There have been several postings on this list regarding those helpful
"UBC write error" console messages that are a new feature of the 3.2
release. For what it's worth, here is the response I got from Digital
CSC on this issue:
[snip]
"This is due to a change in UBC which prints out a message every time
a write error is detected. The messages are mainly to help identify
UFS write error problems, where the device will indicate the actual
major/minor number of the device that failed. This is a (serious)
annoyance with NFS clients as a server may go down at any time and the
system will spew out these messages for every active write. It is
likely that we will generate a patch to reduce (or completely
eliminate) the number of these messages for NFS."
For what it is worth, if the error(#) part of the message is errno
value that corresponds to the error it is ESTALE, "Stale NFS file
handle". This may offer a clue what the problem is. Unfortunately,
the device number seems to be zero (which may be a bug), so it will be
hard to track down the offending file system.
Digital UNIX V3.2 was the first version of UNIX that I've seen that
actually reported write errors from the buffer cache. Usually an
error just silently keep the data dirty in the hope that it would
succeed later. V3.2 appears to be noisy about the failures, if a bit
in error.
[snip]
Received on Wed May 17 1995 - 13:42:52 NZST