Comment 30 for bug 932687

Any news on this? We're experiencing exactly the same problems as described by Peter, except that the workaround doesn't work for us.
We have a lot of Ubuntu 10.04 LTS clients running with /home mounted through NFSv4, with a Debian 6.0 server. We also had a single test machine running 12.04 for several months now without problems. Last friday, I upgraded a second machine and the described problems began.
We also had a server crash on friday, where I'm not sure whether it is related. The server stopped with "Out of memory and no killable processes left." Apparently, it started killing processes to free up memory. The logs say it was due to imapd claiming more memory, but that could well be wrong. What we also see on the server is that two out of four rpciod kernel threads are stuck in the 'D' state, which apparently also causes a permanent load level of at least 2.0. It doesn't seem to have any real performance impact, though. These stuck threads are obviously resolved when you reboot the server, but return as soon as you fire up the 12.04 boxes.
We already had network cards configured by /etc/network/interfaces, so Peters workaround doesn't work for us. I have now removed the /home line from fstab and instead mount /home manually on these two boxes. The clientaddr field is now correct (was 0.0.0.0 before), and everything seems to work now.
That is still something that needs to be resolved quickly. I suspect there are some protocol incompatibilities here; we already went back on the server from kernel 3.2.0 (from Debian backports) to the official sqeeze kernel 2.6.32 because we had problems with ever increasing load on the server. Maybe going again to 3.2.0 on the server would help now, since both client and server would then be running the same kernel version again. But I cannot upgrade all boxes to 12.04 beforehand just to test. I will try and set up a test environment and post the results.