Veritas-bu

[Veritas-bu] yes sir, obviously insane (6.0 upgrade question)

2005-12-01 07:58:43
Subject: [Veritas-bu] yes sir, obviously insane (6.0 upgrade question)
From: thoke AT northpeak DOT org (Tim Hoke)
Date: Thu, 1 Dec 2005 06:58:43 -0600
What you refer to is a problem NetBackup runs into because of the 
limited kernel support for passthrough devices.

Basically, for a given tape drive, NetBackup needs to be able to 
quickly map the data path (/dev/nst) to the passthrough path (/dev/sg) 
so that SCSI Reserve/Release (among other things) can happen properly.  
In order to do this, there's a "make_scsi_dev" command which will 
determine the mappings and create links in the appropriate /dev/st/ and 
/dev/sg/ dirs.  In pre-2.6 kernels, this is the only way to do things 
and as such, since the make_scsi_dev also includes a scsi inquiry 
command, devices which are reserved won't be linked.  In 2.6 kernels, 
there's native mapping which makes this processing unnecessary, but I'm 
not sure what it looks like if one of the devices is busy.  Seems to me 
this is a limitation on any OS... If the drive's reserved to one host, 
another host can't inquire it and therefore can't determine what it is 
in order to configure it.

HTH
-Tim

On Nov 30, 2005, at 5:55 PM, Mark.Donaldson AT cexp DOT com wrote:

> In our case, we could.  However, if the tape drive was reserved by 
> another
> server, then it wouldn't be properly discovered & built into 
> Netbackup's
> config stuff.
>
> The OS could see an unassigned one just fine.
>
> -M
>
> -----Original Message-----
> From: Jeff Lightner [mailto:jlightner AT water DOT com]
> Sent: Wednesday, November 30, 2005 3:59 PM
> To: Mark.Donaldson AT cexp DOT com; DDobbs AT mutualmaterials DOT com;
> Veritas-bu AT mailman.eng.auburn DOT edu
> Subject: RE: [Veritas-bu] yes sir, obviously insane (6.0 upgrade
> question)
>
>
> Actually the problem was at the OS level.  It wouldn't see tape drives
> through the bridges properly.
>
> -----Original Message-----
> From: Mark.Donaldson AT cexp DOT com [mailto:Mark.Donaldson AT cexp DOT com]
> Sent: Wednesday, November 30, 2005 2:25 PM
> To: Jeff Lightner; DDobbs AT mutualmaterials DOT com;
> Veritas-bu AT mailman.eng.auburn DOT edu
> Subject: RE: [Veritas-bu] yes sir, obviously insane (6.0 upgrade
> question)
>
> The problems were mostly related to SSO - problems discovering devices 
> &
> managing scsi reserves placed by other media servers on shared tape
> drives.
>
> Those using one-drive-one-server architectures didn't seem to have
> complaints.
>
> -M
>
> -----Original Message-----
> From: veritas-bu-admin AT mailman.eng.auburn DOT edu
> [mailto:veritas-bu-admin AT mailman.eng.auburn DOT edu]On Behalf Of Jeff
> Lightner
> Sent: Wednesday, November 30, 2005 12:00 PM
> To: Dan Dobbs; Veritas-bu AT mailman.eng.auburn DOT edu
> Subject: RE: [Veritas-bu] yes sir, obviously insane (6.0 upgrade
> question)
>
>
> I don't have answers to what you wrote but did want to caution you 
> about
> using RH for your master.  I've not been able to get RH EL AS 3 to
> properly recognize my SAN environment.   From an earlier thread on this
> list it appears I wasn't the only one.
>
> Not sure if you're running SAN or direct SCSI but if the former thought
> it best to mention it.
>
> -----Original Message-----
> From: veritas-bu-admin AT mailman.eng.auburn DOT edu
> [mailto:veritas-bu-admin AT mailman.eng.auburn DOT edu] On Behalf Of Dan Dobbs
> Sent: Wednesday, November 30, 2005 1:21 PM
> To: Veritas-bu AT mailman.eng.auburn DOT edu
> Subject: [Veritas-bu] yes sir, obviously insane (6.0 upgrade question)
>
> Greetings, list.
>
> Here's my situation. I have a HP-UX media/master server on 5.1 with a
> newer Neo SDLT bot and an ancient DLT4000 Surestore bot. We're going to
> be getting a shiny new Red Hat box soon to take over migrate backup
> services.
>
> I've just read the upgrade docs for 6.0, and I have to say that I'm a
> little overwhelmed. If I read it right, an upgrade to 6 requires all 
> the
> clients to get the security software installed and upgraded to 6 as 
> well
> before they can get backed up, something I'm not sure I can accomplish
> in a day.
>
> In short, I'm considering leaving my HP-UX box at 5.1, and building the
> new RH box on 6. My plan was to remove the Neo SDLT juke out of the 5.1
> environment, plug it into my RH box,and import tapes there (as we've
> used less than 30 SDLT's, it shouldn't be a huge deal). This would 
> allow
> me to do the icky client-side upgrades piecemeal, as I moved them to
> 6.0, they would then back up to the new box. Eventually, as all the
> clients moved over, I could then move the old jukebox and stop
> operations on the HP-UX box.
>
> The obvious problem is recovery; until I import every single tape, I
> wouldn't know what tapes had backups from the 'old' system.
>
> I guess the thrust of the matter is:
>
> 1) Is the upgrade of an in-place 5.1 server to 6 easier than it reads
> on paper?
> 1a) Do I really have to upgrade my clients to 6 before the server will
> talk to them?
> 2) If I was running 6 on the 'old' system, could I follow the previous
> advice, and move /usr/openv/netbackup/db and
> /usr/openv/volmgr/database, run the proper vmglob and bpmedia commands
> on the new box and be done?
>
> Thanks again for all your help!
>
> -dd
>
>
> _______________________________________________
> Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
>
> _______________________________________________
> Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
> _______________________________________________
> Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
>


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