Merge pull request #1373 from jimcooley/master

pagerduty.com notifications

diff --git a/conf.d/health_alarm_notify.conf b/conf.d/health_alarm_notify.conf
index aa73c948a1b56957d021274eeb7f3345d6bbd0a2..dea4ea0511244604f62dae9f996b69a29af6545e 100644 (file)
--- a/conf.d/health_alarm_notify.conf
+++ b/conf.d/health_alarm_notify.conf
@@ -9,6 +9,7 @@
 # - messages to your slack team (slack.com),
 # - messages to your telegram chat / group chat (telegram.org)
 # - sms messages to your cell phone or any sms enabled device (twilio.com)
+# - notifications to users on pagerduty.com
 #
 # The 'to' line given at netdata alarms defines a *role*, so that many
 # people can be notified for each role.
@@ -62,6 +63,7 @@ curl=""
 #  - telegram chat ids
 #  - slack channels
 #  - sms phone numbers
+#  - pagerduty.com (pd) services
 #
 # You can append |critical to limit the notifications to be sent.
 #
@@ -73,9 +75,10 @@ curl=""
 #  telegram: "111827421 112746832|critical"
 #  slack   : "alarms disasters|critical"
 #  twilio  : "+15555555555 +17777777777|critical"
+#  pd      : "<pd_service_key_1> <pd_service_key_2>|critical"
 #
 # If a recipient is set to empty string, the default recipient of the given
-# notification method (email, pushover, telegram, slack) will be used.
+# notification method (email, pushover, telegram, slack, pd) will be used.
 # To disable a notification, use the recipient called: disabled
 # This works for all notification methods (including the default recipients).
 
@@ -212,6 +215,25 @@ KAFKA_URL=""
 KAFKA_SENDER_IP=""
 
 
+#------------------------------------------------------------------------------
+# pagerduty.com notification options
+#
+# pagerduty.com notifications require the pagerduty agent to be installed and 
+# a "Generic API" pagerduty service.
+# https://www.pagerduty.com/docs/guides/agent-install-guide/
+
+# multiple recipients can be given like this:
+#              "<pd_service_key_1> <pd_service_key_2> ..."
+
+# enable/disable sending pagerduty notifications
+SEND_PD="YES"
+
+# if a role's recipients are not configured, a notification will be sent to
+# the "General API" pagerduty.com service that uses this service key.
+# (empty = do not send a notification for unconfigured roles):
+DEFAULT_RECIPIENT_PD=""
+
+
 ###############################################################################
 # RECIPIENTS PER ROLE
 
@@ -231,6 +253,8 @@ role_recipients_slack[sysadmin]="${DEFAULT_RECIPIENT_SLACK}"
 
 role_recipients_twilio[sysadmin]="${DEFAULT_RECIPIENT_TWILIO}"
 
+role_recipients_pd[sysadmin]="${DEFAULT_RECIPIENT_PD}"
+
 # -----------------------------------------------------------------------------
 # DNS related alarms
 
@@ -246,6 +270,8 @@ role_recipients_slack[domainadmin]="${DEFAULT_RECIPIENT_SLACK}"
 
 role_recipients_twilio[domainadmin]="${DEFAULT_RECIPIENT_TWILIO}"
 
+role_recipients_pd[domainadmin]="${DEFAULT_RECIPIENT_PD}"
+
 # -----------------------------------------------------------------------------
 # database servers alarms
 # mysql, redis, memcached, etc
@@ -262,6 +288,8 @@ role_recipients_slack[dba]="${DEFAULT_RECIPIENT_SLACK}"
 
 role_recipients_twilio[dba]="${DEFAULT_RECIPIENT_TWILIO}"
 
+role_recipients_pd[dba]="${DEFAULT_RECIPIENT_PD}"
+
 # -----------------------------------------------------------------------------
 # web servers alarms
 # apache, nginx, etc
@@ -278,6 +306,8 @@ role_recipients_slack[webmaster]="${DEFAULT_RECIPIENT_SLACK}"
 
 role_recipients_twilio[webmaster]="${DEFAULT_RECIPIENT_TWILIO}"
 
+role_recipients_pd[webmaster]="${DEFAULT_RECIPIENT_PD}"
+
 # -----------------------------------------------------------------------------
 # proxy servers alarms
 # apache, nginx, etc
@@ -293,3 +323,5 @@ role_recipients_telegram[proxyadmin]="${DEFAULT_RECIPIENT_TELEGRAM}"
 role_recipients_slack[proxyadmin]="${DEFAULT_RECIPIENT_SLACK}"
 
 role_recipients_twilio[proxyadmin]="${DEFAULT_RECIPIENT_TWILIO}"
+
+role_recipients_pd[proxyadmin]="${DEFAULT_RECIPIENT_PD}"

diff --git a/plugins.d/alarm-notify.sh b/plugins.d/alarm-notify.sh
index 17b009c965a0c7ad46d0b9993a96d7307786dd67..9f5ca1b09d0aae71ba2fc1fdfa406c5d06421aa4 100755 (executable)
--- a/plugins.d/alarm-notify.sh
+++ b/plugins.d/alarm-notify.sh
@@ -21,6 +21,7 @@
 #  - telegram.org notifications by @hashworks PR #1002
 #  - twilio.com notifications by Levi Blaney @shadycuz PR #1211
 #  - kafka notifications
+#  - pagerduty.com notifications by Jim Cooley @jimcooley PR #1373
 
