Showing results for 
Search instead for 
Did you mean: 
Sign up Log in

It's not the same without you

Join the community to find out what other Atlassian users are discussing, debating and creating.

Atlassian Community Hero Image Collage

Zombie reaping in pipelines

I have a phpunit test that checks a function which in turn uses `posix_getpgid` to find out if a process exists (given its pid). The tested process is launched (in the background) from the phpunit test, and simply sleeps for a while, then exits.

The test kept failing when run inside a pipeline, though it consistently works in other environments. In the pipeline the function reports that the process is running even after it has finished.

I believe I have tracked down the problem to the pid1 not reaping the process after it exits. Therefore the process does still exist (forever). This seems to be a well-known problem inside docker containers (to which I am fairly new).

I can demonstrate the basic problem with a single-file repo using the default image:


    - step:
          - ps aux
          - PID=$(ping -c 2 localhost > /dev/null 2>&1 & echo $!)
          - ps -p $PID && sleep 4 && ps -p $PID

This will show what the pid1 is, then launch a background process, show that it is running and that it becomes a zombie after it is done (I would expect the last one to not exist).

Results can be seen here.

For my personal use case I could obviously just detect that I am running inside the pipeline and skip the test, since I know my functions and tests do work in any other environment I need them to... but I feel that this is a problem that should be solved / solvable in the pipeline system.

As far as I understand I have no control over the actual `docker run` command launching the pipeline, so I've only tried to solve it inside the container - by using images that include some sort of init-system that takes care of zombie reaping (s6-overlay, fpco/pid1 and my_init)... but I'm slowly reaching the conclusion that it is impossible, because the pipeline launcher obviously overrides whatever the docker image specifies as `ENTRYPOINT` / `CMD`.

I think the solution could be as simple as adding an `--init` argument to the `docker run` command for a "docker-native" solution - but that is not a change I can make...

Does anyone have a solution for this that will work without bitbucket making any changes? Some ideas?

1 answer

0 votes

Hey @Mikk3lRo,

I am not entirely sure what the root cause is of this. If you think that you might be able to get it to work by specifying the `docker run` command, you could try this from within your build. As part of the build, just execute `docker run <your-image> <your-params>` and see if this works. 

Is it also by now possible to pass the --init option to docker for bitbucket pipelines?

Suggest an answer

Log in or Sign up to answer
Community showcase
Published in Bitbucket

Powering DevOps with Bitbucket Server & Data Center

Hi everyone, The Cloud team recently announced 12 new DevOps features that help developers ship better code, faster   ! While we’re all excited about the new improvements to Bitbucket ...

1,849 views 0 7
Read article

Community Events

Connect with like-minded Atlassian users at free events near you!

Find an event

Connect with like-minded Atlassian users at free events near you!

Unfortunately there are no Community Events near you at the moment.

Host an event

You're one step closer to meeting fellow Atlassian users at your local event. Learn more about Community Events

Events near you