Sometimes, the NFSv4 client's state mangaer gets into a bad state.
This causes insane amounts of error messages which can quickly fill up
/var in common syslog configurations (i've seen 1000's of lines per
second)
I brought this up with upstream and Trond Myklebust offered the
attached patch to rate-limit the error messages:
http://thread.gmane.org/gmane.linux.nfs/47832
It was included in Trond's pull request last Thursday to Linus for
inclusion with upstream:
http://thread.gmane.org/gmane.linux.nfs/48150
Please consider this patch for inclusion in the debian kernel.