[154] in Info-AFS_Redistribution
Full AFS partion on RS/6000
daemon@ATHENA.MIT.EDU (Hakan Torbjar)
Tue Jun 18 03:07:42 1991
Date: Tue, 18 Jun 91 06:51:11 GMT
From: Hakan Torbjar <torbjar@utc.chalmers.se>
To: info-afs@transarc.com
The other day we encountered a problem when an AFS-partition on an
RS/6000 became full. In the process of correcting it we didn't save
any log files so some parts below are unfortunately from memory only.
When referencing a file on the partition there was a long delay before
the next prompt. Bos status reported that the fileserver was killed by
signal 6 and that it had restarted. One user was missing a volume
that fileserver reported was unattched and should be salvaged. The
salvage failed with a (three line?) message containing `... core
dumped!' on the last line.
Now the real problem began: It was impossible to move a volume away
from the partition, because the cloning of the source volume failed
because of the lack of disk space (but I don't think that fileserver
aborted in this case). Attempts to (re-)move files resulted in a dump
and restart of the fileserver.
The solution this time was to extend the file system, but what happens
when the partition spans the whole disk? Having a "buffer file" is no
solution since it apparently can't be removed. Perhaps a "buffer
volume" (containg the "buffer file") can be removed without requiring
the extra disk space needed for a volume move?
Is this a problem only with the AIX 3 file system that doesn't have
the "stretchability" of a traditional UNIX file system?
--
H}kan Torbj{r : Dept. of Computer Engineering/UTC : qui me amat,
torbjar@utc.chalmers.se : Chalmers University of Technology : amat et
(+46) 31-72 40 69 : 412 96 G\TEBORG, Sweden. : canem meam
torbjar@ctp.se.ibm.com : IBM/4C, G\TEBORG, Sweden. :
