Re: Changing Node & Virtual IP's for Print Server Cluster




Bazooka-Joe wrote:
I have a 2 node active-active Windows Server 2003 (SP1) cluster. The
hardware is identical in every way. The only resource it is providing
are clustered spooler resources. It's a clustered print server. The
configuration is as such:

Cluster Name: cluster1.domain.com
Cluster Node 1 Name: cluster1a.domain.com
Cluster Node 2 Name: cluster1b.domain.com
Cluster Virtual Print Server 1 Name: print1.domain.com
Cluster Virtual Print Server 2 Name: print2.domain.com

All of the IP addresses are sequential (I'm anal about that kind of
thing) and are in the same subnet. For arguments' sake, let's just
pretend those addresses are:

100.1.1.10 (physical node 1)
100.1.1.11 (physical node 2)
100.1.1.12 (cluster address)
100.1.1.13 (virtual server 1 address)
100.1.1.14 (virtual server 2 address)

And let's also pretend the subnet mask is 255.255.255.0 (24 bit) and
the gateway is 100.1.1.1.

Now, I'm server operations, and the networking team is telling me that
this cluster, in its entirety, now needs to be moved to 100.1.200.0
subnet and has given me IP address range of 100.1.200.5 through
100.1.200.9. My mask is changing too (255.255.252.0 [22 bit]) as is my
gateway (100.1.199.1). I also need to switch my physical cable
connection as I'm going to a different network switch & VLAN.

These are all made up numbers for security sake. I haven't done the
binary math to see if 199 is in the same subnet as 200 given that mask,
but let's just assume it is, since it's a made up number anyway.

MY QUESTION IS THIS:
What's the cleanest way to change the physical and virtual addresses?
Here's what I'm thinking:

1. Move all resources to Node A.
2. Take all resources offline.
3. Swap out the cable going to the public network for Node A.
4. Change IP address of Node A (verify communication with the domain).
5. Add a new IP address resource to the Cluster group (while it's
still offline).
6. Make this new IP resource the correct address/mask/gateway for the
new subnet.
7. Add a new IP address resource to the 1st Virtual server (while it's
still offline).
8. Make this new IP resource the correct address/mask/gateway for the
new subnet.
9. Add a new IP address resource to the 2nd Virtual server (while it's
still offline).
10. Change the "Network Name" resources so they all depend on the new
IP addresses for all virtual servers including the cluster itself.
11. Swap out the cable going to the public network for Node B.
12. Change IP address of Node B (verify communication w/domain).
13. Bring cluster online.
14. Bring virtual print servers online.
15. Test connectivity to all physical and virtual servers via new
IP's.
16. Cleanup DNS and/or WINS as necessary.


Obviously there's no SQL involved. I have (er...will have) solid
printmig back ups of the existing printing subsystem before I attempt
any of this too. What do you think of that sequence of events though?


:Bazooka-Joe



I just spoke with a colleague of mine who believes this would be the
better sequence:

1. Move all resources to Node A.
2. Take all resources except the quorum offline (can't really take the
quorum offline).
3. Change all the virtual IP addresses while they're offline (no new
ones, just modify existing ones).
4. Change the public network cable and the physical IP for Node B
(test connectivity).
5. Move all clustered resources to Node B (virtual servers remain
offline though)
6. Change public network cable and physical IP for Node A (test
connectivity).
7. Bring virtual servers online.
8. Test connectivity to all cluster virtual servers.

That sounds a bit easier and just as clean I'm thinking.

:Bazooka-Joe

.



Relevant Pages

  • Re: Exch 2003 SP2 - applied on one node, but cant move resources
    ... resources to Node2, the failover did not complete because 'system attendant' ... Virtual Exchange server and failover occurred normally again upon taking ... cluster resources oline. ...
    (microsoft.public.exchange.admin)
  • Re: [Xen-devel] Re: [ANNOUNCE] Xen high-performance x86 virtualization
    ... >> that each virtual machine only gets the resources that its sponsor ... This allows companies providing virtual servers to ... Ensim's 'private virtual server' product is the most complete ... even in the buffer cache. ...
    (Linux-Kernel)
  • Re: clustering exchange + dhcp + file share
    ... on a Windows 2000 cluster. ... owners for all resources. ... This way whatever server comes up first takes on ... > being the preferred owner for exchange groups so they are the active ones ...
    (microsoft.public.windows.server.clustering)
  • Re: file server clustering
    ... During the failover I get a message 'Delayed Write Failed'. ... You will have to declare all shares (cluster resources in this case, ... > I am building a clustered server that only provide drive access... ...
    (microsoft.public.windows.server.clustering)
  • Re: file server clustering
    ... You will have to declare all shares (cluster resources in this case, ... I am building a clustered server that only provide drive access... ... I have setup the basic cluster and setup 2 disk resources (other than the ...
    (microsoft.public.windows.server.clustering)