Large number of zombie processes in docker

[niaow]

When using this plugin in docker, there is a large accumulation of zombie processes. I am using an ssh repo. In one container, I have a large accumulation of bash and git-remote-http zombies. In another, only sh zombies appear.

Somehow, processes are not being waited on properly.

[niaow]

system monitor screenshots:

[niaow]

Sorry for not responding on this for a while.

Basically what happens is that go only waits on only the child process, and not the grandchildren processes. These grandchildren processes exit but are not removed from the process table due to the lack of a wait syscall.

Git will normally wait on child processes. However, it skips this and bails out early if there is an error - creating zombie orphan granchild processes.

[jessicah]

I'm seeing this with ssl_client as well...

jessica@homecloud:~$ ps -axl | grep defunct
0     0  1042  1941  20   0      0     0 -      Z    ?          0:00 [ssl_client] <defunct>
0     0  1050  1941  20   0      0     0 -      Z    ?          0:00 [ssl_client] <defunct>
0     0  1089  1941  20   0      0     0 -      Z    ?          0:00 [ssl_client] <defunct>
0     0  1096  1941  20   0      0     0 -      Z    ?          0:00 [ssl_client] <defunct>
0  1000  6207  6185  20   0   6256   904 pipe_w S+   pts/1      0:00 grep --color=auto defunct
0     0 27448  1941  20   0      0     0 -      Z    ?          0:00 [ssl_client] <defunct>
0     0 29873  1941  20   0      0     0 -      Z    ?          0:00 [ssl_client] <defunct>
0     0 29918  1941  20   0      0     0 -      Z    ?          0:00 [ssl_client] <defunct>

where the parent process is:
4 0 1941 1888 20 0 103416 1336 - Ssl ? 0:00 /bin/parent caddy --conf /etc/Caddyfile --log stdout --agree=$ACME_AGREE

A new zombie process seems to get created on every request that hits my CGI script... normal web requests are unaffected.