I have a systemd service called sitesdb-job-runner.service
with the following config:
[Unit]
Description=SitesDB Background Job Runner
After=network.target mariadb.service
[Service]
Type=simple
Restart=always
User=sitesdb
ExecStart=/bin/scl enable rh-git29 rh-php71 "/var/www/sitesdb/current/bin/console --env=prod jms-job-queue:run -v"
[Install]
WantedBy=multi-user.target
When I run systemctl stop sitesdb-job-runner
, a SIGTERM is not sent to my process. I know this because the process logs when it receives a SIGTERM so it can gracefully shut down.
After some investigation, I suspect this is related to the scl
wrapper I'm using to actually execute the script. When I run /bin/scl enable rh-git29 rh-php71 "/var/www/sitesdb/current/bin/console --env=prod jms-job-queue:run -v"
directly, I get these three processes spawned:
(1) /bin/scl enable rh-git29 rh-php71 /var/www/sitesdb/current/bin/console --env=prod jms-job-queue:run
(2) /bin/bash /var/tmp/sclSdT0wu
(3) php /var/www/sitesdb/current/bin/console --env=prod jms-job-queue:run -v
If I manually target the PID of the third process (my actual command) and sent it a SIGTERM, it works as expected. But when I sent a SIGTERM to the parent process, my process does not get the SIGTERM.
My understanding of the default behavior of systemd is that it will send a SIGTERM to all processes and child processes, but I guess something else is going on here.
Any ideas?