2

I have an application written in python which I've used many times with CentOS 6, and am now setting it up on CentOS 7. I've created a systemd service file that contains the following:

#
# Systemd unit file for Application
#

[Unit]
Description=Application Daemon

[Service]
User=svc_app
Group=svc_app

Type=forking
ExecStart=/usr/bin/python /apps/application/run_app.py --daemon --pidfile=/var/run/apps/application.pid
ExecStop=pkill --pidfile /var/run/apps/application.pid
PIDFile=/var/run/apps/application.pid

[Install]
WantedBy=multi-user.target

If I run systemctl start application.service, the app starts. If I run systemctl stop application.service, the app stops. If I run systemctl status application.service, I get the correct status.

However, I'm experiencing two log messages that are concerning, and not sure why I'm getting them.

Jun 10 18:36:43 centos7-test systemd[1]: Starting Application Daemon...
Jun 10 18:36:44 centos7-test systemd[1]: PID file /var/run/apps/application.pid not readable (yet?) after start.
Jun 10 18:36:44 centos7-test systemd[1]: Started Application Daemon.

App starts, but the message application.pid not readable had me concerned. Should I be concerned about this, or can I ignore it? My application is successfully creating a pid file that contains the pid of the running instance, so I know that's working correctly.

Second issue is when I run systemctl stop application.service. 

Jun 10 18:49:33 centos7-test systemd[1]: Stopping Application Daemon...
Jun 10 18:49:43 centos7-test systemd[1]: Stopped Application Daemon.
Jun 10 18:49:43 centos7-test systemd[1]: [/etc/systemd/system/application.service:14] Executable path is not absolute, ignoring: pkill --pidfile /var/run/apps/application.pid
Jun 10 18:49:43 centos7-test systemd[1]: [/etc/systemd/system/application.service:14] Executable path is not absolute, ignoring: pkill --pidfile /var/run/apps/application.pid
Jun 10 18:49:54 centos7-test systemd[1]: [/etc/systemd/system/application.service:14] Executable path is not absolute, ignoring: pkill --pidfile /var/run/apps/application.pid

Now the service is dead, pidfile is gone. So it stopped, but those messages concern me. What's going on, or where else can I look? Is there something other than pkill I should use to stop my python app? on CentOS 6 I used killproc, but that doesn't seem to be available in CentOS 7 by default so I switched to pkill.

Thanks in advance!

tycoonbob
  • 21
  • 1
  • 1
  • 2
  • Your `ExecStop` did nothing. `Systemd` knows how to kill services (by killing everything in the relevant `cgroup`). We no longer have to resort to killing by PIDs contained in files. Just remove your `ExecStop`. – jordanm Jun 11 '15 at 00:20
  • @jordanm Huh, you're absolutely right. I removed my `ExecStop` and the service was still stopping as it should, and I was no longer getting that message. Any idea on that other message, about the pid not readable after start? Should I be concerned with this? – tycoonbob Jun 11 '15 at 13:09
  • @tycoonbob – If I had to take a shot, I'd say the first warning is caused by your app taking too long to create its pidfile, and as for the second, it was right in front of your nose: *__Executable path is not absolute__, ignoring: __pkill__ --pidfile /var/run/apps/application.pid*. The given command, `pkill`, is not defined by an absolute path. You needed `/usr/bin/pkill`, or perhaps `/bin/pkill`. And, if your app catches `SIGTERM` and deletes the pidfile as part of its cleanup procedure, systemd probably would have moaned again when it no longer existed after `ExecStop`. – Adrian Günter Dec 15 '15 at 19:52

0 Answers0