clock related, intermittent failure in lease.TestPackage

Bug #1788550 reported by Anastasia
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Canonical Juju
Fix Released
Medium
Unassigned

Bug Description

From http://ci.jujucharms.com/job/github-merge-juju/975/ and more detailed http://ci.jujucharms.com/job/github-merge-juju/975/testReport/junit/github/com_juju_juju_worker_lease/TestPackage/:

----------------------------------------------------------------------
FAIL: manager_expire_test.go:194: ExpireSuite.TestClaim_ExpiryInFuture_TimePasses

manager_expire_test.go:219:
    fix.RunTest(c, func(manager *lease.Manager, clock *testclock.Clock) {
        // Ask for a minute, actually get 63s. Expire on time.
        err := getClaimer(c, manager).Claim("redis", "redis/0", time.Minute)
        c.Assert(err, jc.ErrorIsNil)
        clock.Advance(justAfterSeconds(newLeaseSecs))
    })
util_test.go:86:
    c.Fatalf("Store test took way too long")
... Error: Store test took way too long

Revision history for this message
Joseph Phillips (manadart) wrote :

This and its relatives below are stubborn heisenbugs.

- https://bugs.launchpad.net/juju/+bug/1631369
- https://bugs.launchpad.net/juju/+bug/1748381

I attempted to get at the latter of those. As noted in the report, there is a threshold of logging detail, that if passed makes failure unreproducible. I ran the test in a loop for a day.

Revision history for this message
Joseph Phillips (manadart) wrote :

And you are right. Something isn't playing ball in the clock implementation.

Revision history for this message
John A Meinel (jameinel) wrote : Re: [Bug 1788550] Re: clock related, intermittent failure in lease.TestPackage

Is running under '--race' useful at all? I'm just curious if the problem is
actually that we have a small race condition in the test. Though probably
it isn't a variable race as much as general concurrency issues. ('--race'
tells you if you're looking at a variable from 2 goroutines without
synchronization, but it won't tell you that you assume X changes before you
look at it, when actually it changes at a later time.)

On Thu, Aug 23, 2018 at 4:00 PM, Joseph Phillips <<email address hidden>
> wrote:

> And you are right. Something isn't playing ball in the clock
> implementation.
>
> --
> You received this bug notification because you are subscribed to juju.
> Matching subscriptions: juju bugs
> https://bugs.launchpad.net/bugs/1788550
>
> Title:
> clock related, intermittent failure in lease.TestPackage
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/juju/+bug/1788550/+subscriptions
>

Revision history for this message
Anastasia (anastasia-macmood) wrote :

I cannot reproduce this at all - we must have fixed it as part of other work.

I am marking this as Fix Released. Please open a different report if you are able to reproduce, attaching newer logs and possibly repro steps.

Changed in juju:
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.