MS Cluster over multiple c7000 chassis

[Justin Hannan]

Does anyone have information on how best to create a MS cluster over two chassis?

I have Node1 in chassis 1 and Node2 in chassis 2. In each chassis I have 4 VC-Enet modules. As the MS documentation for creating a cluster says a direct cross-over cable should be used to create the heartbeat network, this doesn't quite work for VC-modules...

Am I going to have to create a direct ethernet link to each VC-Enet module on each chassis and create a non-shared link for each hearbeat network?

Thanks

[Bruno J. Melo]

I don't know about VC-Enet, haven't worked with them yet. But with the GbE2c Layer 2/3 Ethernet Blade Switches you can create the heartbeat network using a separate VLAN.

Then you should connect the switches of both enclosures through the use of LACP.

[Graeme Chambers]

Hi Justin,

I'm trying to do the same thing but thinking of linking the VC modules across chassis and creating a shared ethernet network between the nodes. Apparently with the new firmware versions you can link the modules but I'm not sure how it would work in practice.

Just wondering how you solved the problem?

Cheers

Graeme

[Justin Hannan]

Hi Graeme

Apparently you can use the uplink/downlink chassis connection to create a virtual connect domain and setup a heartbeat that way, but my kit went live before I heard about that in the later versions of the VC.

I've got two MS clusters (SQL & Exchange 2007) and one ISA 2006 array. So three heartbeats in all.

I have a 2Gb trunk into each VCe module which I then tag the VLANS in the virtual connect manager. Each of the heartbeat networks is it's own VLAN, so I then have just created three virtual connect networks from these VLANs and assigned them to a specific NIC in a blade profile.

Not ideal as these are still not server/server connections, but the dedicated vlan per heartbeat network is as good as it's going to get for me at the moment. I've had the kit running for a year now and I've not had any problems with the heartbeat networks causing the clusters to failover.

Hope that helps.

Justin

[Graeme Chambers]

Thanks Justin, I've ended up using a similar method but using the X2 ports on the VC modules to link the two chassis together.

I then created two ethernet networks on VC and created two private networks on each host. I've got two clusters (five nodes in total) sharing the two cables but with 20 gig of available bandwidth it really shouldn't be a problem.

I've subnetted all of the private networks separately to ensure they don't get confused.

Thanks for your help.

Cheers

Graeme