model-migration pre-checks failed to catch issue: not idle

[Christopher Lee]

As seen in issue: http://reports.vapour.ws/releases/issue/57ce2d2b749a563a65a52631

We can see in the machine-0.log that the precheck failed, but it didn't make the 'juju migrate' command fail.
Example log:
2016-09-05 04:18:32 INFO juju.worker.migrationmaster:ef70bc worker.go:254 performing source prechecks 2016-09-05 04:18:32 ERROR juju.worker.migrationmaster:ef70bc worker.go:262 source prechecks failed: unit mongodb/0 not idle (failed)
2016-09-05 04:18:32 INFO juju.worker.migrationmaster:ef70bc worker.go:228 setting migration phase to ABORT

[Menno Finlay-Smits (menno.smits)]

Already fixed by this: https://github.com/juju/juju/pull/6155

[Curtis Hovey (sinzui)]

Sorry. CI continues to see this issue. The frequency does not appear to show a regression, rather, the issue is intermittent.

[Nicholas Skaggs (nskaggs)]

Further investigation needs to occur as to why this seems ok in LXD (on some hosts), but otherwise is rather intermittent on other LXD hosts (silcoon), as well as in the cloud. All of these should work.

[Tim Penhey (thumper)]

We now allow idle and executing, as any sufficiently large model will almost certainly have something running all the time.

[Tim Penhey (thumper)]

https://github.com/juju/juju/pull/6631

[Anastasia (anastasia-macmood)]

CI is still seeing this issue... http://reports.vapour.ws/releases/issue/57ce2d2b749a563a65a52631

[Menno Finlay-Smits (menno.smits)]

Here's the fix: https://github.com/juju/juju/pull/6712

The reason is quite involved. Details are on the PR.

[Anastasia (anastasia-macmood)]

PR against 2.2 (develop): https://github.com/juju/juju/pull/6779