Coda File System

Re: more backup woes

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Thu, 15 Apr 2004 13:31:25 -0400
On Sat, Apr 10, 2004 at 02:43:47AM -0700, Steve Simitzis wrote:
> when the disk fills up on the backup server, it causes codasrv to
> freeze up and hang indefinitely, until i kill it with -9. and the
> venus clients are stuck waiting forever. note that i'm still running
> 6.0.3 - perhaps this isn't a problem in the latest rev.

No that pretty much is still a problem. One more advantage of using
Amanda backups, because it tends to track the space used on the holding
disk and if necessary wait until some dumps to are spooled to tape, or
if the tapestreamer is unavailable it tries to make fewer (or no) full
dumps to reduce disk usage.

> 00:05:58 GetVolObj: Volume (1000052) already write locked
> 00:05:58 GrabFsObj, GetVolObj error Resource temporarily unavailable
... 28 more times.

This is kind of strange I guess the cloning is taking so long that the
backup server times out and retries the lock/backup sequence.

> 00:09:33 NewDump: file /vice/backup/7f000004.1000005.newlist volnum 7f000004 id 

It actually looks like it is nicely iterating over all the volumes,
eventhough the actual dumps are failing because of lack of diskspace.

There is no real indication why (or even that) the server got stuck.
Maybe there is an errant 'close(fd)' in the failure path that
accidentally closed the network socket.

Jan
Received on 2004-04-15 13:32:51