On Tue, Aug 14, 2001 at 04:21:49PM +0000, irbis_at_orcero.org wrote:
>  Hello, Jan!
> 
> > >  In fact, CML belongs to root, and all the servers are running as root.
> > > Should I run venus and the different server daemons as coda user?
> >
> > Not really necessary, but you will have to give root a Coda token so
> > that it can reintegrate the changes. Coda doesn't allow a user with
> > userid '0', so you need to authenticate by using the <username> argument
> > to clog.
> 
>  This does not work, I have tested it. In fact, I thing that it is not
> related to venus. All this things happed on a replicated server, and I am
> begining to thing that it is something related to replication, because I
> lost on the tests the contact with one node with cmon -telnet and ping
> see the "lost replicated node", after this an error:
> 
> kernel: coda_read_super: device index: 0
> kernel: coda_read_super: rootfid is (0x7f000000,0x1,0x1)
> kernel: Failure of coda_cnode_make for root: error -110

Erhm, -110 is ETIMEDOUT. So the object that should get mounted on /coda
wasn't cached by venus and venus was unable to contact any of the
servers listed as 'rootservers' in venus.conf, or unable to contact the
server(s) that store replicas of the 7f000000 volume.

> CHILD: mount system call failed. Killing parent.

Yup, venus was unable to get the object that is mounted, so the mount
call fails and venus commits mass suicide, or at least the child that is
performing the mount is killing it's parent.

>  It looks that there is no way to mount any coda volume, neither to
> recover any data.
> 
>  It looks that under normal pressure, my coda replicated servers fails.

Are you sure that 7f000000 is a replicated volume? Could you send me a
tcpdump during a failing venus startup.

    # tcpdump -i any -w trace.dump port codasrv or port codasrv-se

>  Any sugestion?

Compress the dump before sending it to me (not to the list).

Jan