Page MenuHomePhabricator

cirrussearch role doesn't clean up after itself
Closed, DeclinedPublic

Description

Author: jgonera

Description:
If you enable cirrussearch role, you're stuck with elastic search server hogging your resources forever, even after cirrussearch role is disabled. If you run top, you're likely to see something like this in the first line:
1129 elastic+ 20 0 1623836 195716 0 S 3.6 19.2 0:16.94 java

Over time this process seems to be slowing down the virtual machine so much that pages take more than 30 seconds to load (or fail with HTTP 503 Service unavailable).

As a workaround you can run this in vagrant ssh:

sudo service elasticsearch stop && sudo rm -f /etc/init/elasticsearch.conf


Version: unspecified
Severity: normal

Details

Reference
bz69158

Event Timeline

bzimport raised the priority of this task from to Needs Triage.Nov 22 2014, 3:37 AM
bzimport set Reference to bz69158.
bzimport added a subscriber: Unknown Object (MLST).

[00:12] < ori> bd808: giuseppe's puppet compiler can output a diff of a host's catalog pre/post application of some patch
[00:12] < ori> if that's not expensive to generate, maybe the removal of role classes could be acted on somehow
[00:12] MatmaRex_mobile (~MatmaRex@wikipedia/matma-rex) left IRC. (Ping timeout: 240 seconds)
[00:13] < ori> though i dunno, maybe we should just add an exec that kills elastic if the cirrussearch isn't loaded and call it a day

Bulk unassigning bugs from Ori.

bd808 triaged this task as Low priority.Dec 30 2014, 2:01 AM

To cleanup after disabling role(s), rebuild the virtual machine with vagrant destroy -f && vagrant up.