 # -----------------------------------------------------------------------------
 # testing notifications
@@ -175,6 +176,7 @@ SEND_TELEGRAM="YES"
 SEND_EMAIL="YES"
 SEND_PUSHBULLET="YES"
 SEND_KAFKA="YES"
+SEND_PD="YES"
 
 # slack configs
 SLACK_WEBHOOK_URL=
@@ -207,6 +209,10 @@ declare -A role_recipients_telegram=()
 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=()
@@ -266,6 +272,7 @@ declare -A arr_pushover=()
 declare -A arr_pushbullet=()
 declare -A arr_twilio=()
 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
@@ -323,6 +330,14 @@ do
     do
         [ "${r}" != "disabled" ] && filter_recipient_by_criticality slack "${r}" && arr_slack[${r/|*/}]="1"
     done
+
+    # pagerduty.com
+    a="${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)
@@ -345,6 +360,10 @@ to_twilio="${!arr_twilio[*]}"
 to_telegram="${!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 "${!arr_email[@]}"
@@ -376,6 +395,20 @@ done
 # check kafka
 [ -z "${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 [ \( "${SEND_PUSHOVER}" = "YES" -o "${SEND_SLACK}" = "YES" -o "${SEND_TWILIO}" = "YES" -o "${SEND_TELEGRAM}" = "YES" -o "${SEND_PUSHBULLET}" = "YES" -o "${SEND_KAFKA}" = "YES" \) -a -z "${curl}" ]
     then
@@ -408,9 +441,10 @@ if [   "${SEND_EMAIL}"      != "YES" \
     -a "${SEND_TWILIO}"     != "YES" \
     -a "${SEND_PUSHBULLET}" != "YES" \
     -a "${SEND_KAFKA}"      != "YES" \
+    -a "${SEND_PD}"         != "YES" \
     ]
     then
-    fatal "All notification methods are disabled. Not sending notification to '${role}' for '${name}' = '${value}' of chart '${chart}' for status '${status}'."
+    fatal "All notification methods are disabled. Not sending notification to '${roles}' for '${name}' = '${value}' of chart '${chart}' for status '${status}'."
 fi
 
 # -----------------------------------------------------------------------------
@@ -638,6 +672,62 @@ send_kafka() {
     return 1
 }
 
+# -----------------------------------------------------------------------------
+# pagerduty.com sender
+
+send_pd() {
+    local recipients="${1}" sent=0
+    unset t
+    case ${status} in
+        CLEAR)    t='resolve';;
+        WARNING)  t='trigger';;
+        CRITICAL) t='trigger';;
+    esac
+
+    if [ ${SEND_PD} = "YES" -a ! -z "${t}" ]
+        then
+        for PD_SERVICE_KEY in ${recipients}
+        do
+            d="${status} ${name}=${value} ${units} - ${host}, ${family}"
+            ${pd_send} -k ${PD_SERVICE_KEY} \
+                       -t ${t} \
+                       -d "${d}" \
+                       -i ${alarm_id} \
+                       -f 'info'="${info}" \
+                       -f 'value_w_units'="${value} ${units}" \
+                       -f 'when'="${when}" \
+                       -f 'duration'="${duration}" \
+                       -f 'roles'="${roles}" \
+                       -f 'host'="${host}" \
+                       -f 'unique_id'="${unique_id}" \
+                       -f 'alarm_id'="${alarm_id}" \
+                       -f 'event_id'="${event_id}" \
+                       -f 'name'="${name}" \
+                       -f 'chart'="${chart}" \
+                       -f 'family'="${family}" \
+                       -f 'status'="${status}" \
+                       -f 'old_status'="${old_status}" \
+                       -f 'value'="${value}" \
+                       -f 'old_value'="${old_value}" \
+                       -f 'src'="${src}" \
+                       -f 'non_clear_duration'="${non_clear_duration}" \
+                       -f 'units'="${units}"
+            retval=$?
+            if [ ${retval} -eq 0 ]
+                then
+                    info "sent pagerduty.com notification using service key ${PD_SERVICE_KEY::-26}....: ${d}"
+                    sent=$((sent + 1))
+                else
+                    error "failed to send pagerduty.com notification using service key ${PD_SERVICE_KEY::-26}.... (error code ${retval}): ${d}"
+            fi
+        done
+
+        [ ${sent} -gt 0 ] && return 0
+    fi
+
+    return 1
+}
+
 # -----------------------------------------------------------------------------
 # twilio sender
 
@@ -928,6 +1018,13 @@ send_kafka
 SENT_KAFKA=$?
 
 
+# -----------------------------------------------------------------------------
+# send the pagerduty.com message
+
+send_pd "${to_pd}"
+SENT_PD=$?
+
+
 # -----------------------------------------------------------------------------
 # send the email
 
@@ -1033,6 +1130,7 @@ if [   ${SENT_EMAIL}      -eq 0 \
     -o ${SENT_TWILIO}     -eq 0 \
     -o ${SENT_PUSHBULLET} -eq 0 \
     -o ${SENT_KAFKA}      -eq 0 \
+    -o ${SENT_PD}         -eq 0 \
     ]
     then
     # we did send something