You say the backup has been running for 2 weeks. I guess the
question I would ask is, "Are you getting a decent throughput? i.e. if you
do a 'q occ' 1 hour apart, are you getting a reasonable increase or are you
getting very little?". If you get very little data, I'd pursue analyzing the
network and NIC settings. If you are getting decent throughput, but it never
completes, then your situation might be similar to what we had last week.
Lemme tell you what happened. Our NT admins move a large (300GB)
fileserver onto a NetApp. They back it up through an NFS mount, and of
course, the drive letter changed, so it had to do a full backup. (We are not
using NDMP at this time, but I'm willing to listen to any reports of the
good/bad points of NDMP on TSM. I believe that using NDMP, I will have to
move all 300GB once a week for the full backup, and that sounds just
painful).
Anyways, we let it run for a couple of days, and it finally gets to
about 300GB of data pushed to the TSM server. It should be about done, yet
it keeps on going. What? how can the initial backup of a 300GB filesystem
backup 350 GB? We get the NT admins to look at it and surprise! The NetApp
is doing a nightly snapshot and the inclexcl on the client is not excluding
any of the '.snapshot' directories that are all renamed/moved nightly, so
TSM is trying to backup each file on the host 7 times, regardless of whether
it has changed or not. Ouch!
We got the inclexcl cleaned up to only backup 1 copy of the data and
it should be OK now.
Thanks,
Ben
-----Original Message-----
From: Guillaume Gilbert [mailto:guillaume.gilbert AT DESJARDINS DOT COM]
Sent: Wednesday, July 31, 2002 2:31 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Backup of Win2`K Fileserver
Hello
I have a BIG performance problem with the backup of a win2k fileserver. It
used to be pretty long before but it was managable. But now the sysadmins
put it on a compaq
storageworks SAN. By doing that they of course changed the drive letter. Now
it has to do a full backup of that drive. The old drive had 1,173,414 files
and 120 GB of
data according to q occ. We compress at the client. We have backup retention
set to 2-1-NL-30. The backup had been running for 2 weeks!!! when we
cancelled it to try to
tweak certain options in dsm.opt. The client is at 4.2.1.21 and the server
is at 4.1.3 (4.2.2.7 in a few weeks). Network is 100 mb. I know that journal
backups will help
but as long as I don't get a full incremental in it doesn't do me any good.
Some of the settings in dsm.opt :
TCPWindowsize 63
TxnByteLimit 256000
TCPWindowsize 63
compressalways yes
RESOURceutilization 10
CHAngingretries 2
The network card is set to full duplex. I wonder if an FTP test with show
some Gremlins in the network...?? Will try it..
I'm certain the server is ok. It's a F80 with 4 processors and 1.5 GB of
RAM, though I can't seem to get the cache hit % above 98. my bufpoolsize is
524288. DB is 22 GB
73% utilized.
I'm really stumped and I would appreciate any help
Thanks
Guillaume Gilbert
|
