Cluster group and group 0

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

skyline - 24 Jul 2006 17:56 GMT

This is a SQL2005 active/passive cluster.

We noticed that doing a manual failover on Group0 works fine (group0 has all
the SQL volumes, service, and name). There is also a cluster group, which
has the Cluster IP, Name, Q:\ volume (quorum), and the MSDTC. When we
manually fail Group0, the seperate entity Cluster Group does not failover
along with it. So the resources each are responsible for are mixed between
the new active server and the current passive server that was active before
the failover.

We have a SQL2000 cluster that fails both groups when a failover is
initiated on just Group0.

Is there a way to tie them together? Should they be tied together?

Also can anyone tell me if there is a significance between having both a SQL
server virtual IP and a cluster virtual IP? This new SQL2005 cluster has
this, but our SQL2000 cluster only has a single IP.

Thanks!

Reply to this Message

Geoff N. Hiten - 24 Jul 2006 21:57 GMT

The cluster group is exactly what the name implies. It is the cluster
resource that manages the rest of the cluster resource groups. Technically
it exists so the node that owns that resources breaks all ties in resource
arbitration. As a practical matter, it is unrelated to the normal operation
of the SQL group, however it must be operating or in a valid transition
state for all the other cluster resources to run. I usually leave it on an
unoccupied node, just so I know where it is. You most definitely do not
want to tie this to the SQL resource group. You won't gain anything and you
may break something important. A true failure will likely result in both
groups moving independently to the remaining node, even though they are not
tied together.

BTW, I rename my cluster groups to match the SQL instance name. This really
helps on multi-instance clusters.

If your SQL 2000 cluster has only one group and one IP address, someone
built it wrong. It should have separate groups for the cluster and for the
SQL instance.

Signature

Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP

> This is a SQL2005 active/passive cluster.
>
[quoted text clipped - 20 lines]
>
> Thanks!

Reply to this Message

skyline - 25 Jul 2006 14:26 GMT

The only thing that confuses me is the fact that the MSDTC and the drive that
the Quorum resides on are in the cluster group, while all the SQL resources
are in the other group. When I want to fail over to work on one of the
servers, then common sense says I need to fail both groups over so that I can
reboot the passive one after making changes. Only moving the SQL application
will leave my quorum and MSDTC with a server that is getting
patched/updated/etc.

With our SQL2000 cluster, there are 2 seperate groups just as on the 2005
cluster. The only difference is when I force a failover on Group0, both
Group0 AND the cluster group moves at the same time. That way the cluster
group is not stranded on a server that will be under construction. The 2000
cluster has the cluster resources in cluster group and the SQL resources in
Group0 just as the 2005 cluster has.

Is there a particular reason why 2 IP's are a good thing? Is it so that you
can communicate to the cluster via TCP/IP without having to interact with the
SQL virtual IP and create more traffic? Keep in mind, these 2 IP's are both
outward facing and 1 responds to the SQL virtual IP and the other responds to
the Windows cluster virtual IP. I am not talking about the private network
that connects the servers together for heartbeat as one of the IP's.

> The cluster group is exactly what the name implies. It is the cluster
> resource that manages the rest of the cluster resource groups. Technically
[quoted text clipped - 39 lines]
> >
> > Thanks!

Reply to this Message

Geoff N. Hiten - 25 Jul 2006 22:19 GMT

Odd. Your SQL 2000 system should not move both groups when only one is
force failed. Unless you are failing the entire node rather than the group.
If you take a node offline, all resource groups hosted on it will fail over.

As for your SQL 2005 cluster, if the node being upgraded goes offline during
the upgrade, the group will failover anyway. It is still a good idea to
move all groups to the other node before performing an OS upgrade or patch.
You can use the cluster.exe command in a script to force move them together.

As for the two IPs, that is just an artifact of Windows Clustering. Each
resource group with a disk, IP address, and Network name becoms a virtual
server. You use the cluster name to connect to the Cluster itself and the
SQL Virtual Server name to connect to the SQL instance. If you add new
instances, they get their own virtual servers.

Signature

Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP

> The only thing that confuses me is the fact that the MSDTC and the drive
> that
[quoted text clipped - 85 lines]
>> >
>> > Thanks!

Reply to this Message

Russ Kaufmann [MVP] - 26 Jul 2006 18:19 GMT

> The only thing that confuses me is the fact that the MSDTC and the drive
> that
> the Quorum resides on are in the cluster group, while all the SQL
> resources
> are in the other group.

So? This is no big deal.

> When I want to fail over to work on one of the
> servers, then common sense says I need to fail both groups over so that I
> can
> reboot the passive one after making changes.

Yes, that makes perfect sense.

> With our SQL2000 cluster, there are 2 seperate groups just as on the 2005
> cluster. The only difference is when I force a failover on Group0, both
[quoted text clipped - 4 lines]
> in
> Group0 just as the 2005 cluster has.

This is not a good idea. Let's face one simple fact, it takes a matter of
seconds to manually move the default cluster group and the MSDTC contained
in it. Don't sweat the small stuff like this tiny bit of administration.

> Is there a particular reason why 2 IP's are a good thing?

It is a required thing... One IP to be assigned to the cluster name and one
for the virtual server hosting SQL.

Signature

Russ Kaufmann
MVP - Windows Server - Clustering
ClusterHelp.com, a Microsoft Certified Gold Partner
Web http://www.clusterhelp.com
Blog http://msmvps.com/clusterhelp

Reply to this Message