Cluster fails when private network is offline

[Wasser]

We have 3 networks on our cluster. One for the heartbeat, one for the LAN and one {Private) for the communication between the 2 Servers. They have all been set up with different subnets and are all using the mixed network. The Private network uses a switch which has 4 network cables connected. One of the tests is to remove power to the switch, however when this is done the cluster fails (Server Ip Address). The Network priority is Heartbeat first, Private second and the LAN last. When the switch is powered on the Network priority has changed with the Client second and the Private Network last. Clustering will not allow me to change from the Mixed Network. Does anyone have any tips on what Networks to use and to stop the Network priority changing. Thanks in advance for any help.

 

[mpmartin68]

sounds like the Cluster Name has a dependancy upon the Private TCP/IP address. Try changing it to your public TCP/IP address and give your power test a whirl.

 

[Wasser]

There was a dependancy on the Cluster Name for the Private Network. There is also a dependancy on the Public Network. Which one should be in the dependancy list as the Private Network connects two servers together with 4 connections into a switch. We want to keep Network traffic to a minimum when the Web Server is running queries against the Database Server ( these are two seperate CL380 Cluster Boxes). My understanding is that if you set the Network priority and one of the Network fails it should use the next available network in the Network priority list.

 

[mpmartin68]

If your clients are accessing the server via the public network, that is where I would place the dependancy. Classify your private addresses with no dependancies and assign the to your cluster groups (depending if you are active-active or active-passive. If you initiate a failure, all of your resources (private IP's included should fail over to the other node.

As a general practice, I like to set the "allow failback" section to "on" so that the nodes failback to their normal state once all nodes are back up in the event of a failure.

 

[Wasser]

The configuration is active/passive.

If I include the private network in the Cluster Group and failover to the passive node - wouldn't it fail as the private network could not be started on the passive node as the switch failure means that none of the private network ip addresses are available.

I have run a test today, and taking the private network out of the Cluster Group has worked. But how can I verify that it is using the Private Network for Web Server to Database communication when all networks are available?.

Also, Do you know how difficult/easy is it to add MSMQ and TimeService to the Cluster Groups?. I do not think it is as easy as just installing MSMQ from the Win2000 Adv Serv disc.

Thankyou for your time, you have saved me alot of time and I am very appreciative.

 

[mpmartin68]

Sorry I've been underground, I have a lot of projects coming due soon. They definitly made money on me this week...LOL

To find which network your web to DB traffic is using could be done by disconnecting your public network. Sort of getting a sniffer on it, I have no other ideas. That is weird about the private network taken out of the Cluster Group deal. I'll have to chew on that one for a while...

As for the MSMQ/NTP issue, I personally have not done it, but it can be done. MSMQ would be a part of your IIS or SQL install and NTP can be set as a Generic Service. Go to the MS Knowledge base and use "MSMQ cluster" as your keywords and you will get a whole slew of hits.

As far as the help, no problem...anytime. 10% of this job is knowing stuff off the top of your head and the other 90% is knowing where to find everything else!!

Let me know how it goes and I will get back to you when I think of something for the private network...MPM