FishEye 2.8 "performance improvement" - does this also have a positive effect on scanning repositories?

[Daisuke Niwa]

Hi there,

I am happy to learn that FishEye 2.8 has been released, and one of the main feature in this version is performance improvement.

I have been told by many customers that they have a problem with FishEye scanning. They say sometimes it takes even weeks to complete initial scanning.

The "Best practice for FishEye configuration" page below recommends various measures such as using file protocol for SVN, starting scanning from a "sensible" revision, skipping Perforce labels etc... but this may affect the usability of FishEye from the customer's point of view.

Has there been any performance improvement in terms of repository scanning? For example, FishEye 2.8 can complete scanning of large SVN repository x times faster than 2.7? If so, I would be very interested to know the details.

I appreciate your help.

Regards,

Daisuke Niwa

[Nick · Accepted solution]

Has there been any performance improvement in terms of repository scanning?

FishEye 2.8 has improved browse performance of all pages with an Activity Stream, and the projects listing page specifically.

Indexing (scanning) Performance is something we are well aware of as an issue, and quite possibly next on our hit-list ;). For the 2.8 release however we prioritised Browse Performance above indexing performance, since indexing is a background task and is not always blocking a user.

[Daisuke Niwa]

Hello Nick,

Thank you for your response. Understood current situation. Any chance scanning performance will be improved on next release?

Please let me know if I should create a feature improvement request on jira.atlassian.com.

Regards,

Daisuke Niwa

[Nick]

Hi Daisuke,

We do hope we can make some positive improvements to indexing performance for a near-future release. I am unable at this however to provide you with any exact releases or dates.

In the meantime, I would strongly suggest you carefully check through the tips you mentioned: https://confluence.atlassian.com/display/FISHEYE/Best+practices+for+FishEye+configuration .

Most importantly:

• carefully consider whether or not your teams need to search diff text, if not, disable that
• if you are using perforce or svn, ensure that you've correctly configured all branches/tags/trunk symbolic rules
• check you have allocated a decent amount of Java Heap space to your FishEye process
• carefully consider bumping the number of initial indexing threads to 3 or 4, so that longer running repos wont block smaller ones.

Hope those things help out somewhat for the time being!

Cheers,
Nick

[Daisuke Niwa]

Please refer to "Best practice for FishEye configuration" page below. I am hoping our customers no longer need to take as much care of FishEye configuration as described there.

https://confluence.atlassian.com/display/FISHEYE/Best+practices+for+FishEye+configuration