Customize automatic re-run logic

[fkorotkov]

Description

It will be nice to have a customizable re-run policy supporting fancy expressions like following:

task:
  rerun: "status('build') == 'failed'"

or even:

task:
  rerun: "'TimeoutException' in logs('test')"

[saberduck]

hello @fkorotkov ,

any plan to make this available? we have some jobs intermittently timeouting, this would be very useful.

[fkorotkov]

@saberduck do you build Cirrus' timeouts (the default 1 hour one, for example)? Or your build system timeouts to do something?

I'm not 100% sure about the final design here though. I think it seems that the most consis solution will be something like:

rerun_when_log_contains: 
  - Timeout
  - timeout

My initial thoughts about expressions seems unnecessary here and probably can be a part of #496

[pavel-mikula-sonarsource]

Hi @fkorotkov,

We have random timeouts on some intermediate steps like cloning repo, downloading nuget packages, downloading code analyzers, uploading code analysis results.

Some of those timeouts doesn't contain word timeout but stuff like bad gateway, internal server error (actually meaning a timeout), IllegalStateException: Fail to request https://... so it would be great to be able to configure it somehow.

[fkorotkov]

Fixed via #858

See documentation on the website https://cirrus-ci.org/guide/programming-tasks/#hooks