Networker

Re: [Networker] Cloning questions

2005-01-12 13:49:43
Subject: Re: [Networker] Cloning questions
From: Joe Moore <joseph.f.moore AT UGS DOT COM>
To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
Date: Wed, 12 Jan 2005 13:36:43 -0500
Darren Dunham wrote:

A few questions here on cloning.

o We're running an older copy of NetWorker (6.1.1) so I guess cloning
works differently wherein the entire saveset is cloned, including all
the pieces that may live on subsequent volumes, as opposed to just the
piece on the given tape which I wanna say is how nsrclone works under
newer releases?

Cloning is always done at the saveset level.  So all versions will clone
an entire saveset, not a subset.

o Let's suppose I have a save set that is spread across two or more
tapes but maybe could fit on one tape. Obviously, all the pieces have
the same ssid. Okay, now let's say I run nsrclone against that ssid, so
NetWorker loads all the tapes one by one and then clones it off to this
one clone volume I have.

Right.

Now, let's suppose I then recycle (relabel) one
of the affected original volumes. What happens when I try to recover
that saveset?

When you recycle a volume, Networker deletes the record of any saveset
fragments on it from the media database.  So it knows that the original
backup of that saveset is no longer complete and cannot be used for a
restore.

It depends if you mean "recycle" or "relabel".

When you mark a volume recyclable, it marks all the SSIDs on the tape as recyclable. This includes all copies of that SSID, since the retention time is a function of the SSID, not the (SSID/cloneid).

When you relabel the tape, it simply deletes all the media database records that reference that volume ID.

Obviously, one of the original tapes is gone, but pieces
of the save set still live on the other volumes, and the complete
saveset lives on the clone, so what will NetWorker do? Will it matter
that I'm running nwrecover versus save set recover?

No.  It'll always choose the clone.  Further, if you had not cloned it,
you would not be able to recover from any of the other fragments
(however you might be able to use 'scanner' directly on the fragments to
write data directly to disk).

Actually, it will choose whichever non-suspect copy has the lowest "cloneid", which is almost always the original. The exception is on adv_file devices, where it's the .RO copy that has the lowest.

As far as I can tell, the cloneid appears to be the time the clone was started. (in seconds past the epoch)

o What's this business of NetWorker wanting to use original versions and
not clone copies so you have to mark the original suspect to force it to
use the clone copy? Guess that's only true if the original is still
listed in the database, i.e. none of the affected tapes has been
recycled (relabeled)?

Right.  If you have multiple copies, Networker should always choose an
on-line copy for restore.  If none (or multiple) copies are online, then
you might have to use the suspect trick to force the one you want.

I wanna say that when one of the affected volumes is relabeled, all of
the savesets on that tape will now be removed from the database, never
mind whether the savesets are completely contained on the volume or
not.

Actually, I'm not 100% certain if the entire saveset is removed from the
DB, or if only the fragments that were on that volume are removed.  I
can see either one being correct.

I'm pretty sure that the entire saveset is removed, since we have lots of volume-spanning backups, and I often see several recyclable tapes that have no savesets on them. i.e some other part of those savesets have been relabeled.

As a result, none of the pieces would be recoverable from any of the
tapes, so when I run saveset recover or nwrecover on the affected
saveset that was cloned, NetWorker will only use the clone copy, and
everything should behave as if it was the original?

Yup.

--Joe

--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listserv.temple DOT edu or visit the list's Web site at
http://listserv.temple.edu/archives/networker.html where you can
also view and post messages to the list. Questions regarding this list
should be sent to stan AT temple DOT edu
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=