Netbackup Buffer Settings on RHEL 5

[Marcelino Mata]

I can't help you other than telling you I had similar problem with older
Netbackup 3.1 release under AIX.  Netbackup support insisted the problem
was with my network/hardware setup.  Spend 20-30 hours on the problem
with no resolution.  Once I upgraded to Netbackup 4.5 the problem went
away.  As they support so many platforms, you have to be very careful
what your running and try to stick to "mainstream" releases.  

MM

-----Original Message-----
From: redhat-list-bounces at redhat.com
[mailto:redhat-list-bounces at redhat.com] On Behalf Of Chuck
Sent: Thursday, November 15, 2007 5:00 PM
To: General Red Hat Linux discussion list
Subject: Netbackup Buffer Settings on RHEL 5

I was wondering if anyone has tweaked netbackup buffers for LTO2 drives
on RHEL 5 64-bit? (primarily the network buffer, but the number and size
of data buffers would also be nice) I have a 4 x 2.0 Ghz Xeon class
backup server running RHEL 5 64-bit and NBU 6.5. Im getting very poor
network backup performance (using both the NBU client and by mounting
filesystems via NFS and backing them up locally) It's almost 10 times as
slow as the identical backup done by a 2 x 1.5ghz sparc III solaris 10
based system. (both systems are at identical points in the network and
latencies to the backup clients are identical)

All the obvious things have been checked like NIC speed/duplex or memory
shortages. There are no lost packets or any other transport related
errors, its just grindingly slower. =p

My current settings are:

/usr/openv/netbackup/NET_BUFFER_SZ = 65535 (gone as high as 256k and as
low as 32k) /usr/openv/netbackup/db/config/SIZE_DATA_BUFFERS = 262144
(tried everything from 24k - 256k)
/usr/openv/netbackup/db/config/NUMBER_DATA_BUFFERS = 32 (tried 4, 16,
32, and 64 with no visible change in performance)

I have heard rumors from various vendors that there have been various
issues related to the generic scsi driver included in RHEL 5 (NBU uses
the pass-thru sg driver for various things). I was thinking this might
be the source of my performance problems. (however these issues the
vendor brought to my attention were not performance related) Even the
worst possible buffer settings shouldn't be 10 times slower than a
similar system.

Thanks for any info...
CC

--
redhat-list mailing list
unsubscribe mailto:redhat-list-request at redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/redhat-list