Merge pull request #6403 from qzhuyan/backport/william/4.3/systemd-restart-on-failure

Backport/william/4.3/systemd restart on failure

.ci/build_packages/tests.sh

@@ -135,23 +135,6 @@ running_test(){
     pytest -v /paho-mqtt-testing/interoperability/test_client/V5/test_connect.py::test_basic
     # shellcheck disable=SC2009 # pgrep does not support Extended Regular Expressions
     emqx stop || kill "$(ps -ef | grep -E '\-progname\s.+emqx\s' |awk '{print $2}')"
-
-    if [ "$(sed -n '/^ID=/p' /etc/os-release | sed -r 's/ID=(.*)/\1/g' | sed 's/"//g')" = ubuntu ] \
-    || [ "$(sed -n '/^ID=/p' /etc/os-release | sed -r 's/ID=(.*)/\1/g' | sed 's/"//g')" = debian ] ;then
-        service emqx start || ( tail /var/log/emqx/emqx.log.1 && exit 1 )
-        IDLE_TIME=0
-        while ! emqx_ctl status | grep -E 'Node\s.*@.*\sis\sstarted'
-        do
-            if [ $IDLE_TIME -gt 10 ]
-            then
-                echo "emqx service error"
-                exit 1
-            fi
-            sleep 10
-            IDLE_TIME=$((IDLE_TIME+1))
-        done
-        service emqx stop
-    fi
 }
 
 relup_test(){

deploy/packages/deb/Makefile

@@ -13,6 +13,7 @@ TARGET_PKG := $(EMQX_NAME)-$(PKG_VSN)-otp$(OTP_VSN)-$(SYSTEM)-$(ARCH)
 .PHONY: all
 all: | $(BUILT)
 	cp -r debian $(SRCDIR)/
+	cp emqx.service $(SRCDIR)/debian/${EMQX_NAME}.service
 	sed -i "s#<DATETIME>#$(shell date -u '+%a, %d %b %Y %T %z')#g" $(SRCDIR)/debian/changelog
 	sed -i "s#<VERSION>#$(PKG_VSN)#g" $(SRCDIR)/debian/changelog
 	sed -i "s/emqx-pkg/$(EMQX_NAME)-pkg/g" $(SRCDIR)/debian/rules; \

deploy/packages/deb/debian/init.script

@@ -1,150 +0,0 @@
-#! /bin/bash
-### BEGIN INIT INFO
-# Provides:          emqx
-# Required-Start:    $remote_fs $syslog
-# Required-Stop:     $remote_fs $syslog
-# Default-Start:     2 3 4 5
-# Default-Stop:      0 1 6
-# Short-Description: Erlang MQTT Broker
-# Description:       EMQX, a distributed, massively scalable, highly extensible MQTT message broker written in Erlang/OT
-### END INIT INFO
-
-NAME=emqx
-DAEMON=/usr/bin/$NAME
-SCRIPTNAME=/etc/init.d/$NAME
-
-# Read configuration variable file if it is present
-[ -r /etc/default/$NAME ] && . /etc/default/$NAME
-
-# Load the VERBOSE setting and other rcS variables
-. /lib/init/vars.sh
-. /lib/lsb/init-functions
-
-# `service` strips all environmental VARS so
-# if no HOME was set in /etc/default/$NAME then set one here
-# to the data directory for erlexec's sake
-if [ -z "$HOME" ]; then
-    export HOME=/var/lib/emqx
-fi
-
-#
-# Function that starts the daemon/service
-#
-do_start()
-{
-    # Return
-    #   0 if daemon has been started
-    #   1 if daemon was already running
-    #   2 if daemon could not be started
-
-    # Startup with the appropriate user
-    start-stop-daemon --start \
-        --name emqx \
-        --user emqx \
-        --exec $DAEMON -- start \
-        || return 2
-}
-
-#
-# Function that stops the daemon/service
-#
-do_stop()
-{
-    # Identify the erts directory
-    ERTS_PATH=`$DAEMON ertspath`
-
-    # Attempt a clean shutdown.
-    $DAEMON stop
-
-    # waiting stop done sleep 5
-    sleep 5
-    
-    # Return
-    #   0 if daemon has been stopped
-    #   1 if daemon was already stopped
-    #   2 if daemon could not be stopped
-    #   other if a failure occurred
-    # Make sure it's down by using a more direct approach
-    start-stop-daemon --stop \
-                      --quiet \
-                      --retry=TERM/30/KILL/5 \
-                      --user emqx \
-                      --exec $ERTS_PATH/run_erl
-    return $?
-}
-
-#
-# Function that graceful reload the daemon/service
-#
-do_reload() {
-    # Restart the VM without exiting the process
-    $DAEMON restart && return $? || return 2
-}
-
-# Checks the status of a node
-do_status() {
-    $DAEMON ping && echo $"$NAME is running" && return 0
-    echo $"$NAME is stopped" && return 2
-}
-
-case "$1" in
-    start)
-        log_daemon_msg "Starting $NAME"
-        $DAEMON ping >/dev/null 2>&1 && echo $"$NAME is already running" && exit 0
-        do_start
-        case "$?" in
-            0|1) log_end_msg 0 ;;
-            2) log_end_msg 1
-                exit 1
-                ;;
-        esac
-        ;;
-    stop)
-        log_daemon_msg "Stopping $NAME"
-        do_stop
-        case "$?" in
-            0|1) log_end_msg 0 ;;
-            2) log_end_msg 1
-                exit 1
-                ;;
-        esac
-        ;;
-    ping)
-        # See if the VM is alive
-        $DAEMON ping || exit $?
-        ;;
-    reload|force-reload)
-        log_daemon_msg "Reloading $NAME"
-        do_reload
-        ES=$?
-        log_end_msg $ES
-        exit $ES
-        ;;
-    restart)
-        log_daemon_msg "Restarting $NAME"
-        do_stop
-        case "$?" in
-            0|1)
-                do_start
-                case "$?" in
-                    0) log_end_msg 0 ;;
-                    1) log_end_msg 1 && exit 1 ;; # Old process is still running
-                    *) log_end_msg 1 && exit 1 ;; # Failed to start
-                esac
-                ;;
-            *)
-                # Failed to stop
-                log_end_msg 1 && exit 1
-                ;;
-        esac
-        ;;
-    status)
-        do_status && exit 0 || exit $?
-        ;;
-    *)
-        echo "Usage: $SCRIPTNAME {start|stop|ping|restart|force-reload|status}" >&2
-        exit 3
-        ;;
-esac
-
-

