On Wed, 05 May 2004 13:56:41 -0400, Tom Lane <tgl@sss.pgh.pa.us> wrote: > Lonni Friedman <netllama@gmail.com> writes: > > On Wed, 05 May 2004 12:31:21 -0400, Tom Lane <tgl@sss.pgh.pa.us> wrote: > >> Once the complaint starts appearing, I'd expect it to continue until you > >> reindex the index. > > > That's exactly what happens. It consistantly errors until reindexed. > > Any suggestions? thanks. > > You are seemingly triggering some sort of bug in the backend's > large-object code, causing extra index entries to be made ... > but I sure haven't the foggiest idea how that would happen. > > Perhaps you could look at the parts of your application code that work > with large objects and see if you are doing anything "off the beaten > track" that might suggest a way to trigger the bug reproducibly. > I'm sure we could fix it if we could see it happening. hrmmm, i'm not sure what would constitute 'off the beaten track'. is there something specific that i could look for? or what is considered to be normal? I can tell you that the large objects that are getting chucked into the database are generally under 1MB in size (each), as they're mostly M$ documents. There are a few that are between 1MB & 10MB but I don't think there's anything larger than 10MB. ---------------------------(end of broadcast)--------------------------- TIP 8: explain analyze is your friend
