TestAgentConnectionDelaysShutdownWithPing fails

[Aaron Bentley]

As seen here:
http://reports.vapour.ws/releases/issue/55564982749a5650f87680a9

[Curtis Hovey (sinzui)]

This bug is the most common reason unit tests fail in lxc. We want to run in lxc because it is faster to bring up and parallelise.

[Curtis Hovey (sinzui)]

We can see this issue affects the large windowss machine as well as lxc
   http://reports.vapour.ws/releases/issue/55564982749a5650f87680a9

[Alexis Bruemmer (alexis-bruemmer)]

Moving this to high and targeting it to 1.25 given the impact on test efforts

[Dimiter Naydenov (dimitern)]

I hope this http://reviews.vapour.ws/r/2454/ might fix it (in net-cli, soon to be merged in master).

[Dimiter Naydenov (dimitern)]

My "fix" didn't work, but gave me enough information to propose a better one: https://github.com/juju/juju/pull/3094 which should make is pass more reliably as it should be much less sensitive to machine load / power.

[Aaron Bentley (abentley)]

I'm sorry to report that we are still seeing this in master, even with the second fix landed:
http://reports.vapour.ws/releases/3006/job/run-unit-tests-win2012-amd64/attempt/1126
(3006 was a test of bdc631, which has 3be8e2f as an ancestor)

[Dimiter Naydenov (dimitern)]

Well, I've discovered there's a race condition somewhere around suite initialization, which is causing the test to fail intermittently, and not due to heavy machine load (i.e. my fix for calculating the timeout according to time it takes on the actual machine running it *was* correct, but it uncovered the race).

[Cheryl Jennings (cherylj)]

Reopening as this issue is still being seen:

http://reports.vapour.ws/releases/4108/job/run-unit-tests-win2012-amd64/attempt/2566

[Cheryl Jennings (cherylj)]

Here's a link to the issue to see all the cases where this is being hit:

http://reports.vapour.ws/releases/issue/55564982749a5650f87680a9

[Nicholas Skaggs (nskaggs)]

Just a note that this also affects juju-2.0.