004_basic_sanity.test_gen test failed for ovirt-master in vdsm project

Description

The patch that CQ marked seems to be related.
the patch that failed CQ is: https://gerrit.ovirt.org/#/c/93046/
you can look at the build's logs here:
https://jenkins.ovirt.org/job/ovirt-master_change-queue-tester/8725/artifact/exported-artifacts/basic-suit-master-el7/test_logs/basic-suite-master/post-004_basic_sanity.py/

the engine log shows several tasks that fail and then there is an exception:
https://pastebin.com/6f5QTZUH

Activity

Show:

Milan Zamazal July 20, 2018 at 1:35 PM

There has already been experience in OST with a similar problem, resolved by using a different check. So I posted a patch changing the check in preview_snapshot_with_memory: https://gerrit.ovirt.org/93216

Dafna Ron July 20, 2018 at 11:28 AM

thanks for looking in to this.
looking again at the last failure it's not in ovirt-engine but in ovirt-ansible-disaster-recovery and in 4.2
usually I would say that because its not in a specific project its probably the test but because its a completely random failure there is a chance that its been merged and not failed CQ.
So lets wait to see if its an actual regression and not skip at this time.

Milan Zamazal July 20, 2018 at 11:18 AM

I looked into the problem and I'm not sure whether the check in the test is based on wrong assumptions or whether there is a bug in Engine / REST API. I've sent a question to oVirt devel mailing list. Depending on the answer either the test in OST should be fixed or oVirt should be fixed. If the failure causes too much trouble in the meantime, you can disable make_snapshot_with_memory + preview_snapshot_with_memory + check_snapshot_with_memory tests in 004_basic_sanity.py.

Dafna Ron July 20, 2018 at 8:56 AM

we had a second failure for this test (this time with the correct name).
if we cannot resolve this soon I think we should skip the test.

Milan Zamazal July 19, 2018 at 3:37 PM
Edited

The bug is in 004_basic_sanity.preview_snapshot_with_memory test. The following check for successful snapshot creation is used:

snapshot = test_utils.get_snapshot(engine, VM0_NAME, SNAPSHOT_DESC_MEM) return (snapshot is not None and snapshot.snapshot_status == types.SnapshotStatus.OK)

However the check passes even when the snapshot is not finished yet, resulting in the reported bug. There is add_vm_pool test running between snapshot start and that check. If snapshot creation is faster than adding a VM pool, which seems to be the regular case, the check doesn't matter and the failure doesn't occur. So the problem is not much urgent. I don't know yet what's wrong with the check, but I'll try to fix it.

Please note there is an unrelated bug causing the test name reported being test_gen rather than the actual test name. It was introduced in a recent OST change and a fix is posted: https://gerrit.ovirt.org/93167. It can be merged as soon as CI passes (it's currently stuck, waiting on an available executor).

Francesco Romani July 19, 2018 at 10:19 AM

The linked patch (https://gerrit.ovirt.org/#/c/93046/) has nothing to do with vm snapshots, which seem to be the reason for the failure:

2018-07-18 15:55:22,271-04 ERROR [org.ovirt.engine.api.restapi.resource.validation.ValidationExceptionMapper] (default task-2) [] Input validation failed while processing 'POST' request for path '/vms/09566e39-399e-40e6-b22d-9985669c6138/previewsnapshot'.
2018-07-18 15:55:22,271-04 ERROR [org.ovirt.engine.api.restapi.resource.validation.ValidationExceptionMapper] (default task-2) [] Exception: org.ovirt.api.metamodel.server.ValidationException: Parameter 'snapshot.id' is mandatory but was not provided.

Therefore, I think the patch (https://gerrit.ovirt.org/#/c/93046/) is unrelated.

The error seems to be caused by incorrect request, but I don't have any idea where this request comes from. Maybe from OST?

Fixed

Details

Assignee

Reporter

Priority

Created July 19, 2018 at 9:29 AM
Updated September 2, 2018 at 3:50 PM
Resolved July 31, 2018 at 8:12 AM

Flag notifications