Re: IndexWriter 2-phase commit usage

2009-02-24 Thread Michael McCandless
owse/LUCENE-1449 Cheers Mark - Original Message From: "fang...@emc.com" To: java-user@lucene.apache.org Sent: Tuesday, 24 February, 2009 9:49:34 Subject: RE: IndexWriter 2-phase commit usage The prepareCommit should do most real works, so the chance index2.commit() failure

Re: IndexWriter 2-phase commit usage

2009-02-24 Thread mark harwood
. See here for some discussion/Junit examples: https://issues.apache.org/jira/browse/LUCENE-1449 Cheers Mark - Original Message From: "fang...@emc.com" To: java-user@lucene.apache.org Sent: Tuesday, 24 February, 2009 9:49:34 Subject: RE: IndexWriter 2-phase commit

RE: IndexWriter 2-phase commit usage

2009-02-24 Thread Fang_Li
The prepareCommit should do most real works, so the chance index2.commit() failure should be slim. I think it's very hard to compensate the changes already committed. One solution is that you create separate indexes for each transaction and merge them later. Merging can fail, but the transaction