On Tue, Dec 28, 1999 at 03:22:11PM -0500, Claudio V. Peredo wrote:
>     It seems that I am able to run the backup only once.  It seems to
> create three of directories (one in the backup partition /backup1 on C
> named after the date (28Dec1999) and the others in the directory where
> the backup program is run also on C names after the machine names
> (A.domain, B.domain).  If I erase these directories then I am able to
> run backup again.
> What am I doing wrong?

We automatically remove those directories wheneven the data has been
streamed to tape. The backup program probably tries to avoid overwriting
backups done on the same day. If you wouldn't remove the directories,
the next days backups would end up in /29Dec1999/ and as this directory
doesn't conflict they will succeed.

>     Also, the backup process created 3 files in /backup1/28Dec1999/
> named:
>       1)  A.domain-7f000000.1000001  (root)
>       2)  A.domain-7f000001.1000002  (sources)
>       3)  B.domain-7f000001.2000001  (sources replica?)

7f000001 is a replicated volume with 2 replicas, 10000002 on server
A.domain and 20000001 on server B.domain. There is no `master-slave'
relation between the replicas so they are equivalent, but could
potentially be different if there are unresolved files or directories.
That is why all replicas are dumped separately instead of assuming that
one will be enough.

>     Is 1) the root volume?  Is 2) the replicated volume? Why is there a
> 3)?

Yes, 1 is the rootvolume, 2 and 3 are both underlying replicas of the
the replicated volume.

> I am trying to perform a restore of this backup, I brought file 2) from
> the backup coordinator to the SCM.  In the SCM, I ran:
>             volutil restore A.domain-7f000001.1000002 /vicepa
> 
> A new entry is created in /vice/vol/VolumeList listed as
> 'sources.0.restored' where sources is the name of the volume that was
> backed up.  Is sources.0.restored the name of the volume which contains
> the backed up data?

Yes, it is a read-only, non-replicated volume.

> I attempted to create a mount point for the new volume
> (sources.0.restored), from a client:
>     cfs mkmount /coda/test sources.0.restored
> 
> It creates an entry which looks like 'test -> #sources.0.restored' when
> I issue an 'ls -l' command.
> cfs lsmount /coda/test says: Dangling sym link; looks like a mount point
> for volume "sources.0.restored".

You have to rebuild the volume location database. Run `bldvldb.sh' and
it will recreate the VLDB from the VolumeList files. The servers will
pick up the VLDB file and the mountpoint should `resolve'.

Jan