Coda File System

Re: Tr: Greetings and First Crash.... :o)

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Mon, 20 Oct 2003 23:23:51 -0400
On Tue, Oct 21, 2003 at 03:22:01AM +0200, Lionix wrote:
> ( stable vs unstable, tried both and a machine upgrade ), fines paquages
> upgrades, postinst preinst script... Greetings !
> Upgraded everything friday ! Could not get out of it since that day !

I am not yet completely happy, first of all, installing the packages
places things in different locations compared to building from source
(and uses a completely different set of init scripts).
Second of all the codasrv pre/post install scripts stop the running
auth2/update/codasrv daemons, but don't start them again once the
installation is done.

> I Followed the process on the web site , dumped-rvm worked fine, rdsinit...
> Got an error message at the end of restoring work : 
> WARNING: Volume ResOn = FALSE, fix norton to create empty log.
> WARNING: Volume ResOn = FALSE, fix norton to create empty log.

For some reason I remember this problem and though it has been fixed a
long time ago. Basically all volumes with only a single replica do not
need a resolution log, so this warning shouldn't really matter in
practice.

However...

> then the scm crash with : 
> scm:/# cat /vice/srv/SrvErr
> codasrv: sftp1.c:1181: sftp_AllocPiggySDesc: Assertion `p->vmfile.SeqBody'
> failed.

I guess the server tried to resolve something and wanted to ship the
current resolution log to another server, but the log doesn't exist so
we end up passing a NULL ptr to the SFTP data transfer. I don't really
understand why it tried to resolve a singly replicated volume though.
Who would it resolve with?

I guess there are still parts of the code that actually expect that we
always have a resolution log around. That's a bit annoying because it
makes it a bit hard to use norton-reinit as a means to upgrade the
servers to a new RVM layout.

> I suppose I got to create_vol a new volume, but it will generate a new vol-id,
> and then the backup won't restore !

I would say, create a new volume and copy the data from the original
rootvolume or the backup in there and reset the ACLs and volume mount
points.

Jan
Received on 2003-10-20 23:25:27