Disk performance for Data Center shared homes

Eddy Fras
Contributor
September 24, 2018

Hi,

My Jira & Confluence DC nodes are very slow with no user connected, pages can take several seconds to display.

I tried to analyse the disk performance on my linux VMs and was redirected to https://confluence.atlassian.com/kb/testing-disk-access-speed-for-a-java-application-818577561.html by Atlassian support.

The problem is that the shared home directory is a NFS shared disk on each cluster nodes so it can't have the same performance as a local disk.

The results I have are :

Benchmark: RandomAccessFile

TOTALS
────────────────────────────────────────────────────────────────────────────────
stat avg median tmin tmax
────────────────────────────────────────────────────────────────────────────────
open 1,761,813 763,637 470,733 102,386,127
────────────────────────────────────────────────────────────────────────────────
r/w 19,746 16,230 7,317 236,447
────────────────────────────────────────────────────────────────────────────────
close 878,427 426,411 263,622 55,607,967
────────────────────────────────────────────────────────────────────────────────
delete 957,823 457,462 277,536 56,907,399
────────────────────────────────────────────────────────────────────────────────
All times are in nanoseconds.

Are these results ok for Jira/Confluence nodes or do I have really poor NFS disks ?

Thanks,

Eddy

1 answer

Suggest an answer

Log in or Sign up to answer
0 votes
Andrew Morin
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.
May 9, 2019

If I recall we had poor results when analyzing our NFS disk speed. We focused on the local disk speed to make sure we were hitting the numbers in the table at the page link you have in your post. We actually used a VM with similar specs as our app nodes as our NFS for the first year we were on JDC and CDC. Each app had its own dedicated VM for this. We had >1 million issues and over 20k users running 4 app nodes for each app. Focus on tuning Java JVM and database connection settings. How big is your JVM in the <install>/bin/setenv.sh file? How many nodes are you running? CPU's per node? We have moved to Isilon storage but didn't see that big of a difference in performance. What did your local disk results look like?

TAGS
AUG Leaders

Atlassian Community Events