Yggdrasild fails to start if a stale worker pid file is left behind

[adamruzicka]

To Reproduce

1. Start yggdrasild with at least one worker
2. kill -9 $yggdrasil_pid
3. kill -9 $worker_pid
4. Start yggdrasild again

Expected behavior
Yggdrasil checks that the worker pid file points to a process that does not exist anymore, removes the file and moves on.

Actual behavior
Yggdrasil fails to start

# /usr/sbin/yggdrasild
[yggd] 2022/10/10 10:23:31 /root/yggdrasil/cmd/yggd/main.go:207: starting yggd version 0.2.98
cannot stop workers: cannot stop worker: cannot stop worker: cannot stop process: os: process already finished

# echo $?
1

Additional Information

• OS: Rocky Linux 8.5 but that shouldn't be relevant
• Version main @ 88a5f78

[subpop]

I wonder too if those pid files are clobbering each other with different worker-dir values. If two separate instances of yggd are running with different worker-dir values, they don't currently differentiate run dir values, so they both attempt to write "directive.pid" file names. :\

[subpop]

Is this reproducible on the yggdrasil-0.2 branch? Or with shipping versions of rhc-0.2.x, or did this only happen on main?

[adamruzicka]

IIRC I only encountered this when messing around with main. Right now I only have yggdrasil-0.2.0 at hand and it seems to be safe from this