> Date: Mon, 10 Oct 2011 08:59:08 -0400 > From: "Michael J. Flickinger" <[email protected]> > CC: Martin Pool <[email protected]>, [email protected] > > Hmm, all on the machine's side looks good. > > Low load, no stale bzr serve processes, nothing blowing up... > > Is this still a problem for you right now?
Just tried that again, and the problem still persists. The connection was reset after 3594 seconds, again very close to 1 hour mark. The end of the bzr log file looks identical to what I posted yesterday: 3593.800 RemoteTCPTransport.readv 2 offsets => 2 coalesced => 1 requests (2) 3593.800 hpss call w/readv: 'readv', '/emacs/.bzr/repository/packs/cbc955c848fe27435bbc01a59b4808b0.pack' 3593.800 21 bytes in readv request 3593.800 decoder state: buf[:10]='', state_accept=_state_accept_expecting_protocol_version So I think we do have some 1-hour timeout somewhere. If so, I wonder why the timeout isn't smarter: the connection is obviously active, with vast amounts of data being moved down the wire. So it makes little sense to drop it.
