + # disable pagerduty.com + SEND_PD="NO" + fi +fi + +# if we need curl, check for the curl command +if [ \( \ + "${SEND_PUSHOVER}" = "YES" \ + -o "${SEND_SLACK}" = "YES" \ + -o "${SEND_DISCORD}" = "YES" \ + -o "${SEND_HIPCHAT}" = "YES" \ + -o "${SEND_TWILIO}" = "YES" \ + -o "${SEND_MESSAGEBIRD}" = "YES" \ + -o "${SEND_TELEGRAM}" = "YES" \ + -o "${SEND_PUSHBULLET}" = "YES" \ + -o "${SEND_KAFKA}" = "YES" \ + \) -a -z "${curl}" ] then curl="$(which curl 2>/dev/null || command -v curl 2>/dev/null)" if [ -z "${curl}" ] then + # no curl available + # disable all curl based methods SEND_PUSHOVER="NO" SEND_PUSHBULLET="NO" SEND_TELEGRAM="NO" SEND_SLACK="NO" + SEND_DISCORD="NO" + SEND_TWILIO="NO" + SEND_HIPCHAT="NO" + SEND_MESSAGEBIRD="NO" + SEND_KAFKA="NO" fi fi +# if we need sendmail, check for the sendmail command if [ "${SEND_EMAIL}" = "YES" -a -z "${sendmail}" ] then sendmail="$(which sendmail 2>/dev/null || command -v sendmail 2>/dev/null)" @@ -350,18 +526,22 @@ if [ "${SEND_EMAIL}" = "YES" -a -z "${sendmail}" ] fi # check that we have at least a method enabled -if [ "${SEND_EMAIL}" != "YES" -a "${SEND_PUSHOVER}" != "YES" -a "${SEND_TELEGRAM}" != "YES" -a "${SEND_SLACK}" != "YES" -a "${SEND_PUSHBULLET}" != "YES" ] +if [ "${SEND_EMAIL}" != "YES" \ + -a "${SEND_PUSHOVER}" != "YES" \ + -a "${SEND_TELEGRAM}" != "YES" \ + -a "${SEND_SLACK}" != "YES" \ + -a "${SEND_DISCORD}" != "YES" \ + -a "${SEND_TWILIO}" != "YES" \ + -a "${SEND_HIPCHAT}" != "YES" \ + -a "${SEND_MESSAGEBIRD}" != "YES" \ + -a "${SEND_PUSHBULLET}" != "YES" \ + -a "${SEND_KAFKA}" != "YES" \ + -a "${SEND_PD}" != "YES" \ + ] then - fatal "All notification methods are disabled. Not sending a notification." + fatal "All notification methods are disabled. Not sending notification for host '${host}', chart '${chart}' to '${roles}' for '${name}' = '${value}' for status '${status}'." fi -# ----------------------------------------------------------------------------- -# get the system hostname - -[ -z "${host}" ] && host="${NETDATA_HOSTNAME}" -[ -z "${host}" ] && host="${NETDATA_REGISTRY_HOSTNAME}" -[ -z "${host}" ] && host="$(hostname 2>/dev/null)" - # ----------------------------------------------------------------------------- # get the date the alarm happened @@ -369,7 +549,7 @@ date="$(date --date=@${when} 2>/dev/null)" [ -z "${date}" ] && date="$(date 2>/dev/null)" # ----------------------------------------------------------------------------- -# URL encode a string +# function to URL encode a string urlencode() { local string="${1}" strlen encoded pos c o @@ -377,14 +557,14 @@ urlencode() { strlen=${#string} for (( pos=0 ; pos \ +${alarm} ${info_html}
\ +${chart} (family ${family})
\ +${date}${raised_for_html}
\ +View netdata dashboard \ +(source of alarm ${src}) \ +" + +SENT_HIPCHAT=$? + # ----------------------------------------------------------------------------- # send the email @@ -892,8 +1393,23 @@ SENT_EMAIL=$? # ----------------------------------------------------------------------------- # let netdata know -# we did send something -[ ${SENT_EMAIL} -eq 0 -o ${SENT_PUSHOVER} -eq 0 -o ${SENT_TELEGRAM} -eq 0 -o ${SENT_SLACK} -eq 0 -o ${SENT_PUSHBULLET} -eq 0 ] && exit 0 +if [ ${SENT_EMAIL} -eq 0 \ + -o ${SENT_PUSHOVER} -eq 0 \ + -o ${SENT_TELEGRAM} -eq 0 \ + -o ${SENT_SLACK} -eq 0 \ + -o ${SENT_DISCORD} -eq 0 \ + -o ${SENT_TWILIO} -eq 0 \ + -o ${SENT_HIPCHAT} -eq 0 \ + -o ${SENT_MESSAGEBIRD} -eq 0 \ + -o ${SENT_PUSHBULLET} -eq 0 \ + -o ${SENT_KAFKA} -eq 0 \ + -o ${SENT_PD} -eq 0 \ + ] + then + # we did send something + exit 0 +fi # we did not send anything exit 1 +