Question on network problems and JIRA data center

Question

Hello Support,

We have a standalone JIRA site hosted on a physical windows 2008 server with a 64GB RAM, 2 physical processors, 12 virtual processors, jira_home on a 200GB SAN disk (pool consists of SAS and SSD drives), anti-virus exclusions on jira_home and application installation folders, SQL database.

The application is hosted in OMAHA and we have users from Cape Town, Singapore, Australia, UK accessing the site. Though the application performs good we have users from other regions complaining about 1. Page loading for a long time, 2. A page does not shows all the details. Eg: A few fields present but a few are not. 
After thorough analysis we cornered the problem to be related to network latency and had a word with the network team and the good news is they changed a few things in the network riverbed and the latency problem seemed to have improved.

Right now the management want the riverbed changes to be rolled back as per the policy so the network problem will instantly arise after that.

My question is do you people suggest something good that can help the other regions to access the network without any latency? The users have tried connecting through their office network and also through VPN but the same latency issue persists.

After studying, clustering seems to be a better approach to tackle network problems but i think even the latest JIRA data center does not support "Geo-Clustering". Does it? I just read that disaster recovery on different geographical locations is supported but "Geo Clustering" is not.

May i request your help in providing some details on how get rid of network related issues for remote regions?

Many Thanks,
Kishore Srinivasan

Kishore Srinivasan · Answer

Hello Paulo,

First of all Many Thanks for your time! I'm sorry to have missed to mention the JIRA version. JIRA 6.2.4 is being used and is installed as a windows service. The java heap is set to 6144m and please see below the GC and the java startup options:

-Dcatalina.base=D:\jira
-Dcatalina.home=D:\jira
-Djava.endorsed.dirs=D:\jira\endorsed
-Djava.awt.headless=true -Datlassian.standalone=JIRA
-Dmail.mime.decodeparameters=true
-Dorg.apache.jasper.runtime.BodyContentImpl.LIMIT_BUFFER=true
-Djava.io.tmpdir=D:\jira	emp
-Djava.util.logging.manager=org.apache.juli.ClassLoaderLogManager
-Djava.util.logging.config.file=D:\jira\conf\logging.properties
-XX:+UseParallelOldGC
-XX:MaxPermSize=512m
-XX:+DisableExplicitGC
-XX:+UseCompressedOops
-XX:NewSize=3000m
-Dsun.rmi.dgc.client.gcInterval=900000
-Dsun.rmi.dgc.server.gcInterval=900000
-Dfile.encoding=UTF-8
-verbose:gc
-Xloggc:E:\jira_data\log\gc.log
-XX:+PrintGCDetails
-XX:+PrintTenuringDistribution
-XX:+PrintGCDateStamps
-XX:+PrintGCTimeStamps
-Dhttp.nonProxyHosts=10.*.*.*
-Djavamelody.storage-directory=E:\jira_data\javamelody
-XX:+HeapDumpOnOutOfMemoryError
-XX:HeapDumpPath=E:\jira_data\log

Garbage collection is behaving good with this setting and i'm also monitoring the GC activity everyday so the young generations happen very frequent (most often seconds apart and sometimes milliseconds apart) and takes an average of 0.08s to complete and the Full GCs happen almost every 1 hour and take an average of 7.5s toi complete. I'm planning to increase the heap to 7G if the Full GCs tend to go weird.

Many Thanks for the HRA suggestion! We have planned for a benchmark testing and i'll propose the HRA option to the users as the network team could analyze further with that.

We've anyway planned to update JIRA to v6.4 in 2 months so if you think this problem might be due to an application BUG, would 6.4 perform better and can remedy the network slowness?

Many Thanks,

Kishore Srinivasan

Paulo Hennig · Answer

Hello Kishore,

VPN should be the best option to use it, but as you are saying that this method has no effect (users have the same latency than when they’r not connected in the VPN), it’s hard to say which is the best option for your company as I have no knowledge about your environment. Your server hardware (CPU, disk and memory) seems to be very good, so there’s no worry about that. My suggestion is to make a test from users who are facing network problems instructing them to generate a HAR file and compare the latency with users who are not facing the problem, it’s important to say that this could be a BUG with the JIRA application, so if you could say what’s your JIRA version should help the troubleshooting.

About the geo-clustering, I’d suggest you to reach some of our experts and choose the best solution according to your environment and needs. In additional, you can also take a look it here.