JVM Sizing question

[Jim Winsor]

We are planning a large (Enterprise scale) Jira instance, based on the sizing guidence from Atlassian. The server is equipped with more than 128 GB of memory.

During a recent planning session, a developer noted that JVM's larger than 8 GB can have performance issues due to garbage collection, and other memory management, taking too long.

Has anyone configured a JVM with large memory (64 GB and up) and experienced this type of issue, and was there any remedy?

Thanks!

[Harry Chan · Accepted solution]

Hi Jim, do you anticipate a need for JIRA having >64GB of memory set for the heap? It's quite a rare situation even in large organizations. We've had instances running with >8GB of RAM.

Whilst garbage collection would be consideribly slower, you'd also require less garbage collection (as you'll less often fill up all the memory). I'd consider using Java 7 and G1C as a concurrent garbage collector to avoid any huge performance spikes due to GC.

If performance is an issue though, I'd consider other methods of performance tuning than simply increasing the RAM.

[Jim Winsor]

Thanks!

Jim

[Mizan]

Hi Jim ,

I had a similar query few months back and I had raised this to Atlassian support they advised to use 6gb instead of 32gb which was hitting the performance .

[Daniel Alonso]

Hi @Jim Winsor2 @Mizan any final decision after working with this configuration for almost 5 years? Thanks in advance.

[darylchuah]

I found a few useful documentation on JIRA performance tuning and scaling:

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

Hope it helps

Cheers :)

[Jim Winsor]

Thanks, Daryl.

While these links provide helpful information, I was hoping to find someone who has actually implemented Jira using a large JVM (>64 GB) to understand the issues involved.

Jim