Networker

Re: [Networker] SAN Backup Block size Woes..

2003-08-27 12:00:43
Subject: Re: [Networker] SAN Backup Block size Woes..
From: Davina Treiber <treiber AT HOTPOP DOT COM>
To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
Date: Wed, 27 Aug 2003 12:00:38 -0400
On Wed, 27 Aug 2003 10:04:40 -0500, Kent Holloway <Kent.Holloway AT ONVOY DOT 
COM>
wrote:

>Actually I think the problem has only to do with block size.

Suit yourself if that's what you want to think. I think you should consider
other possibilities. I'm not making this stuff up, I have been there myself.

>I can backup clients of any size if they are Solaris to this tape library.
>Its only when I start backing up Windows hosts that the problem appears.

Solaris is always more robust on a SAN because it supports SCSI
reserve/release, and Windows just doesn't. Thus when your Solaris box is
using the tape drive, it has a lock on it. When the Windows box is using
the tape drive, it just trusts to luck, and if something else on the SAN
decides to send a SCSI reset or some other SCSI command to the drive, it
has a good chance of succeeding. This could result in the tape drive
rewinding, or having superfluous data written to it. NetWorker will
complain at this because it finds unexpected blocks of unexpected sizes.
Windows just isn't good at this sort of stuff, it's not really a server
class OS in this sense. Complain to Bill.

>We are also using zoning on the SAN switches.
>I have 2 zones that contain the following
>
>CrossRoads10K and Networker Server
>CrossRoads10K and Windows DDS Storage Node
>
>The Storage node sees the tapes and changer just fine as does the Solaris
machine.
>All of the Solaris hosts I have tested with show no problem writing large
or small amounts of data to the tapes.
>The block size (On the Networker Solaris server) via scanner shows up as
128K.
>If I label a tape on the Windows DDS Storage Node the tape shows up as 64K
instead of 128K.
>
>This is terribly frustrating because there is so little documentation for
this and what I have found all points to Adaptec SCSI cards registry
settings and thats it.

Most people with pure SCSI environments are using Adaptec. Emulex fibre
cards don't have this problem, that's why you're not seeing anything
written about it. I assume that you have up to date firmware and drivers on
your Emulex card?

>
>I have made the changes that the Legato Admin guides say you need to make
when you have this setup (NSR_DEV_BLOCK_SIZE_SDLT=128) on both the
Networker server and the Storage Node.

Truth is, you shouldn't even need to do this. It will take the default of
128KB for sdlt if the hardware allows.

>Thanks again for all the replies!
Well, we're trying...

>No solution yet but still hoping.. :)
If you don't get anywhere, you might need to get a specialist in to analyse
the packets on your SAN. Crossroads should have the right sort of kit. This
is the best advice I can offer at the moment.

--
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.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=