Coda File System

Re: Finally, 5.3.17...

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Thu, 29 Nov 2001 13:26:08 -0500
On Thu, Nov 29, 2001 at 11:50:19PM +0900, Stephen J. Turnbull wrote:
> Something that does seem buggy: doing a purgevol_rep on it removed it
> from the VRList, but not from the BigVolumeList.  Then creating
> another volume with the same name gave two volumes with the same name
> in BigVolumeList.  I edited out the old one by hand.  Is there a
> correct way of getting those lists recreated?

BigVolumeList is created by bldvldb.sh. We query the servers for
available volumes and concatenate the replies.

Check the BigVolumeList after running bldvldb.sh, if there is any volume
name duplication some volume replica didn't get purged from the servers.
If that is the case, make sure which is the actual replica used by the
replicated volume, and which is the not-yet-purged volume. The list of
volumeid's in /vice/db/VRList should help identifying the right one.

> I don't know when cfs gmt appeared, but I love it.  Quick shell hack:
> 
> for i in `cut -f 2 -d ' ' /vice/db/VRList`; do cfs gmt $i; done
> 
> Comes in handy when you've been mkm/rmm and create/purge broken
> volumes, and don't remember what's currently mounted or where exactly.
> But you seem to have to actually list the directories (ie, at the
> mount points) to get anything?  When venus first comes up, all you get
> is the root volume.

Yes, we only know the mountpoint once we've traversed the 'magic
symlink'. I believe network disconnections might actually trigger the
client to turn all mounts back into 'magic symlinks' to force
revalidation of all mountpoints so it would lose the "where is this
volume mounted" information too.

Jan
Received on 2001-11-29 13:26:12