I need some help with Fisheye, as usual.
Our indexes are around 200G, and still only 15% of total repositories indexed. We're averaging a ratio of the fisheye index being twice the size of the svn repo. Performance sucks at best. For once I'm not ranting about Fisheye, I think this is totally our fault, actually the users' fault.
The size of the indexes is heavily dependent on "correct", or at least conventional usage of svn, ie your repository is has /tags/trunk/branches, or a number of components with tags/trunk/branches under each first level directory. The cost for a tag or branch correctly made, ie by copying trunk directly under branches or tags, is pretty cheap, but the cost of copying it to say /releases is very expensive, as it seems that each file gets completely reindexed.
As it happened we had other problems with svn in this area, and wrote a hook suite to ensure that only the conventional usage was allowed, but we still have the problem of all the changesets that had happened before this point.
I have had a dig around the lucene indexes using Luke, but have not hit on the query that will show me the number of roots. Based on that I would add exclusions or use the "symbolic" rules to get the indexes efficient. Eventually this whole thing will need to be automated. It's not as simple as just looking at the current layout, as I'd need to see the layout for every single revision.
What I'm looking for really is a tool that will open the index and show me the number of Documents and their disk usage under each logical svn root, so that I can exclude anything but the conventional roots.
Is this something Atlassian support might have?
PS. I promise I hate svn more than you so no need to recommend git or hg at this point ;-)
PPS. Also note that global includes don't support antglobs.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.