Networker

Re: [Networker] Windows 2000 cluster backups not working as expected.

2003-09-15 04:19:50
Subject: Re: [Networker] Windows 2000 cluster backups not working as expected.
From: "Renty, Bart" <bart.renty AT HP DOT COM>
To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
Date: Mon, 15 Sep 2003 10:19:32 +0200
I've seen this when there is no "Physical Disk" resource type assigned
to a shared disk in cluster admin.
If you look at the virtual groups in cluster admin, do you see a line
concerning these disks E: and Q: , containing "Physical Disk" as
resource type ?

Bart

-----Original Message-----
From: Stuart Lamble [mailto:networker AT CAROUSEL.ITS.MONASH.EDU DOT AU] 
Sent: donderdag 11 september 2003 8:54
To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
Subject: [Networker] Windows 2000 cluster backups not working as
expected.


Greetings, folks. We have a little bit of a problem here, and I'm
utterly stumped. First, the system details: we have two W2k clusters,
configured essentially identically (one is used for QAT, the other, for
production. Or they will be, when the software they were bought to run
goes live.) Both clusters are in their own Active Directory domain, in a
distinct, invalid (from an Internet point of view) domain name. In both
cases, the only members of the domain are the physical nodes in the
cluster. All systems have been patched to at least service pack three
(and generally have a number of other hotfixes applied over and above
that.)

Each physical node has two drives local to itself -- drives C: and S:.
Each cluster has two shared drives -- drives E: (holding SQL Server
data) and Q: (the cluster quorum). The nature of Windows clustering
being what it is, these latter two drives are both held by only one
physical host at a time (although not necessarily the same physical
host).

For the sake of discussion, the two physical nodes are called
"physical1.its.monash.edu" and "physical2.its.monash.edu" in the
Internet-visible DNS; within the domain, they know each other as
"physical1.virtual.local" and "physical2.virtual.local". Each cluster
defines two virtual hosts. The first is associated with the cluster
quorum, and is named "virtual.its.monash.edu" in the Internet-visible
DNS. The Windows Cluster name in this case is "VIRTUAL-CLUSTER", and the
Active Directory DNS knows it as "virtual-cluster.virtual.local".

The second virtual host is associated with the SQL Server that is
running on these systems, and is named "virtual-sql.its.monash.edu.au",
"VIRTUAL-SQL", and "virtual-sql.virtual.local" (in the same order as
above.)

The Networker servers are running either Networker 6.1.2 Power Edition
on Solaris 2.7, or Networker 7.0 on Solaris 9 (why we have two Networker
servers is a long story that isn't relevant to the matter at hand.) Each
Networker server is responsible for backing up one of these clusters.

The clients we have defined in Networker are:

  physical1.its.monash.edu (aliases "physical1",
      "physical1.its.monash.edu.au", "physical1.its.monash.edu", and
      "physical1.virtual.local")
  physical2.its.monash.edu (aliases "physical2",
      "physical2.its.monash.edu.au", "physical2.its.monash.edu", and
      "physical2.virtual.local")
  virtual-sql.its.monash.edu (aliases "virtual-sql",
      "virtual-sql.its.monash.edu", and "virtual-sql.its.monash.edu.au")
  virtual.its.monash.edu (aliases "virtual", "virtual.its.monash.edu",
      "virtual.its.monash.edu.au", and "virtual-cluster")

Now, the problem itself (having defined the environment, hopefully with
enough detail):

With this setup, my expectation is that each physical client will backup
its drive C:, S:, and other, node-specific data (SYSTEM DB:, SYSTEM
FILES:, and SYSTEM STATE:, as Networker defines them). The virtual
clients would back up their appropriate shared drive (E: in the case of
virtual-sql, Q: in the case of virtual).

What I'm _actually_ seeing is Networker backing up the shared drives as
part of the backup of the physical node.

In other words, if virtual and virtual-sql happen to currently be hosted
by physical1, I will see physical2 backup drives C: and S: (as
expected), and physical1 backup drives C:, E:, Q:, and S:. The virtual
hosts are both failing with the error, "nothing to save".

Does anybody have any suggestions? I've tried changing the virtual hosts
to point to "VIRTUAL-CLUSTER" and "VIRTUAL-SQL" (in the hope that it was
a string comparison on the client failing somehow), creating an entry in
/etc/hosts for these addresses, but with no success. As somebody who has
very little hands-on experience with clustering in general, let alone
Windows clustering, I'm not sure how best to solve this one.

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

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

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