[LU-16473] track when failloc was set but not triggered on unset

Fri, 13 Jan 2023 19:30:54 +0000

There's a somewhat common test problem where a failloc is being set, but does not trigger not leading to desired action and leading to cryptic errors, or sometimes false successes.

We already have functionality to see if failloc was triggered - used to implement a ONCE flag in particular.

The idea is to export this to userspace somehow on unset and then as a first step run once final unsetting would print a warning (perhaps aggregated across multiple nodes?) in test log if nothing was triggered.

Then tests like https://testing.whamcloud.com/test_sets/89df8dad-7b3f-414a-9c2a-7328807139c2 would have a much more visible message about what happened leading to much faster way to diagnose it.

Potentially other improved uses of this functionality will come up later

Whamcloud Community JIRA

[LU-16473] track when failloc was set but not triggered on unset