Networker

[Networker] Backing up ISCSI LUNS

2009-09-15 13:46:15
Subject: [Networker] Backing up ISCSI LUNS
From: dfields <networker-forum AT BACKUPCENTRAL DOT COM>
To: NETWORKER AT LISTSERV.TEMPLE DOT EDU
Date: Tue, 15 Sep 2009 13:42:14 -0400
FYI, turns out that Legato was doing the right thing after all .  I ran some 
controlled tests and it behaves just like it does for NFS mounts, and doesn't 
back up mount points if they are under another ISCSI mount.

Turns out what was happening was just a very confusing coincidence.  I had two 
that had the directory structure I described earlier, and both of them had 
failures on the main /u00 ISCSI mountpoint/LUN.  The combined size of the 
retries and good backup added up to almost exactly the same size as /u00 + all 
of the other LUN's which were mounted below /u00.  Given that I was trying to 
debug this around 5:00 am after three nights of watching backups all night, I 
read the data wrong and came to the wrong conclusion.  

Turns out that at least the biggest underlying problem I had was that one of my 
nics that I have teamed together on my backup server was having problems, and 
that kept causing my backups to fail.  None of this was evident in any event 
logs, and I only found out by watching my network switch throw up some errors 
on that connection while the backups were running.  I disabled that NIC, and so 
far my backups have been more stable.

My other issues all stemmed from the fact we've outgrown the Network edition of 
Networker, and just having a single backup server that can handle the backup 
databases and writing to the tape library at the same time.  We added a SAN 
earlier this year, and the amount of data and the number of servers we have to 
back up has grown to the point where I need to add at least one storage node, 
maybe two.  I budgeted for an upgrade on Networker, and for an expanded tape 
library and at least one storage node.  For now I'm working around it by 
splitting up backups so that I can run smaller sets of servers over the weekend 
full such that I don't overload the backup server with any one set.

I'm not quite sure how to size/judge the number of storage nodes we need 
though, as we've never needed any in the past.  I'm trying to find some guides, 
so if anyone has any suggestions/rule of thumb, I would gladly appreciate the 
help.

Thanks,

Dave





Preston de Guise wrote:
> 
> > 
> > > 
> > > > During backups, the client is backing up the server with a saveset  
> > > > of All, which works ok in that it backs up /u00, /u00/work, and / 
> > > > u00/staging as different save sets, which is what you would expect  
> > > > it to do.  But what it's not doing is skipping the "work" and  
> > > > "staging" directories when it's backing up the /u00 save set.  I  
> > > > think it's because the client doesn't know  that the work and  
> > > > staging directories are ISCSI mounts.  If these were NFS mount  
> > > > points, then they would be skipped as part of the backup of /u00.
> > > > 
> > > > I ran some reports after the backup finished, and the backup for / 
> > > > u00 was much larger than it should have been.
> > > > 
> > > > I'm going to open up a ticket with EMC once I do some more testing  
> > > > and I can give them more information.
> > > > 
> > > Opening a case with EMC on this would be a waste of time because  
> > > NetWorker is behaving as intended and as we'd expect.
> > > iSCSI is effectively meant to be presented as local storage, in the  
> > > same way that a SAN/FC presented LUN should appear as local storage.
> > > As such, NetWorker is doing the right thing - it wants to back it up.
> > > To work around it - e.g., if you want to backup the LUN via the  
> > > RedHat host or in some other fashion, you need to setup directives  
> > > to skip/ignore those filesystems.
> > > 
> > 
> > I'm going to disagree with Preston.  Either, NetWorker should  
> > backup /u00 and not see /u00/work and /u00/staging as separate  
> > savesets, or it should see them as separate savesets and not backup / 
> > u00/work and /u00/staging as part of /u00 as well as separately.
> > 
> > You should definitely open an issue with EMC -- one piece or the  
> > other is wrong.
> > 
> 
> Ah, my apologies to Dave. Francis, you're entirely right - and this  
> will serve me right to read and respond to the mailing list _before_  
> I've had my morning coffee and put my glasses on. I misread what was  
> going wrong, and completely agree - if the filesystems are appearing  
> as 3 separate units, but NetWorker is traversing the mountpoints to  
> backup (and you're not using save -x), then it's definitely a problem.
> 
> Cheers,
> 
> Preston.
> 
> --
> Preston de Guise
> 
> 
> "Enterprise Systems Backup and Recovery: A Corporate Insurance Policy":
> 
> http://www.amazon.com/Enterprise-Systems-Backup-Recovery-Corporate/dp/1420076396
> 
> http://www.enterprisesystemsbackup.com
> 
> NetWorker blog: http://nsrd.wordpress.com
> 
> 
> 
> via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER


+----------------------------------------------------------------------
|This was sent by david.fields AT acs-inc DOT com via Backup Central.
|Forward SPAM to abuse AT backupcentral DOT com.
+----------------------------------------------------------------------

To sign off this list, send email to listserv AT listserv.temple DOT edu and 
type "signoff networker" in the body of the email. Please write to 
networker-request AT listserv.temple DOT edu if you have any problems with this 
list. You can access the archives at 
http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER

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