Why is this Fisheye thread consuming CPU cycles?

[Chad Barnes]

Fisheye 3.7.1
16G Heap
4x CPU
1 / 2 indexing threads configuration 
Mostly SVN repos, some Git 

Since our recent upgrade to 3.7.1, Fisheye has gotten into a state where a few threads will run constantly, saturating the CPU. After a restart, it calms down and then after some time, hours or days, will abruptly consume all of the CPU cores again (and not let go).

Fisheye is usable right now (after a restart) but there has been one constant thread consuming one CPU since the restart. What is its purpose? When the CPU cores become saturated, and Fisheye becomes unusable, there are several of these similarly-named threads competing for CPU:

Thread:
qtp1351478315-1318 [1318] (RUNNABLE)

Stack Trace
com.cenqua.fisheye.diff.MyersOND.buildPath line: 123
com.cenqua.fisheye.diff.MyersOND.diff line: 82
com.cenqua.fisheye.diff.Diff.diff line: 8
com.cenqua.fisheye.diff.DiffHelper.getHunkList line: 27
com.cenqua.fisheye.diff.DiffHelper.getHunkList line: 142
com.cenqua.crucible.revision.source.Source.getHunkList line: 230
com.cenqua.crucible.view.FRXDO.getHunkList line: 1287
com.cenqua.crucible.view.FRXDO.getTetrisGrid line: 781
com.cenqua.crucible.view.FRXDO.mapInlineComments line: 809
com.cenqua.crucible.view.FRXDO.<init> line: 204
com.cenqua.crucible.revision.managers.DefaultContentManager.makeFRXDO line: 114
com.atlassian.crucible.actions.ReviewBaseAction.makeFRXDO line: 597
com.atlassian.crucible.actions.ViewFRXAction.execute line: 188

...

GC is not an issue, not even when Fisheye becomes non-responsive. We have a 16G heap that collects down to 6G and there is no thrashing (of GC).

What is the purpose of this thread? Can we correct something to remedy the CPU utilization or is it necessary to assign more processors to this machine?

[Dominic toretto]

Fisheye server appears to be very busy with the CPU pegged at 100% utilization. There are no other applications running on the machine that appear to be consuming resources. There are no apparent exceptions reported in the Fisheye logs. for any CPU related Problem to check out here:https://www.techsupportdubai.com/led-tv-repair/">Tv 

Cause

The FISHEYE_INST/config.xml has a similar configuration:

CONFIG.XML

<resources>
  <incrementalIndexThreads max="30"/>
  <initialIndexThreads max="300"/>
</resources>
<crowd auto-add="true" resync="true" refreshExistingUsers="true" sso-enabled="

true" resyncPeriod="1 hour" jiraInstance="false">

[gustavo_refosco]

Hi Chad,

As Sergey mentioned from your description and stack trace it looks you're facing the issue described by https://jira.atlassian.com/browse/CRUC-6593, though you mentioned it happens even not loading any of the files.

For the review in which you observe this happening can you please confirm:

• What's the size of the files added to the review.
• How many revisions of these files are added to the review.

I've also asked for updates in CRUC-6593 in a private comment - please add yourself as a watcher to it so you can keep track of any progress on it.

Regards,

Gustavo Refosco

[Sergey Svishchev]

Likely this bug -- https://jira.atlassian.com/browse/CRUC-6593, support could confirm that for you.

[Chad Barnes]

Update: The thread starts when the review begins to load. Before any file is selected to view. This review has only 3 files. The thread never stops. A different review, with 700+ files, does not cause this issue.

[Chad Barnes]

More information: these threads start taking a full CPU core when a particular file is viewed in a review. The file is never viewable and the UI eventually times out. However, the thread never goes away. If we try to view that file in a review again (new window), a new thread pops up taking a full CPU core. Now we have 2X threads each taking up full CPU cores. We can keep doing this until the instance becomes non-responsive. The offending file is an xxx.sql file (with SQL content). We don't immediately see anything suspicious about it.