[Services] Restart Telemetry service upon unexpected critical process…

… exit. (#3768) Signed-off-by: Yong Zhao <yozhao@microsoft.com>
sonic-net · Nov 19, 2019 · df11b2b · df11b2b
1 parent f9e36d3
commit df11b2b
Show file tree

Hide file tree

Showing 5 changed files with 18 additions and 3 deletions.
diff --git a/dockers/docker-sonic-telemetry/Dockerfile.j2 b/dockers/docker-sonic-telemetry/Dockerfile.j2
@@ -35,5 +35,7 @@ RUN apt-get clean -y      && \
 
 COPY ["start.sh", "telemetry.sh", "dialout.sh", "/usr/bin/"]
 COPY ["supervisord.conf", "/etc/supervisor/conf.d/"]
+COPY ["files/supervisor-proc-exit-listener", "/usr/bin"]
+COPY ["critical_processes", "/etc/supervisor"]
 
 ENTRYPOINT ["/usr/bin/supervisord"]
diff --git a/dockers/docker-sonic-telemetry/critical_processes b/dockers/docker-sonic-telemetry/critical_processes
@@ -0,0 +1,2 @@
+telemetry
+dialout
diff --git a/dockers/docker-sonic-telemetry/supervisord.conf b/dockers/docker-sonic-telemetry/supervisord.conf
@@ -3,6 +3,12 @@ logfile_maxbytes=1MB
 logfile_backups=2
 nodaemon=true
 
+[eventlistener:supervisor-proc-exit-listener]
+command=/usr/bin/supervisor-proc-exit-listener
+events=PROCESS_STATE_EXITED
+autostart=always
+autorestart=unexpected
+
 [program:start.sh]
 command=/usr/bin/start.sh
 priority=1
@@ -15,22 +21,22 @@ stderr_logfile=syslog
 command=/usr/sbin/rsyslogd -n
 priority=2
 autostart=false
-autorestart=false
+autorestart=true
 stdout_logfile=syslog
 stderr_logfile=syslog
 
 [program:telemetry]
 command=/usr/bin/telemetry.sh
 priority=3
 autostart=false
-autorestart=true
+autorestart=false
 stdout_logfile=syslog
 stderr_logfile=syslog
 
 [program:dialout]
 command=/usr/bin/dialout.sh
 priority=4
 autostart=false
-autorestart=true
+autorestart=false
 stdout_logfile=syslog
 stderr_logfile=syslog
diff --git a/files/build_templates/telemetry.service.j2 b/files/build_templates/telemetry.service.j2
@@ -3,12 +3,16 @@ Description=Telemetry container
 Requires=database.service
 After=database.service swss.service syncd.service
 Before=ntp-config.service
+StartLimitIntervalSec=1200
+StartLimitBurst=3
 
 [Service]
 User={{ sonicadmin_user }}
 ExecStartPre=/usr/bin/{{docker_container_name}}.sh start
 ExecStart=/usr/bin/{{docker_container_name}}.sh wait
 ExecStop=/usr/bin/{{docker_container_name}}.sh stop
+Restart=always
+RestartSec=30
 
 [Install]
 WantedBy=multi-user.target
diff --git a/rules/telemetry.mk b/rules/telemetry.mk
@@ -3,3 +3,4 @@
 SONIC_TELEMETRY = sonic-telemetry_0.1_$(CONFIGURED_ARCH).deb
 $(SONIC_TELEMETRY)_SRC_PATH = $(SRC_PATH)/telemetry
 SONIC_DPKG_DEBS += $(SONIC_TELEMETRY)
+$(SONIC_TELEMETRY)_FILES += $(SUPERVISOR_PROC_EXIT_LISTENER_SCRIPT)