On Wed, Oct 13, 1999 at 03:57:02AM +0200, Kurt Huwig wrote:
> Thanx for your fast answer.
> 
> The documentation about backup doesn´t say anything about how and where
> to run the 'backup' program - the documentation about backup is somehow
> outdated - there are no file like 'host' and 'file.. I supposed to have
> it to run on the SCM.

Not necessarily, we have a dedicated backup machine, which uses the
updateclnt to pull in /vice/db from the SCM, just like a non-SCM server.

These are the mounted partitions,

    /dev/sda1 on /backup type ext2 (rw)
    /dev/sda2 on /backup1 type ext2 (rw)
    /dev/sda3 on /backup2 type ext2 (rw)
    /dev/sda4 on /backup3 type ext2 (rw)

The vicetab on backupbox contains:

    backupbox     /backup1    backup
    backupbox     /backup2    backup
    backupbox     /backup3    backup

It then runs backup.sh from a cronjob, which calls:

    backup -t 135 /vice/db/dumplist /backup > $BACKUPLOG 2>&1

After this program is finished we end up with something like:
/backup contains symlinks to the actual places where backed up volumes
are stored in the backup volumes.

    /backup/<date>/server1.coda.cs.cmu.edu/7f000001.c1000001 -> /backup1/...
    /backup/<date>/server2.coda.cs.cmu.edu/7f000001.c1000001 -> /backup3/...
    /backup/<date>/server3.coda.cs.cmu.edu/7f000001.c1000001 -> /backup2/...

The partitions specified in vicetab contain the actual dumped volumes,
the backup program puts any dumped volume in the emptiest partition to
balance the diskusage.

    /backup1/<date>/server1.coda.cs.cmu.edu-7f000001.c1000001
    /backup2/<date>/server3.coda.cs.cmu.edu-7f000001.c3000001
    /backup3/<date>/server2.coda.cs.cmu.edu-7f000001.c2000001

So backup is essentially a pull-model, /vice/db/dumplist contains the
replicated volume id's and the backup schedule. The VRDB and VLDB (and
servers?) files are used to figure out what servers have replicas of
these volumes.

> where 'smuf' is the SCM. Without the last line, it bails out with 'No
> diskpartitions for backup!! Bailing out.'

That's a trick to be able to use the same vicetab on all machines. If
the first line doesn't match the hostname of the current machine, it is
ignored during the parsing.

> BTW: There is no documentation about how to set up a replicating server,
> i.e. a server that is in the same VSG and replicates volumes. I can
> write a HOWTO and put it on my webserver, but where to send it for
> public use?

We don't really have a lot about replicated servers, the best bit is
probably in the coda-HOWTO.
 (http://www.coda.cs.cmu.edu/doc/html/coda-howto-4.html#ss4.8)

Too bad that large parts of the more indepth "Coda User and
Administration Manual", have become severely outdated in less than 2
years.

If you got something like a howto for replication, we can mirror it on
the coda-website and/or have a link pointing back to your site.

Jan