[13609] in Athena Bugs
sun4 7.7R:
daemon@ATHENA.MIT.EDU (Diane Delgado)
Fri Jun 23 16:25:15 1995
To: bugs@MIT.EDU
Date: Fri, 23 Jun 1995 16:25:12 EDT
From: Diane Delgado <delgado@MIT.EDU>
System name: green-acres
Type and version: SPARC/Classic 7.7R (1 update(s) to same version)
Display type: cgthree
AFS client corrupts cache when local cache partition fills up.
Each time I've run across corrupted files, its been after the local cache
has filled up.
Included are the cache usage stats, since some people have expressed
skepticism regarding the fact that this could occur (This is during
the link phase for the backup system).
>AFS using 34617 of the cache's available 34637 1K byte blocks.
Shortly after this I got the AFS cache full message; I then
attempted to re-execute the link, and ld core dumped because
it claimed one of the *.o's (which was fine before this)
was now somehow "bad".
ld: warning: file master_xdr.o: section *** Signal 11 - core dumped
make: Fatal error: Command failed for target `master'
The workaround for this which is either to truncate the cache
or flush possibly bad data back to the server isn't acceptable
since one can loose actual data, if AFS happens to select a file
that you've been actively editing as the thing it decides to
corrupt (this has also happened).
