+1 The improvements are huge. There are other fixes since this thread started that finally allowed my stress test to pass, I think they all should go into 0.10.
B. On Tue, Aug 25, 2009 at 3:13 PM, Damien Katz<[email protected]> wrote: > +1 > > On Aug 24, 2009, at 11:26 PM, Adam Kocoloski wrote: > >> Hi all, I committed a good bit of replication-related code today and >> wanted to check with the list regarding the 0.10 branch. Here's what I >> changed .. >> >> one flat-out bugfix, >> r807308, r807354: more precise and accurate calculation of replication >> progress >> >> one new feature that could be classified as a bugfix depending on your >> point-of-view, >> r807342, r807345: follow 302 redirects during replication >> >> and two significant performance improvements (thanks rnewson for all the >> stress testing): >> r807320, r807360: checkpoint at most once per 5 seconds >> r807208, r807459, r807461: minimize the number of full commit operations >> >> There's another new feature hiding in that last set of commits, namely the >> ability to do a restricted full commit >> >> POST /db/_ensure_full_commit?seq=N >> >> which guarantees that everything up to and including update_seq N will be >> synced to disk. This is a weaker guarantee than a vanilla >> _ensure_full_commit, and as a result it can be much faster in certain >> situations (e.g. replication from a server experiencing a heavy batch=ok >> write load). >> >> I'd like to merge all of these into the 0.10 branch if that abides by our >> release policies. Best, >> >> Adam > >
