Monitoring JIRA access

[Abhi Vaishnav]

Hi,

I was wondering what tools JIRA/Confluence admins use to monitor JIRA load.

We have JMX and JavaMelody to know when the load gets high; however, I want to know what is happening on the server when the load gets high.

Does anyone have tools they can recommend?

My primary objective here is to start noticing any patterns that causes high loads.

We had a user make large queries against the server every day at 3 and 10 PM; which caused JIRA to slow down at those times. We debugged that by looking at JavaMelody graphs and asking JIRA users if they ran automated scripts and if so when... There has gotta be a better way.

We also had confluence intermittently go slow; we'd take thread dumps, send them to Atlassian and neither us nor Atlassian have found anything. (Though this problem seems to have gone away after upgrading confluence).

We want to start getting ahead of these issues and are wondering what tools (if any) are out there to help.

[Abhi Vaishnav]

There is a tool call Traceview from AppNetta.

We are demo-ing it right now and it is pretty cool.

It captures everything from the JMX metrics to the incoming requests to the DB queries your app runs.

It also measures how much time each request spends in the various layers: HTTP request, Tomcat processing, DB query to HTTP response.

We think it is really cool and def is helping us ask Atlassian the right questions.

Maybe I'll write a blog on how we use this tool at some point...

[Mirek]

Hi Abhi,

Slowdowns are very complex to investigate. It could be a multiple reason... user load, database connection, virtualization, garbage collector, huge export/import, low on memory.. and more..

I do not think that you will be able to predict what will happen. Everything depends of your instance configuration and enviroment. We are also dealing with a lot of performance problems. Unfortunately there are no ideal tools that could exactly say what is going wrong.

This requires a lot.. I will say again.. a lot of preparation, testing, verifiaction. And as you say all problems could be gone after an upgrade. But this is not always a way. Sometimes after an upgrade your instance will be slower.

In general you will need to to a lot of testing and as a rule you cannot change more than one variable at one time. If you will debug memory.. do only memory.. if virtualization .. then completly scan virtualization.. etc.

If you will in the meantime upgrade the system or do a change to something else everything that you tested before will be useless.. You need to start over again.

Below documentation is always a good start:

https://confluence.atlassian.com/display/JIRAKB/Troubleshooting+Performance+Problems

https://confluence.atlassian.com/display/JIRAKB/JIRA+Performance+Tuning

but finding a root cause it is very hard. However I hope that you will find finally a anwers on your slownes problem.

Best Regards,

Mirek