"Steven G. Parker" wrote:
> I somehow managed to corrupt my 18 gig XFS filesystem. It ran
> fine for several months, but now it is quite trashed. I filled it up
> with a huge write() call, and then shortly thereafter things started
> failing. I got I/O error reading ".", and couldn't really do much
> so I rebooted.
> The first time I rebooted, it wouldn't mount the partition. So I
> ran xfs_check on it and got 485k lines of crud. xfs_repair core
> dumped in phase 3, agno = 0 after several "imap claims a free inode..."
> and "bad version number 0x0..." and "bad magic number 0x0..." messages.
> After subsequent reboots, I can now mount the filesystem, but xfs_repair
> still dumps core. The mounted filesystem has now lost much of it's
> contents, and I get messages like this from ls:
> ls: /home/sparker/News: No such file or directory
> So the question is: How can I turn this catastrophe into something
> useful for the XFS team? Do you want a copy of the filesystem image?
> Will you be able to do anythign useful with it if I send it? It may take
> me a few days to get it there, so could I give somebody an account on
> this machine to poke at it? Do you want the output of xfs_check?
> Please let me know how I can be useful.
Sorry for not getting back to you sooner...
Ahh lets see, well I really don't don't think sending
and 18gig image across the wire would be a good idea.
The best thing would be if we could access the system directly
and poke around a bit with xfs_db.
Other things that might be insightful:
any console messages that might have appeared when the FS went bad.
and/or messages in /var/log/messages.
You could send us the output from xfs_check and or xfs_repair.
It's doubtful it will revile anything useful but it is always worth a