Row lock contention in Fisheye 3.2

Eli Kim
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.
December 8, 2013

I have upgraded to Fisheye 3.2 and am running on Oracle 11. I am finding that I am having massive row lock contentions with Fisheye. Specifically my issue is with updates to the cru_revision table. I have Fisheye connected to 74 repositories for Crucible reviews.

How could I reduce the "enq: TX - row lock contention"?

Also is it possible to throttle the indexing? I tried deleting FISHEYE_INST/var/cache and it triggered the re-indexing, but all the diff commands it runs and with there being so many revisions..... it is just beating against subversion.

4 answers

1 accepted

0 votes
Answer accepted
Eli Kim
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.
December 19, 2013

After watching a little, we found out that adding the indexes on foreign keys actually didn't help so we removed those indexes.

This is seen on Oracle 11 and MySQL 5.6.12

After discussions with support, we were able to show the bug to get them to replicate. This affects only LightSCM, Fisheye Native does not have this problem. If you only have a Crucible license, do not upgrade until the bug fix is released

0 votes
Eli Kim
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.
December 10, 2013

Thanks for the suggestion. It did help, but unfortunately it didn't eliminate the problem.

0 votes
Nick
Atlassian Team
Atlassian Team members are employees working across the company in a wide variety of roles.
December 9, 2013

Hi Eli,

FishEye 3.2 will perform an upgrade of your crucible (ie review) data when it first starts up. This should not take more than 5-10 minutes. Has the row-locking settled down since then ?

You can indeed throttle SVN commands, using the Throttle setting on each repository:

Eli Kim
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.
December 9, 2013

Hello,

I did see the migration of the users & groups out of the data0.bin; however we had 0 cause we are using Crowd. It did move something out though. Leaving it on over the weekend, it had major row-locking sessions and became stuck. Main culprit being with CRU_REVIEW. We placed an index on all the foreign keys as Sergey suggested. We are waiting to see if that resolved the issue.

We did see the Throttle Connections, but description also says that throttling is not advised in atlassian documentation. Is that just pre-cautionary warning?

Eli Kim
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.
December 10, 2013

Creating a index for all foreign keys did help but we are still getting row lock contention.

0 votes
Sergey Svishchev
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.
December 8, 2013

Look for missing index(es) on foreign keys. Confluence had the same problem with Oracle 11g -- https://jira.atlassian.com/browse/CONF-21909

Suggest an answer

Log in or Sign up to answer
TAGS
AUG Leaders

Atlassian Community Events