Shutting down because other poller took lock: None

[tkellogg]

Shutting down because other poller took lock: None

[tkellogg]

I'll see if I can reproduce this in a test.

I'm not sure how it would go into deadlock like this. The algorithm is:

1. start up, write own PID to config.json
2. start loop, check if PID in config.json is same as own PID. If not, exit loop and fail process.
3. do work, go to 2.

If a process dies unexpectedly, the new process will overwrite & take control. dura kill overwrites the PID with None/null so it effectively stops all processes.

The behavior you're describing seems to be that dura kill is running on repeat.

Are you running on Docker or an overlay filesystem? Ive seen strange behavior like this on Docker (overwrites not being recognized/consistent).

[jazzdan]

@tkellogg ah sorry, I should have included the environment. I'm running dura on macOS 12.0.1 arm64.

I tried git pull'ing my dura repo and re-installing. Now when I run dura serve or dura kill both hang with no output. Interesting!

[tkellogg]

can you check the logs in ~/.config/dura/logs/

also i wouldn't mind seeing a ls -lathr ~/.config/dura/ and also the config file.

[tkellogg]

I just tracked down the issue. I'll push out a fix as soon as I can. The problem is that the json file gets corrupted when overwriting with a shorter file. i.e. "overwrite" isn't what I normally think of as overwrite since it doesn't reduce the length of the file.

In your case, i think you might have restarted your laptop, which caused the PID count to reset to a shorter number, shortening the length of the file.

[tkellogg]

Thanks for submitting this one. I don't know how I got so far without seeing it.

[jazzdan]

Ahh good catch @tkellogg !! Thanks for the fix and for making dura :)

[jazzdan]

@tkellogg you're theory was right by the way, here's what my config.json looks like:

{"pid":57878,"repos":{}}sers/dan/devel/backend":{}}}}%