I guess I'm just in the habit of not setting up indexing on static data
that we run one-time fulls on since these guys are typically rather
large directories, so this does conserve space on the client's index. As
such, I'm in the mind frame to simply turn it off on all static fulls
even though in this case they're single files, and they're only backed
up a few a month. Actually, it's the pools that we send static data to
that I tend turn indexing off on, so it's just a habit. Seems easy
enough to run saveset recover if anyone needed anything since all the
savesets are named in a manner to make them easily identifiable. In the
case of a directory, we'd just recover the whole thing even though they
might only need one file. They don't mind waiting, and at some point it
would no longer be in the index anyway since it's static. All it would
buy us is the browse policy which is 2 months, and in most cases it's
worth the savings.
George
"Shelley L. Shostak" wrote:
>
> Hi George,
>
> While turning off indexing can save lots of space, I am betting that there are
> so few files from the database dumps that it does not have much on an impact
> on space or performance. Now my filesystem with 8 million inodes would have
> greatly benefited from no index entries!
>
> Shelley
>
> On Tue, 25 Nov 2003, George Sinclair wrote:
>
> > Date: Tue, 25 Nov 2003 13:32:14 -0800
> > From: George Sinclair <George.Sinclair AT NOAA DOT GOV>
> > To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
> > Subject: Re: [Networker] wild cards with mminfo?
> >
> > We have indexing turned off because because there is no need to have it
> > turned on for these files. These are Oracle database dumps. They can be
> > easily recovered via the saveset recover tool. We can instantly identify
> > them by their name "name.mm.dd.yy.dmp" and "name.mm.dd.yy.dmp.log"
> > respectively. We have no need to browse on these because we are not
> > backing up directories, only these individual files. Here is another
> > important aspect to all this. These go to their own pool that has
> > indexing turned off. These are skipped from the nightly pool. Tapes in
> > this pool can be recycled based on request only from the DBA. We clone
> > these tapes, too. I wanted them off of the nightly pool so if and when
> > we can recycle them, they won't affect other non-related data; data that
> > we might not want to recycle yet. The idea was to rotate the tapes. So
> > Turning indexing on just didn't seem necessary.
> >
> > George
> >
> > Pat OBrien wrote:
> > >
> > > I think that you need to utilize mminfo -query 'name> = < value and then
> > > into a -report. I have been following your recent posts and suspect I
> > > can forward advice. i too, recieve many requests for backups. We once
> > > attempted to train users to submit requests by directory which in our
> > > case is usually a mount point. We in IT actually launch the request with
> > > a script that I can not share. I am sorry. The script just parses the
> > > requested info bu mountpoints and requested directories, and launches a
> > > bunch of smaller "save jobs" in parralell (save params params directory &
> > > (ampersand)).There are a couple things to consider though, max chars per
> > > line being one which we had problems with and throttling the number of
> > > parralel jobs when independant saves were several hunderd per client. We
> > > over come this with our super scripter to actually created sub-scripts,
> > > and controller software scheduler ( basically a monitor of this running
> > > and needing to be run and using unix "at comm!
> and")for
> > > each save. Parameters at our site to be considered were that different
> > > saves have different browse/ retention periods. This is a lot of words
> > > getting to the point which I think you are looking for . save -LL will
> > > add a extra line of output to the save which is basically the nsavetime
> > > for the client. This param can be fed directly in "nsrinfo" to review
> > > files in the saveset, though with indexing off, I confess to not know
> > > exactly what will happen. Why do you have indexing off?
> > >
> > > -----Original Message-----
> > > From: George Sinclair [mailto:George.Sinclair AT NOAA DOT GOV]
> > > Sent: Tuesday, November 25, 2003 12:20 PM
> > > To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
> > > Subject: [Networker] wild cards with mminfo?
> > >
> > > Anyone know any way to use wild cards with the mminfo command?
> > >
> > > For example, if I wanna look for savesets from a particular client that
> > > are all of the form:
> > >
> > > /raid/home/user/Backups/fname.11.??.03
> > >
> > > do I have to narrow the search as much as possible and then pipe to grep
> > > like:
> > >
> > > mminfo -s server -q
> > > 'client=client_name,savetime>=start_date,savetime<=enddate' | grep
> > > fname.11.
> > >
> > > or is there some way I can use wild card notation inside the query?
> > >
> > > Thanks.
> > >
> > > George
> > >
> > > --
> > > 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.
> > > =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
> >
> > --
> > 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.
> > =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
> >
--
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.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
|
