Bug 1146208 - Watchman throttler plugin defaults are very aggressive
Summary: Watchman throttler plugin defaults are very aggressive
Keywords:
Status: CLOSED DEFERRED
Alias: None
Product: OpenShift Container Platform
Classification: Red Hat
Component: Unknown
Version: 2.1.0
Hardware: x86_64
OS: Linux
medium
low
Target Milestone: ---
: ---
Assignee: Brenton Leanhardt
QA Contact:
URL:
Whiteboard:
: 1167893 (view as bug list)
Depends On:
Blocks: 1167895
TreeView+ depends on / blocked
 
Reported: 2014-09-24 18:45 UTC by Timothy Williams
Modified: 2018-12-09 18:39 UTC (History)
4 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2016-04-18 19:18:39 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Red Hat Knowledge Base (Solution) 1202813 0 None None None 2016-04-18 19:18:12 UTC

Description Timothy Williams 2014-09-24 18:45:36 UTC
Description of problem:
The watchman throttler plugin's defaults are very aggressive for a default value. The defaults are:

  apply_percent: 30
  restore_percent: 70

This means that when a gear is using 30% (or more) of its allocated cpu limit for 120 seconds, it will be throttled. This is very aggressive, as the gear has not reached near its normal cgroup limit before being throttled. If a healthy gear consistently uses ~32% cpu, it will be throttled without justification. 

The defaults make sense for OpenShift online, but may not make sense for what we expect a normal OpenShift Enterprise deployment to be. 

Version-Release number of selected component (if applicable):
2.1

Additional info:
Since the throttler is barely documented, it is also confusing to customers when healthy gears are throttled. https://bugzilla.redhat.com/show_bug.cgi?id=1146147

Comment 2 Randall Theobald 2014-09-24 19:13:10 UTC
For many OSE installers, one of the first things they will do is some sort of performance testing. When they see the throttling and have no idea what is going on, there will be some WTF moments...

I'd suggest you carefully consider what the defaults should be for OSE. One option should be that throttling is disabled completely by default, and admins can enable it if they are targeting a high-density environment with development activities (instead of pure production deployments).

Comment 3 Timothy Williams 2016-04-18 19:18:39 UTC
This is intended behavior and has been documented.

Comment 4 Timothy Williams 2016-04-18 19:19:18 UTC
*** Bug 1167893 has been marked as a duplicate of this bug. ***


Note You need to log in before you can comment on or make changes to this bug.