Unexpected SIGTERM stops Docker standalone-chrome container

[Aditya Joshi]

Hi, I am unable to keep the Docker Selenium StandaloneChrome container running https://github.com/SeleniumHQ/docker-selenium. The standalone chrome container abruptly terminates in my Oracle Linux 7 machine. It is unclear what sends the SIGTERM signal to the chrome container. The machine has sufficient disk space and is able to run other Docker containers (e.g the selenium/hub:3.141.59-zirconium) without any problem.

Reproduction steps:

docker run -p 4444 selenium/standalone-chrome:3.141.59-zirconium

** I don't want to use the default 4444 port in the host machine. 

The container starts running for some time, and then terminates due to a SIGTERM. 

INFO Included extra file "/etc/supervisor/conf.d/selenium.conf" during parsing
INFO supervisord started with pid 9
INFO spawned: 'xvfb' with pid 12
INFO spawned: 'selenium-standalone' with pid 13
INFO [GridLauncherV3.parse] - Selenium server version: 3.141.59, revision: e82be7d358
INFO success: xvfb entered RUNNING state, process has stayed up for > than 0 seconds (startsecs)
INFO success: selenium-standalone entered RUNNING state, process has stayed up for > than 0 seconds (startsecs)
INFO [GridLauncherV3.lambda$buildLaunchers$3] - Launching a standalone Selenium Server on port 4444
INFO::main: Logging initialized @495ms to org.seleniumhq.jetty9.util.log.StdErrLog
INFO [WebDriverServlet.] - Initialising WebDriverServlet
INFO [SeleniumServer.boot] - Selenium Server is up and running on port 4444
Trapped SIGTERM/SIGINT/x so shutting down supervisord...
WARN received SIGTERM indicating exit request
INFO waiting for xvfb, selenium-standalone to die
INFO stopped: selenium-standalone (terminated by SIGTERM)
INFO stopped: xvfb (terminated by SIGTERM)

I am not sure who sends the SIGTERM. I enabled audit logging following this post https://www.thegeekdiary.com/how-to-find-which-process-is-killing-mysqld-with-sigkill-or-sigterm-on-linux/ and found that the supervisor process sends the SIGTERM to "bash" that started the container. However, it is unclear who sent SIGTERM to supervisor. 

type=PROCTITLE msg=audit(1582575602.648:1389598): proctitle=62617368002F6F70742F62696E2F656E7472795F706F696E742E7368
type=SYSCALL msg=audit(1582575602.649:1389599): arch=c000003e syscall=62 success=yes exit=0 a0=d a1=f a2=0 a3=1 items=0 ppid=21866 pid=21919 auid=4294967295 uid=1000 gid=1000 euid=1000 suid=1000 fsuid=1000 egid=1000 sgid=1000 fsgid=1000 tty=(none) ses=4294967295 comm="supervisord" exe="/usr/bin/python2.7" key="audit_kill"
type=OBJ_PID msg=audit(1582575602.649:1389599): opid=21927 oauid=-1 ouid=1000 oses=-1 ocomm="bash"

Other Docker containers on this machine run fine, including the Docker Selenium Hub https://github.com/SeleniumHQ/docker-selenium. 

I have tried disabling/adding https proxies and no proxy without any success. Please suggest ways to further investigate and resolve the problem.

Thank you.

[Aditya Joshi]

Hi all,

By using strace and regenerating audit logs, I was able to find an odd cron job that was silently terminating the chrome container. So problem solved and the lesson is to create cron jobs carefully.

Thanks