The idea is the same as exposed in the issue.
In case of sanity failures, we execute also the performance stage but in dry-run mode suppressing any non-aborting exception. To achieve so, we also extended the dry-run mode inside the performance stage to also generate the perfvalues which are required by the logging handlers for emitting the performance record.
In case of sanity failures, the performance values will be logged as None.
Todos
[x] Add a check_fail_reason log format placeholder for allowing the fail reason to be also logged.
The idea is the same as exposed in the issue. In case of sanity failures, we execute also the performance stage but in dry-run mode suppressing any non-aborting exception. To achieve so, we also extended the dry-run mode inside the performance stage to also generate the
perfvalues
which are required by the logging handlers for emitting the performance record.In case of sanity failures, the performance values will be logged as
None
.Todos
check_fail_reason
log format placeholder for allowing the fail reason to be also logged.Closes #3186.