Persistent worker failed to start the task: Parallels isolation failed

[real-or-random]

https://cirrus-ci.com/task/4823373415972864
https://cirrus-ci.com/task/6102191808184320
https://cirrus-ci.com/task/4694816924631040

from bitcoin-core/secp256k1#925 and bitcoin-core/secp256k1#926

[fkorotkov]

It does seem like a false alarm. @edigaryev seems it's hitting this case https://github.com/golang/crypto/blob/83a5a9bb288b4c9523e182ed4f621411a11de3df/ssh/session.go#L456-L460.

I've fixed the issue on the server side: now Cirrus ignores failures from the persistent worker in case of a task being already in a final state. In @real-or-random cases all the tasks finished successfully but then CLI reported the failure.

@edigaryev once you are back from PTO next week let's think if CLI should just ignore ExitMissingError.

[real-or-random]

Thanks for the quick fixes again!

Sorry, here's another failure:
https://cirrus-ci.com/task/6378104290541568
Not sure if this is related or not. The error message is different but it shows the same behavior of marking a task failed even though it has run successfully.

Maybe this is already fixed here but the fixed is simply not deployed? If yes, ignore me.

[fkorotkov]

@real-or-random this is a bit different issue. Seems there was an issue while updating clone command status (this is why it's reported as failed). It caused a confusion between service that is responsible for status updates of tasks/builds and the service that runs persistent workers.

I've fixed the issue. Sorry for the inconvenience!

[real-or-random]

I've fixed the issue. Sorry for the inconvenience!

Thanks!