Indexing since over half a year now and still nowhere near finished

Question

Our crucible/fisheye server has direct file:// access to our svn repo and was setup to index our user's feature branches and trunk.

We have over 400000 revisions and in the last 6 months the indexing reached to revision 126080.

Server CPU: Intel(R) Xeon(R) CPU E5-2640 v4 @ 2.40GHz
Server Memory: 32GB
Java cmd: /usr/lib/jvm/java-8-openjdk-amd64/bin/java -Xmx8192m -XX:MaxPermSize=256m -Dcrucible.review.content.size.limit=999999 -Dfecru.configure.from.env.variables=false -Dfisheye.library.path= -Dfisheye.inst=/atlassian/data/crucible -Djava.awt.headless=true -Djava.endorsed.dirs=/atlassian/apps/crucible/fecru-4.8.0/lib/endorsed -jar /atlassian/apps/crucible/fecru-4.8.0/fisheyeboot.jar run

Is this expected, that crucible needs multiple years to finally finish indexing?

We have very wrong diffs, when the indexing doesn't finish. I tried creating another repo which only indexes a very small subdirectory of personal branches of mine and since it finished indexing after a few hours the diffs were all correct and reliable.

So we really want the indexing to finish, but with all branches, not only my private ones. But indexing all will need about 8 years if we count correctly here.

Daniel Brunner_2ud9k8 · Answer

Yes, I use svnsync to create a complete local svn database copy. Crucible is accessing this via the file:// protocol.

The repo has 60GB in size. We tried indexing only the last 100 revisions and then create reviews, but the diffs are all wrong.

Mikael Sandberg · Answer

Do you have a local copy of the repository on the Fisheye/Crucible server? How big is the repository?

I used to have a Fisheye server and was trying to index a SVN repository that was about 110 Gb with 80,000 revisions, and I had to modify the indexing to only do the last 10,000 revisions, and it still took over a week for it to finish. Mind you that the server did not have a local copy of the repository.