Greetings and thanks for the replies. I'll try and answer all of them
here.
1. The systems are not in a cluster.
2. The tape should have been written to by the media server that is
trying to do the restore. The tape drives are in a SAN and the servers
all have a dedicated hba for the SAN. The master server can see all
drives in each library and the media servers are zoned so that they only
see drives in one of the two libraries.
3. It does not appear that the media server is trying to do a network
restore, but since it never actually loads a tape, I can't be certain.
On other restores it has shown up in the device manager with the proper
media server name for the server using that drive and media.
4. Since these are nfs only server (no user logins, no databases, etc)
we only put in 4 gig of swap with 4 gig of physical ram. The
maxswapchunks variable is set for 8192, so I should be able to add
additional swap once I round up some more drives.
I was hoping the memory was for the master server since it will be
cheaper and easier to upgrade that than to upgrade 6 servers and
possibly the master.
I'll do some experimenting with adding some additional swap to the media
server and see what happens.
Thanks,
Jeff
________________________________
From: Eric Ljungblad [mailto:Eric.Ljungblad at CopleyPress.com]
Sent: Monday, February 05, 2007 9:21 PM
To: Jeff Lightner; CLEVERLEY,JEFF; veritas-bu at mailman.eng.auburn.edu
Subject: RE: [Veritas-bu] Allocation failed error
Is this a cluster with alternating nodes?
-----Original Message-----
From: veritas-bu-bounces at mailman.eng.auburn.edu on behalf of Jeff
Lightner
Sent: Mon 2/5/2007 1:29 PM
To: Jeff Cleverley; veritas-bu at mailman.eng.auburn.edu
Subject: Re: [Veritas-bu] Allocation failed error
Left out: HP-UX has kernel parameters (swapchunks, maxswchunks...)
that control how large swap can get. If you exceed the amount of swap
these kernel parameters allow for it will just allocate the amount of
swap it can and won't give much complaint. You may want to check to
see if you have already allocated enough swap but aren't able to use it
all due to the above parameters.
-----Original Message-----
From: Jeff Lightner
Sent: Monday, February 05, 2007 4:25 PM
To: 'Jeff Cleverley'; veritas-bu at mailman.eng.auburn.edu
Subject: RE: [Veritas-bu] Allocation failed error
I believe the memory is for the media server itself assuming the tape
mounted there. The master would presumably only need enough memory to
issue commands - it is the actual data handling of the tape restore
eating up memory.
Was the tape originally backed up on the media server you're trying to
restore to? If not you may unintentionally be doing a network restore
where it is actually mounting on the original server and restoring from
there via network. To get around this you set the
FORCE_RESTORE_MEDIA_SERVER option in bp.conf e.g.
FORCE_RESTORE_MEDIA_SERVER = <original media/master> <target
media/master>. You then have to run bpadm g (global) m (modify) d
(notify request daemon) to reread the bp.conf - this is done on the
master server.
As to swap - my rule of thumb has always been 2 x Physical Memory. With
11.x theoretically you don't need that much but I always do it for the
virtual memory allowances it makes.
We do this kind of cross server restore in our HP-UX environment on a
disgustingly regular basis mainly for database refreshes.
-----Original Message-----
From: veritas-bu-bounces at mailman.eng.auburn.edu
[mailto:veritas-bu-bounces at mailman.eng.auburn.edu] On Behalf Of Jeff
Cleverley
Sent: Monday, February 05, 2007 3:56 PM
To: veritas-bu at mailman.eng.auburn.edu
Subject: [Veritas-bu] Allocation failed error
Greetings,
I'm running 5.1 mp4 in a hpux environment. I have a master and 6 SAN
media server systems. When I tried to do a restore of a large file
system on a media server this past weekend, the job failed with status
code 10. The master server has the catalog database and the restore was
being performed from a media server to a new file system on the same
media server. The book states that status 10 means you need to add more
memory or swap.
In a case like this, do I need to add the memory to the master or the
media server? Is there some type of formula that tells how much
memory/swap will be needed? This one file system in particular has over
8 million inodes in use. We have several others like it.
I don't remember what the settings were, but we did make sure that we
had the recommended tuning parameters for Veritas. The hardware on both
servers is hpux 11.11, rp5470, 4 gig of ram.
Thanks,
Jeff
_______________________________________________
Veritas-bu maillist - Veritas-bu at mailman.eng.auburn.edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
_______________________________________________
Veritas-bu maillist - Veritas-bu at mailman.eng.auburn.edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://mailman.eng.auburn.edu/pipermail/veritas-bu/attachments/20070206/e6df578b/attachment-0001.html
|
