[logging] Don't try to kill autoscaler during log monitor cleanup

[FarzanT]

Why are these changes needed?

If the worker_pid is the string autoscale, the os.kill function will through a type error as it only accepts type int.

Related issue number

Closes #12565

Checks

• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[wuisawesome]

Thanks for the contribution! Will merge once the tests (presumably) pass.

[rkooo567]

Actually, it might be better to add this as well #12565 (comment) to avoid the future regression. Can we add some try except that checks if the pid is a string and add some logger.error if that's the case?

[wuisawesome]

Wait I think this is fine. I retract that comment lol. It was uninformed (I didn't knows the context that the original issue was fixed and then resurfaced). I was scared of hunting down deserialization bugs, but this way is more precise.

(@rkooo567 I'm assuming you're ok with this since you approved it, and we can make it more defensive later if we need to).

[rkooo567]

I think we need at least a better assertion here. Otherwise we will have regression every time we add a new string pid. I can make a PR real quick for that.