Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Thu, Apr 16, 2009 at 06:38:42PM -0700, Tom Brown wrote: > > > > Simon Graham explained an interesting effect when using scatter gather > > and DRBD and xen... > > That seems to be this thread: > http://www.gossamer-threads.com/lists/drbd/users/16962 > > unfortunately, in that case, Valentin apparently reported good results with > protocol C and this tweak: > > ethtool -K xenbr0 sg off > > I can't get that. Even when using > > for if in eth0 peth0 virbr0 xenbr0 ; do > ethtool -K $if rx off tx off sg off > done add "tso off". or try to just set everything to "off" that is reported by ethtool -k. > I DO seem to get a slightly more stable config... in > fact, it sometimes seems to be stable until I bring the secondary back > online, at which point I get a crash that looks _very_ similar to the > initial one. > > But, using protocol A is robust... (but slow?). No, I don't think so. > hhmm, yes. I just managed to crash the box during a resync > (proto C). I have the trace for that too if it is relevent. > (drbd.take4) > > hhmm, I'm not convinced this test config is at all stable under > protocol C. I wanted to do a speed test, but can't keep it alive long > enough :( > > Anyhow, I need to put this down for now, I'm seriously in need of some > exercise. I just added a module parameter "disable_sendpage", which will disable zero copy sendpage use for protocols B and C as well, patch going to you via private mail. please test. -- : Lars Ellenberg : LINBIT HA-Solutions GmbH : DRBD®/HA support and consulting http://www.linbit.com DRBD® and LINBIT® are registered trademarks of LINBIT, Austria. __ please don't Cc me, but send to list -- I'm subscribed