On Fri, Sep 29, 2000 at 04:50:25PM -0400, Douglas C. MacKenzie wrote:
> backup questions.

A backup volume is a readonly replica that is kept on the server. It
normally doesn't take up that much space because all the files are
shared with the volume it is a backup off (copy-on-write semantics).

So for a doubly replicated volume, let's say vmm:u.hmpierce, we have two
underlying replicas. To distinguish between the two, the createvol_rep
script has created them with numeric suffixes, vmm:u.hmpierce.0 and
vmm:u.hmpierce.1. These underlying volume replicas are the volumes that
actually contain the data on the servers, the replicated volume is
simply a logical volume that only exists in the mind of the Coda client.

So the server that stores the first replica vmm:u.hmpierce.0 will (after
the backup is made) also have a volume vmm:u.hmpierce.0.backup. This is
the readonly clone which is dumped during backups.

A client can mount this volume, because it is still exported by the
servers, but it is unrelated to the volumes on the backup-server (or
tape).

Codaservers keep track of previous dumplevels so that they can correctly
select the files required for the incremental (or differential) dump.
The backup program use the dumplevels in /vice/db/dumplevels to pick the
wanted incremental, so that it doesn't always make a full dump.

Does this answer the question, or am I just rambling?
I better go home now.

l8r,
    Jan