Page MenuHomePhabricator

Send EXIT events more consistently from daemons
ClosedPublic

Authored by epriestley on Mar 24 2017, 3:56 PM.
Tags
None
Referenced Files
F15445772: D17553.id42215.diff
Thu, Mar 27, 3:58 PM
F15442149: D17553.id.diff
Wed, Mar 26, 9:45 PM
F15435754: D17553.id.diff
Tue, Mar 25, 10:02 AM
F15433076: D17553.diff
Mon, Mar 24, 9:13 PM
F15431064: D17553.id42220.diff
Mon, Mar 24, 10:18 AM
F15407459: D17553.id.diff
Tue, Mar 18, 5:36 PM
F15396492: D17553.id42220.diff
Sun, Mar 16, 2:43 PM
F15320693: D17553.diff
Thu, Mar 6, 10:15 PM
Subscribers
None

Details

Summary

Ref T12298. When we bin/phd restart daemons, we currently may miss sending EXIT events for daemons which aren't actively running at the time, since the EXIT event only happens when the subprocess exits, not if we tear down a waiting/sleeping/hibernating daemon.

The most common case of this is hibernating daemons. A less common case is daemons which have exited because of an error (usually, taskmasters).

In these cases, the web UI doesn't get cleaned up properly. Instead, send EXIT from any kind of teardown so we knock daemons out of the web UI more often.

Test Plan

Ran bin/phd restart a bunch, saw fewer hanging daemons in the web UI at /daemon/.

Diff Detail

Repository
rPHU libphutil
Branch
daemon3
Lint
Lint Passed
Unit
Tests Passed
Build Status
Buildable 16128
Build 21403: Run Core Tests
Build 21402: arc lint + arc unit

Event Timeline

This revision is now accepted and ready to land.Mar 24 2017, 4:06 PM
This revision was automatically updated to reflect the committed changes.