s6-notifyonchec zombie process

[klieber]

I'm seeing the following zombie process when using this container:

PID  PPID  C STIME TTY      TIME        CMD
405 32589 30 Sep09 ?        3-00:41:55  /usr/local/crashplan/bin/CrashPlanService
412   405  0 Sep09 ?        00:00:00    [s6-notifyonchec] <defunct>

This appears to be a similar issue that was reported with your docker-handbrake container back in April: jlesage/docker-handbrake#59

Wondering if the same fix needs applied to this one?

Also, thank you for all your work on this container!

[jlesage]

Which version of the image are you using?

[klieber]

Which version of the image are you using?

2.7.3

[jlesage]

The fix has been applied on version 2.6.4... I will check if I have the same thing on my side.

[XDGFX]

Just wanted to say I'm observing the same on 2.9.0

~> ps axo stat,ppid,pid,comm | grep -w defunct
Z    22580 22591 s6-notifyonchec <defunct>

[GreatHolyCow]

This appears to have come back

**** 421 7.5 1.0 10232528 351960 ? Ssl 12:36 0:28 /usr/local/crashplan/bin/Code42Service root 432 0.0 0.0 0 0 ? Z 12:36 0:00 [s6-notifyonchec]

Z 421 432 s6-notifyonchec

Running Latest: 2.14.0 I just logged into ubuntu informing me of zombie processes tried restarting docker comes back, restarted entire system comes back. Verified latest version from docker image.

[agners]

Just stumbled upon this while researching a similar problem.

By default, s6-notifyoncheck forks itself before exec'ing into the program you trying to wait for readiness (I guess CrashPlanService in your case). Now s6-notifyoncheck, your process likely won't reap the child process, hence it will stay as a zombie process.

s6-notifyoncheck has an option to avoid this:

-d : doublefork. s6-notifyoncheck will run as the grandchild of prog... instead of its direct child. 
     This is useful if prog... never reaps zombies it does not know it has. 

[jobenvil]

@agners Thanks a lot, it worked! I had same issue with the container docker-firefox.

root@arkanoid:/usr/local/etc/firefox-docker/config# docker exec -it firefox-alpine /bin/bash
root@firefox-alpine:/# ps -ef
UID          PID    PPID  C STIME TTY          TIME CMD
root           1       0  0 17:44 ?        00:00:00 s6-svscan -t0 /var/run/s6/services
root          38       1  0 17:44 ?        00:00:00 s6-supervise s6-fdholderd
root         434       1  0 17:44 ?        00:00:00 s6-supervise xrdp
root         435       1  0 17:44 ?        00:00:00 s6-supervise guacd
root         436       1  0 17:44 ?        00:00:00 s6-supervise web
root         437       1  0 17:44 ?        00:00:00 s6-supervise autostart
root         439       1  0 17:44 ?        00:00:00 s6-supervise xrdp-sesman
root         441     434  0 17:44 ?        00:00:00 /usr/sbin/xrdp --nodaemon
abc          442     436  0 17:44 ?        00:00:00 node app.js
root         443     435  0 17:44 ?        00:00:00 /bin/bash ./run
root         444     437  0 17:44 ?        00:00:00 bash ./run
root         445     439  0 17:44 ?        00:00:00 /usr/sbin/xrdp-sesman --nodaemon
root         446     441  0 17:44 ?        00:00:00 [s6-notifyonchec] <defunct>
root         455     443  0 17:44 ?        00:00:00 /usr/sbin/guacd -f -b 0.0.0.0 -l 4822
root         464     444  0 17:44 ?        00:00:00 sleep infinity
root         483       0  0 17:45 pts/0    00:00:00 /bin/bash
root         491     483  0 17:45 pts/0    00:00:00 ps -ef

adding the -d on the line s6-notifyoncheck solved me problem too.

root@firefox-alpine:/# cat /etc/services.d/xrdp/run
#!/bin/execlineb -P

# Move stderr to out so it's piped to logger
fdmove -c 2 1

# Notify service manager when xrdp is up
s6-notifyoncheck -d -w 500 -c "true &>/dev/null </dev/tcp/127.0.0.1/3389"

# Wait until other services are running
if { s6-svwait -t 10000 -U /var/run/s6/services/xrdp-sesman/ }

# set env
s6-env DISPLAY=:1

/usr/sbin/xrdp --nodaemon