lucene indexing

[Eric Huss]

I noticed that while the full index is running, the incremental indexes
fail with this exception:

lucene.JavaError: org.apache.lucene.store.LockObtainFailedException: Lock
obtain timed out:
SimpleFSLock@/data/www/reviewboard/search-index/write.lock

It looks like the default lock timeout is 1 second. Has this been noticed
before? Any thoughts about preventing incremental indexing while the full
index is running?

Also, what's the motivation for doing the full index? Is it trying to
pick up changes that aren't caught by checking the last_updated field?

-Eric

[Christian Hammond]

Hi Eric,

I haven't noticed this, but it's certainly possible. We could look at preventing the incremental index.

What I'd like to do with search is look into switching to Haystack (http://www.haystacksearch.org), which should give the ability to use other search backends, which may be easier to set up. Whether this would help us prevent these problems, I don't know.

I don't know the search code all that well, so I'll let David answer the full vs incremental indexing question.

Christian

--
Christian Hammond - chi...@chipx86.com
Review Board - http://www.reviewboard.org
VMware, Inc. - http://www.vmware.com

To unsubscribe from this group, send email to reviewboard...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/reviewboard?hl=en
-~----------~----~----~----~------~----~------~--~---

[David Trowbridge]

The daily full index is just something I did initially. I don't think
it's actually necessary.

-David