Problems with SQL Server following IP address change

[rpreston]

All,

I have a 2-node passive/active cluster running on Win2000. This cluster has been relocated and I therefore need to change the ip addresses of the two servers, the cluster, and the virtual SQL server.

I've now done this, using Q228480 to change the cluster address, and Q244980 to change the virtual SQL server address. If I look at the two resources in Cluster Administrator they now show the correct addresses.

The two servers are talking to the rest of the network and the DCs happily, and I can ping the cluster addresses as well.

All fine so far, except for when I come to start the cluster, everything comes up except for the SQL Server resource. This generates two events, a 1145 (Cluster Resource SQL Server timed out) followed by a 1069 (Cluster Resource SQL Server failed).

When I look at the cluster log the following two errors seem to be behind the problem :-


ODBC sqldriverconnect failed
sqlstate=08001
native error=11
message=[Microsoft][ODBC SQL Server Driver][TCP/IP Sockets]SQL Server does not exist or access denied

ODBC sqldriverconnect failed
sqlstate=01000
native error=274d
message=[Microsoft][ODBC SQL Server Driver][TCP/IP Sockets]Connectionopen (Connect())


Any help/ideas/suggestions would be greatly appreciated.

TIA


Rob

 

[CarlMartin]

Did you ever find a solution to this problem?

 

[rpreston]

Yes. I re-installed the cluster !

I talked to a couple of other people with MSCS cluster backgrounds who told me that experience had taught them that any time they needed to reconfigure the cluster, it was safer to just re-install it.

Fortunately, the amount of data we have on our cluster is relatively small at the moment, so we were able to pause our live system and copy the SQL databases on the cluster off to a server with a large drive array. We then ran SQL from the one server while we rebuilt the cluster, and we then copied the SQL database back to the clustered array.

One thing to be aware of when re-installing the cluster. It operates in a different way from Win2K SP3 onwards. I had major problems trying to re-install until I found this one. The easiest thing for me was to just go with a Win2K SP2 until the cluster had been rebuilt, and then update the OS.

Good luck !!



Rob

 

[luisrato]

rpreston, tell me something.
Does your private and public networks belong to private Network Address Ranges (like class A 10.x.x.x;class B 172.16.x.x;class C 192.168.x.x)?
Does your private and public network belong to the same network address class?

If you configure your cluster with the public and private network in the same Network address class, you can experience a lot of problems.

In this example, imagine that you configure your public NIC with the following address - IP 10.50.10.0 / Mask 255.0.0.0 / Gateway 10.50.10.254, and your private NIC with the following address - IP 10.0.1.0 / Mask 255.255.255.0. They are both class A, but in a different Subnet, and it was not suposed to got any problem, right?

Wrong! With this configuration, and with this subnets, it is not suposed that you have any problem. But in fact you may have. If you configure the cluster public and private NICs in the same Network Address Class, sometimes they confuse the Network subnets and may loop the network connections and your clusters services.

To avoid this type of problems, configure your private NIC with an IP address from a different Network Address Class from the Public NIC.
If your public NIC has the Address 10.50.10.1, configure the private NIC with the IP 192.168.0.1. The the same in both Cluster Nodes.

Luís Rato
MCP;MCSA;MCSA Messaging;MCSE;CIWA;CIWP;CIWSA;MCIWA

 

[rpreston]

Luis,

That wasn't the case with our configuration. The public and private addresses were in different classes.

Thanks for the info.



Rob