deploy/packages/deb/debian/postinst

@@ -5,9 +5,6 @@
 
 set -e
 
-# install startup script
-update-rc.d emqx defaults >/dev/null
-
 # create group
 if ! getent group emqx >/dev/null; then
         addgroup --system emqx
@@ -38,6 +35,12 @@ chmod -R 0755 /usr/lib/emqx/bin
 ln -s /usr/lib/emqx/bin/emqx /usr/bin/emqx
 ln -s /usr/lib/emqx/bin/emqx_ctl /usr/bin/emqx_ctl
 
+if systemctl status --no-pager; then
+    systemctl enable emqx;
+else
+    echo "systemd is not in use, skip enable emqx"
+fi
+
 case "$1" in
     configure)
     ;;

deploy/packages/deb/debian/postrm

@@ -23,9 +23,6 @@ case "$1" in
     purge)
         rm -f /etc/default/emqx
 
-        # ensure we remove the rc.d scripts installed by postinst
-        update-rc.d emqx remove >/dev/null
-
         if [ -d /var/lib/emqx ]; then
                 rm -r /var/lib/emqx
         fi
@@ -58,6 +55,11 @@ case "$1" in
         if [ -d /usr/lib/emqx ]; then
                 rm -r /usr/lib/emqx
         fi
+
+        if systemctl status --no-pager; then
+          systemctl disable emqx;
+          rm -f /etc/systemd/emqx.service
+        fi
     ;;
 
     remove)

deploy/packages/deb/debian/rules

@@ -46,7 +46,9 @@ install: build
 	cp -R releases debian/emqx/usr/lib/emqx
 	cp -R etc/* debian/emqx/etc/emqx
 	cp -R data/*  debian/emqx/var/lib/emqx
-	install -m755 debian/init.script debian/emqx/etc/init.d/emqx
+
+	install -d debian/emqx/lib/systemd/system
+	install -p -m0644 debian/emqx.service debian/emqx/lib/systemd/system/emqx.service
 
 	dh_shlibdeps
 

deploy/packages/deb/emqx.service

@@ -0,0 +1 @@
+../emqx.service

deploy/packages/emqx.service

@@ -0,0 +1,17 @@
+[Unit]
+Description=emqx daemon
+After=network.target
+
+[Service]
+User=emqx
+Group=emqx
+Type=forking
+Environment=HOME=/var/lib/emqx
+ExecStart=/usr/bin/emqx start
+LimitNOFILE=1048576
+ExecStop=/usr/bin/emqx stop
+Restart=on-failure
+RestartSec=5s
+
+[Install]
+WantedBy=multi-user.target

deploy/packages/rpm/emqx.service

@@ -7,9 +7,11 @@ User=emqx
 Group=emqx
 Type=forking
 Environment=HOME=/var/lib/emqx
-ExecStart=/bin/sh /usr/bin/emqx start
+ExecStart=/usr/bin/emqx start
 LimitNOFILE=1048576
-ExecStop=/bin/sh /usr/bin/emqx stop
+ExecStop=/usr/bin/emqx stop
+Restart=on-failure
+RestartSec=5s
 
 [Install]
 WantedBy=multi-user.target