Coda File System

Re: inconsistent object

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Thu, 30 Mar 2000 15:59:07 -0500
On Thu, Mar 30, 2000 at 10:25:08AM +0200, Alan Scheinine wrote:
> I was not able to connect coda directories and in the file
> 
>    /usr/coda/etc/console
> there was the warning
> 
>    19:46:41 Local inconsistent object at
>    /coda/usr_local_L2.2/stow/pgplot-5.2.0/pgplot, please check!
> 
> but when I do a repair, and in particular the command
> 
>    beginrepair /coda/usr_local_L2.2/stow/pgplot-5.2.0/pgplot
> 
> I am told
> 
>    object not in conflict
> 
> The above object is a directory.  If there is an object inside
> pgplot that is in conflict, how do I find it?

This is a nasty situation that luckily doesn't seem to be occurring that
often lately. There is a server-server conflict on the pgplot directory,
however there are still unreintegrated modifications in the CML which
keep the parent directory pinned on the client. As we can't reintegrate
any further (conflict), and can't refetch the parent directory (dirty
object) Coda is unable to represent the problem.

To fix it, either repair the server-server conflict from another client,
or purge the CML (cfs purgeml /coda/..../pgplot-5.2.0) losing all the
pending updates. The newly created files can be rescued by copying the
checkpointed cml-tar file from /usr/coda/spool/<uid>/xxx.tar to a safe
place.

It is the pgplot-5.2.0 directory that needs to be rewritten, if after
the purgeml the conflict doesn't yet appear run 'cfs endrepair' on the
pgplot-5.2.0 dir. This sends a downcall to the kernel to have it
revalidate the directory contents. And if that fails, some internal
reference count went wrong and restarting venus is necessary.

Jan
Received on 2000-03-30 16:02:04