On Fri, Feb 13, 2009 at 3:18 PM, Adam Kocoloski <[email protected]> wrote: > Hi Chris et al., trunk should definitely be better in this respect. We now > use a pipelined 10 connection pool instead of negotiating a new connection > for each request. We do need to fix a bug in trunk (just filed a patch) > that is causing connections to be closed far too frequently, though, so you > may see lots of connection_closed errors in the logs if you try it now.
It certainly is much better - I can't provoke the previous error with trunk at all. I can confirm the large number of connection_closed errors, although the replicator does seem to cope with it more or less. Is there a JIRA I could follow for the connection_closed issue? COUCHDB-197 maybe? Thanks! Chris.
