Networker

[Networker] Making NDMP selective backups faster

2007-04-04 11:27:03
Subject: [Networker] Making NDMP selective backups faster
From: Stan Horwitz <stan AT TEMPLE DOT EDU>
To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
Date: Wed, 4 Apr 2007 11:20:14 -0400
I have been experimenting with NDMP selective backups of a Mirapoint message store that's available to me for testing purposes. For my experiments, I recovered around 500GB worth of data to it from a production Mirapoint server. In production, we do regular image based full backups nightly of our six production Mirapoint servers, but what we want to do is selective backups which will enable us to recover individual mailboxes.. As it is now, if a user needs to recover a specific message or calendar entry, we have to recover the entire message store, which is an overnight process.

I appended the most recent savegroup report onto the end of this message. The problem is that when I try to do selective backups on my R&D server, I end up bringing my NetWorker server to its knees while the history data is sorted. Backups also take a lot longer in this mode, compared with the regular NDMP backups.

I am wondering if anyone has any tips speeding up this sort of NDMP backup. I suspect I will need to throw a faster NetWorker server at the problem, so I am wondering how to size a new NetWorker server to allow concurrent NetWorker NDMP selective backups of five message stores, each with around 500GB worth of data.

We are using dedicated NDMP tape drives that are connected to the NetWorker server via a GB NIC into the NetWorker server, but 100MB out of the individual message stores. Going to GB on the message stores themselves has not shown any improvement at all with NDMP backups.

Our NetWorker server is a Sunfire V480 running Solaris 9 with NetWorker 7.2.1. We recently upgraded that server from one processor board to two and doubled the RAM. This has not helped with improving NDMP backups significantly. The tape library involved is a Sony PetaSite with 14 fibre attached S-AIT tape drives, five of which are dedicated to NDMP backups.

The next fiscal year here begins in July and I would like to size up a new NetWorker server to replace our V480 to purchase next fiscal year. I am wondering what would be a good choice for a new server in terms of processor, RAM, and I/O channel. My options are limited to Dell and Sun, for reasons I prefer not to discuss because it just gives me a head ache!

One of the things I expect to do in the next month is add a Sun X4500 with 12TB of usable disk space as a storage node. This machine will be fibre attached to my PetaSite's SAN fabric and I expect to use it to do disk-to-disk-to-tape NDMP backups of our Mirapoint message stores and a few other servers, and I am wondering if anyone else is doing disk-to-disk-to-tape NDMP backups and if you saw any better throughput or at least better data compression on your tapes.

Here is the  most recent savegroup report from my testing ...xxxxxxxx

NetWorker savegroup: (notice) xxxxxxxx completed, total 1 client(s), 0 Hostname(s) Unresolved, 0 Failed, 1 Succeeded.
Start time:   Mon Apr  2 10:12:44 2007
End time:     Wed Apr  4 00:01:17 2007

--- Successful Save Sets ---

Successful Save Sets written to verified media noted with "V" in first column.

* xxxxxxxx.temple.edu:/usr/store savegrp: suppressed 83 lines of output -
check daemon.log for details.
* xxxxxxxx.temple.edu:/usr/store
* xxxxxxxx.temple.edu:/usr/store nsrndmp_save: The NDMP backup is complete. xxxxxxxx.temple.edu: /usr/store level=full, 557 GB 25:42:52 17529139 files * xxxxxxxx.temple.edu:/usr/store nsrndmp_save: Save session closed with NW server successfully
* xxxxxxxx.temple.edu:/usr/store
* xxxxxxxx.temple.edu:/usr/store nsrndmp_save: Sorting File History....
* xxxxxxxx.temple.edu:/usr/store nsrndmp_save: Sorting File History completed Successfully in 10:56:42 Hours
* xxxxxxxx.temple.edu:/usr/store
* xxxxxxxx.temple.edu:/usr/store nsrndmp_save: Processing NDMP File History... * xxxxxxxx.temple.edu:/usr/store nsrndmp_save: xxxxxxxx.temple.edu:/ usr/store Processing NDMP File History completed Successfully on 'bootz' in 00:58:13 Hours
* xxxxxxxx.temple.edu:/usr/store
V xxxxxxxx.temple.edu:/usr/store nsrndmp_save: browsable savetime=1175523168
* xxxxxxxx.temple.edu:/usr/store nsrndmp_save: Successfully done
* xxxxxxxx.temple.edu:index rcmd localhost, user root: `save -s bootz -S -g xxxxxxxx -LL -f - -m bootz -V -l full -LL -W 78 -N index:e646bd64-00000004-40b81b02-40b81b01-00010000-9bf7a61a /nsr2/ index/xxxxxxxx.temple.edu' * xxxxxxxx.temple.edu:index /nsr2/index/xxxxxxxx.temple.edu/ db6/46031f8c.rec * xxxxxxxx.temple.edu:index /nsr2/index/xxxxxxxx.temple.edu/ db6/455dfccf.rec * xxxxxxxx.temple.edu:index /nsr2/index/xxxxxxxx.temple.edu/ db6/46110f60.rec
* xxxxxxxx.temple.edu:index /nsr2/index/xxxxxxxx.temple.edu/db6/
* xxxxxxxx.temple.edu:index /nsr2/index/xxxxxxxx.temple.edu/
* xxxxxxxx.temple.edu:index
V bootz: index:xxxxxxxx.temple.edu level=full, 4159 MB 00:10:34 5 files

To sign off this list, send email to listserv AT listserv.temple DOT edu and type 
"signoff networker" in the body of the email. Please write to networker-request 
AT listserv.temple DOT edu if you have any problems with this list. You can access the 
archives at http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER

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