Any way to increase Fisheye indexing on large Subversion operations?

[M. Stewart]

Frequently, we copy our entire trunk as a branch. This is a lot of data, and takes Fisheye a long time to index this. While this is being indexed, other users are committing smaller changes that would not take Fisheye too long, but Fisheye is still stuck on the large branch creation. This causes a problem for us because we use Crucible to review, and Crucible cannot see those smaller changes because Fisheye is still indexing the large branch creation. Is there any way for those smaller changes to be seen and reviewed even if Fisheye is still working on the branch?

[Lukasz Pater · Accepted solution]

In general FishEye needs to scan the revisions in order they're created, so it needs to finish the branch creation before it gets to the newer commits.

If your reviews are mostly on trunk, you might consider creating a separate repository that just indexes trunk (or the relevant branches), and use that for reviews.

If you're experiencing excessively long indexing times on branch creation, try raising a support request on https://support.atlassian.com.

[Rene Verschoor [Atlassian]]

By default, FishEye uses one thread to process the incremental changes.
You can increase the number of Incremental Index Threads to 2.
One thread will then be busy processing the large commit, but it's no longer blocking, as the other thread will visit your other repositories to see if there are any commits in them.

See also https://confluence.atlassian.com/display/FISHEYE/Configuring+indexing .

[M. Stewart]

I have threads set to the default which I believe is 3.

[Rene Verschoor [Atlassian]]

The default (i.e. if you haven't filled in any value) is 1.

[Lukasz Pater]

3.0.0 increased the default numer of initial and indexing threads to 3 (FECRU-3329).

[Rene Verschoor [Atlassian]]

I stand corrected...

[Sergey Svishchev]

Are you running Fisheye 3?

[M. Stewart]

Fisheye 3.0.1

[Sergey Svishchev]

FE3 speeds up svn indexing but it's still linear. But you should check if symbolic rules setup matches your actual repository layout.