LAVA jobs and LAVA test cases may fail for a number of reasons, including:
See Writing Tests
If LAVA detected an error during the job, that error will be highlighted at the top of the job output. Some errors can be due to bugs in LAVA and should be reported to LAVA using the link at the bottom of each LAVA page. Other errors may be LAVA detecting an error in the job data (JSON, YAML, parser or scripts) which need to be fixed by the test writer.
See Best practices for writing a LAVA test job
When LAVA detects an error, the job will be marked as Incomplete and will
show up in the failure reports for the device type and the device.
Failure tags allow the same failure reason to be marked on a variety of different jobs which may otherwise be unrelated. If a particular failure starts to become common with a particular piece of hardware or due to a specific cause, a tag can be created by the lab administrators.
Failure tags can be used whether the job was marked as Incomplete in LAVA
or not. The tags will show on the job output but only Incomplete jobs will
show failure tags in the reports.
Failure comments can be used when a failure is unique, rare or to add more detail to an existing tag as it relates to this specific job.
Failure comments can be used whether the job was marked as Incomplete in
LAVA or not. The comments will show on the job output but only Incomplete
jobs will show failure comments in the reports.
This list will be expanded over time.
Errors in test job submissions and errors in admin setup can show up as a failure comment of no deployment or boot method being available. First, attempt to reproduce the failure with one of the standard test jobs. If the error is reproduced, the admin can use the Triage guidelines to identify the problem.
These messages relate to code errors which should either have been caught as invalid input or handled gracefully within the test job. Failure messages containing a traceback should be reported as bugs.
See also
Some updates of the LAVA packages can introduce new code support or changes to
the code support. The master and the worker need to be running the same code
version of lava-dispatcher for the test job to run correctly. For example,
a deprecated option in a test job submission could be removed in an update. If
the master is running an older version of code than the worker, the test job
could fail as the worker would be unable to handle the removed option.
To prevent this, the master and the worker calculate a compatibility number for
each test job. If this number does not match, the test job will fail. So a
compatibility failure indicates that the lava-dispatcher code on the worker
is out of date with respect to the master. This is a problem which needs to be
reported to the admins.
See also
Compatibility failures for more information on how developers set the compatibility for test jobs.
If you have permission to add or edit failure tags and comments, a button will
be displayed on the job output page Comment on failure.
The button displays a form where the current tag(s) or comment(s) are displayed. Additional tags can be selected and the comment (if any) can be edited.
Incomplete jobs will show up in the reports. Reports are generated which cover all jobs, all jobs on a specified device type and all jobs on a specific device.
Reports show two graphs of the number of complete jobs against the number of incomplete jobs over time, separating health checks from other test jobs. Clicking on the time / day link shows the failure tags and failure comments for the incomplete jobs during that timeframe.
Not all test failures will show as incomplete jobs and a Complete job can
still have failure tags and failure comments assigned.
If a test failed due to a problem outside the test definition or supporting files and scripts, use the link at the bottom of each page to report a bug.
