Forums

Articles
Create
cancel
Showing results for 
Search instead for 
Did you mean: 

Confluence 4.x Extremely Slooowwwww

Chris Schmidt
Contributor
February 15, 2012

We are evaluating Confluence as an option for our product support site - I have used confluence for quite a while for several development projects and think it is the perfect match for what we want. So we installed confluence 4.1 into a VM in our Data Center - instantly the CPU spiked to 200% on a Dual-Proc VM with 8g of RAM O_o

Response times render it completely unusable, up to 30 seconds to load a page sometimes! I read somewhere that this may be an affect running confluence within a VM - however, I have been running a 3.5 instance in a VM with less resources (1 Proc, 4g RAM) for a year with no slowness whatsoever.

Anyone had similar expereinces or have recommendations. For the interim I am installing 3.5 in hopes that it will run at least as fast as my other instance - but I really would like to run on the latest.

4 answers

1 accepted

6 votes
Answer accepted
Chris Schmidt
Contributor
February 15, 2012

FYI - There is an obscure setting in ESXi that limits RAM to 512m (under resource allocation). The VMWare ESXi essentially tells the OS that it has however much memory you have allocated, but it appears to only be exposing a 512mb page which it then swaps internally - causing some serious performance issues. Disabling this completely resolved my issue.

1 vote
Adam Saint-Prix
Atlassian Team
Atlassian Team members are employees working across the company in a wide variety of roles.
February 16, 2012

That sounds like it should be a Confluence/JIRA knowledgebase article.


"If you are running VMWare ESXi and having performance issues, check to make sure..."

Chris Schmidt
Contributor
February 16, 2012

+1 - That would have saved several hairs on my head and made my hair a little less grey at the end of this week. Hence why I came back to post the solution

Adam Saint-Prix
Atlassian Team
Atlassian Team members are employees working across the company in a wide variety of roles.
February 16, 2012

Totally get that and was glad to see you post the solution. Michael Seager or any other Atlassian folks: any chance you can make this into a KB per Chris' posted solution?

0 votes
Michael S
Atlassian Team
Atlassian Team members are employees working across the company in a wide variety of roles.
February 15, 2012

Chris,

Is there something in particular that's causing the spike? Usual culprits are scheduled XML backups, exporting to XML or trying to view certain Office files (XLSX and PPT can cause problems).

Try disabling the Office Connector and any scheduled backups as a first step. If that doesn't help, try taking external thread dumps (kill -3 <PID>) every 30 seconds for about for a period of a few minutes while it's slow and then create a support request via support.atlassian.com with those logs - we'll take a look into it.

Michael Seager
Atlassian Support - Sydney

Chris Schmidt
Contributor
February 15, 2012

I had backups disabled, I didn't disable the Office Connector tho. I had the same issue with 2 different distro's of 4 (4.0/4.1) and a botched install with 3.5 - I am currently trying the WAR version of Confluence 4.1.5 rather than the standalone version (since I am also running a tomcat instance on the box for jForum - it is a more ideal deployment anyway) - If the problems continue with the WAR deployment in an otherwise stress tested Tomcat 7 I will follow through with a bunch of thread dumps and raise an issue. Thanks Michael

0 votes
Jobin Kuruvilla [Adaptavist]
Rising Star
Rising Star
Rising Stars are recognized for providing high-quality answers to other users. Rising Stars receive a certificate of achievement and are on the path to becoming Community Leaders.
February 15, 2012

How much is the JAVA heap memory allocated to Confluence? You can find it under Administration > System Info. Try increasing it if it is the default 512MB.

See http://confluence.atlassian.com/display/DOC/Memory+usage+and+requirements

Chris Schmidt
Contributor
February 15, 2012

Nope that was the first thing I tried - I have seen heap swapping cause similar problems. I tried increasing heap to 2g, increasing permgen to 512m, enabled parallel gc, enabled hotspot compiler (-server) - nothing seemed to really make it any better.

Suggest an answer

Log in or Sign up to answer
TAGS
AUG Leaders

Atlassian Community Events