Networker

Re: [Networker] Merits of sending indexes and bootstrap to separate pool?

2004-01-27 17:30:35
Subject: Re: [Networker] Merits of sending indexes and bootstrap to separate pool?
From: Davina Treiber <Treiber AT HOTPOP DOT COM>
To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
Date: Tue, 27 Jan 2004 10:14:51 +0000
George Sinclair wrote:

I have some questions here regarding opinions and techniques on using a
separate pool for indexes and bootstrap.

Question 1. Does anyone have any philosophies, suggestions of criticisms
on having the client indexes instead go to a separate pool of tapes from
the actual data?

It depends on your environment. In lots of cases it is a good idea.

One reason is because of varying retention policies. You may have
different retention policies for data and indices, so mixing them on the
same tape may not be a good idea.

You can extend the above principle by operating a system where an
occasional set of indices (say twice a month) is kept for a very long
time, so that you can easily make a tape browsable again. This means you
don't need to have very long browse policies but never need to scan a
tape. I'm sure I've posted some details of this before, it should be in
the archives.

Another reason is because of having multiple pools and storage nodes. If
you write to a pool on a storage node, you would also need to have a
tape labelled for the same pool on the server in order to write the
index data. If you have just a single drive or a small library this can
cause a problem, but putting index and bootstrap data in a special pool
can avoid this.

There are some good reasons not to do this too, but Thierry has covered
some of these.


Question 2. What about the bootstrap? Is this always a level full
regardless of the level backup that the server is running?

Bootstrap is always level full. It's not very big and doesn't need to be
kept very long either.

Question 3. What about having the server go to a separate pool?
Unless your NetWorker server has other data on it, there's only a minor
difference between defining index and bootstraps to go to your special
pool, and doing the whole server. It just requires a slightly different
pool configuration, either specifying the client name or specifying the
save set names. Your choice.


The way I see it: The only thing on the server that seems worth sending
to a special pool would be the server's index and bootstrap. The client
indexes get backed up when the clients themselves back up, and the rest
can be recovered from the regular pools. I'm thinking we could have the
server's bootstrap and index also go to this same IDX pool?

I agree.


Question 4: How to get the bootstap to go to this same pool?
Partly depends how you are defining the IDX pool. If you specify the
save set as ^index: (haven't we discussed this before) you can also
specify another save set of ^bootstrap$. If you just define the server
in the clients list of the pool, it'll go there anyway.

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

<Prev in Thread] Current Thread [Next in Thread>