Indexing in a virtualized environment

[Adam White]

For those of you that have virtualized your instances (both application and database), what are your indexing speeds like / how many issues do you have?

Ours seems to vary wildly. We have ~80,000 issues and a background index takes a widely variable amount of time.

[MattS]

Background index times depend on the activity in JIRA at the same time. I'd check your VM server statistics for other VMs using resources (CPU/disk) during the same time. You also want to check you have enough -Xmx set for the JVM heap to avoid large GC pauses

[Adam White]

Our  XMX value is "-Xmx6144m", so that seems like plenty, no?

[Adam White]

The database server has 4 Cores and 18GB RAM. MS SQL Server 2012. 

[Boris Georgiev [Appfire]]

What type of database are you using and what version (very important as I noticed great difference between MySQL versions for example) ?

Indexing speed might be affected by a lot of factors.

[Peter T]

Hi Adam.

I would look at the storage performance.What are you using to provision storage to the VM?

On system I use quite often with 50K issues background indexing takes about 8 mins.

This KB Article explains how to start about performance troubleshooting.

https://confluence.atlassian.com/kb/testing-disk-access-speed-for-a-java-application-818577561.html

Cheers,
Peter T

[Adam White]

We've got raided SSDs providing storage to the DB, Machine, and Home Directory servers, something else is amiss here. Mind sharing your virtual machine specs?

[Peter T]

Can you provide some examples of the times you are experiencing? Also what kind of monitoring you have on the system, CPU, I/O, memory, users load?

[Adam White]

Thanks Peter. We have 24GB RAM with 8 Cores. We're using RAID'ed SSD drives for storage. Anti-virus should not be a factor on the machine. The 5 hour background index was on off-hours (sunday afternoon). 

We have 90,000 issues and a background index on sunday afternoon took over 5 hours.

We have  323 custom fields according to System Info, could that be a reason? I've tried to restrict some of them to specific projects, but I still have some room to go there. What is your custom field count at?

[Adam White]

I should also say that general JIRA usage is very fast, including searching. The only painful thing is re-indexing.