I'm using NeoIndexService. I've tried very hard to reproduce the bug outside of tomcat but have failed to do so. Rearranging my code a bit and handling the exceptions better seems to fix this problem, by I don't know if it's by accident or not.
If I write I create nodes, write properties and index them, and then in the same transaction try to read them, what happens? I think this may be what happened for me, but I'm not sure really. Philip On 5/11/08, Mattias Persson <[EMAIL PROTECTED]> wrote: > I guess having each request wrapped in one transaction sounds sane and > to answer your previous question the indices (changes to your > IndexService) aren't really written until you call tx.finish(). So > that's either a bug in the IndexService (btw which one do you use, > LuceneIndexService or NeoIndexService?) or your transaction handling > may be buggy, however your supplied example seems correct... we'll > have to test that more. > > 2008/5/10 Philip Jägenstedt <[EMAIL PROTECTED]>: > > > On 5/10/08, Johan Svensson <[EMAIL PROTECTED]> wrote: > > > >> If tx.finish() is never called the transaction will never be committed > >> or rolled back (holding locks/memory). > > > > I only have a single transaction wrapping everything I do for the > > entire request. Not nice maybe, I'm trying to work out some better way > > to split it. In any event, it is precisely on the form that you've > > written here, but after the request fails once (due to > > ServletException thrown when the MusicBrainz webservice is overloaded) > > it seems that the indexes created in that very transaction block have > > not been rolled back, because next time I run it seems like there's a > > hit in the index, followed by a NotFoundException. I can reproduce > > this seemingly reliably (well, it depends on the musicbrainz > > webservice getting overloaded, but anyway) if I shut down tomcat, > > remove my data store and run from scratch. > > > > In other words, the index seems to have been updated/not rolled back > > even though I never called success(). > > > > Bug? > > > > Philip > > > >> When manually managing transactions always make use of a try finally > block: > > > > Is there a non-manual way to handle transactions? Sounds comfortable... > > > > Philip > > > _______________________________________________ > > Neo mailing list > > User@lists.neo4j.org > > https://lists.neo4j.org/mailman/listinfo/user > > > > _______________________________________________ > Neo mailing list > User@lists.neo4j.org > https://lists.neo4j.org/mailman/listinfo/user > > _______________________________________________ Neo mailing list User@lists.neo4j.org https://lists.neo4j.org/mailman/listinfo/user
