`tail` processes accumulate over time

[shivaraj-bh]

Reproducer

1. Start a new job
2. Open the job log page while it is running
3. Exit the job log page before it successfully finishes
4. Observe the processes under Vira using nix run nixpkgs#pstree -- [PID_OF_VIRA]
5. Notice tail -n +1 … /vira/[JOB_ID]/output.log

The tail processes created using the above steps are never killed — accumulating over time. The only time when the tail process is cleaned up after the CI is if the user stays on the log page until the CI finishes running.

(1456 in the above screenshot is the PID of Vira started using just run in Vira's devShell)

Expected behaviour

Always cleanup spawned tail processes

[srid]

@shivaraj-bh Is the full log available for me to inspect?

[shivaraj-bh]

@srid The initial title and description conflated two separate issues. Created #61 and updated the description and title of the current issue to separate concerns.

[srid]

Exit the job log page before it successfully finishes

Alternatively, you can refresh the page while the job is running. It will spawn yet another tail session, and more if refreshed again. Eventually, they terminate though.

[srid]

May or may not be relevant: bigskysoftware/htmx-extensions#147

[srid]

With #67, this is less of an issue, but nevertheless needs to be addressed.