TestAddresserWorkerStopsWhenAddressDeallocationNotSupported fails on pp64el

Bug #1488576 reported by Curtis Hovey
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
juju-core
Fix Released
High
Unassigned

Bug Description

Since the fix for bug 1487727, we see that ppc64el now fails often. We got a bless from the last net-cli run because of retesting. Ass seen in:
    http://reports.vapour.ws/releases/3002/job/run-unit-tests-trusty-ppc64el/attempt/3729

    c.Check(err, jc.ErrorIsNil)
... value *errors.Err = &errors.Err{message:"checking address deallocation", cause:(*errors.errorString)(0xc210678700), previous:(*errors.Err)(0xc242fedeb0), file:"github.com/juju/juju/worker/addresser/worker.go", line:27} ("checking address deallocation: connection is shut down")
... error stack:
 connection is shut down
 github.com/juju/juju/api/addresser/addresser.go:39:
 github.com/juju/juju/worker/addresser/worker.go:27: checking address deallocation

machine_test.go:744:
    // When the address-allocation feature flag is disabled.
    c.Check(w, gc.FitsTypeOf, worker.FinishedWorker{})
... obtained = nil
... sample worker.FinishedWorker = worker.FinishedWorker{}

The outcome analizer sees this as a likely reason for future cursed revisions
    http://reports.vapour.ws/releases/issue/55d89569749a56415476f23b

Curtis Hovey (sinzui)
description: updated
Revision history for this message
Dimiter Naydenov (dimitern) wrote :

This seems like mainly a ppc64 issue, but also on other architectures occasionally, so with https://github.com/juju/juju/pull/3106 I've disabled the test until I have time to fix it properly (again).

Curtis Hovey (sinzui)
Changed in juju-core:
milestone: 1.25-alpha1 → 1.25-beta1
Revision history for this message
Dimiter Naydenov (dimitern) wrote :

Looking at the logs for cases where it fails: http://reports.vapour.ws/releases/issue/55d89569749a56415476f23b I think we're seeing a race where the API server announces the port it's listening on (e.g. via SetAPIHostPorts) *before* actually starting to listen on that port, which triggers a cascade failure - api can't connect to apiserver, which brings down the agent and runner, causing "connection is shut down", however some time after the apiserver starts listening on that port, but it's too late.

Changed in juju-core:
milestone: 1.25-beta1 → 1.25-beta2
Revision history for this message
Curtis Hovey (sinzui) wrote :

This issue has not been seen in 3 weeks

Changed in juju-core:
milestone: 1.25-beta2 → none
status: Triaged → Incomplete
Curtis Hovey (sinzui)
Changed in juju-core:
status: Incomplete → Fix Released
Revision history for this message
Anastasia (anastasia-macmood) wrote :

The bug was not seen as there were some tests disabled.

Re-enable the tests to confirm that the bug is fixed...

SKIP: machine_test.go:783: MachineSuite.TestAddresserWorkerDoesNotStopWhenAddressDeallocationSupported (Test temporarily disabled as flaky - see bug lp:1488576)

SKIP: machine_test.go:788: MachineSuite.TestAddresserWorkerStopsWhenAddressDeallocationNotSupported (Test temporarily disabled as flaky - see bug lp:1488576)

Changed in juju-core:
status: Fix Released → Triaged
Curtis Hovey (sinzui)
Changed in juju-core:
status: Triaged → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.