#Converting NX9336C-FX2 Cluster Switch to Cluster/Storage Switch

1 messages · Page 1 of 1 (latest)

vagrant imp
#

We are running two 10-Node Clustes, each with 6x A300 and 4x A400 Nodes using NX9336C-FX2 Cluster Switches (Cluster-HA-Breakout RCF 1.8)
The A300's are connected using 10G Breakouts on ports 1, 2 and 3. The A400's are connected with 100G to Ports 7-10. The single NS224 Shelf per A400 HA Pair is directly connected to the Nodes.

We were wondering, if it is supported / possible to add a second NS224 Shelf to the A400's using the NX9336C as both Cluster and Storage Switches.
There is a RCF (NX9336C-FX2-RCF-v1.10-Shared) but it would have to be modified:

Original:

  • Ports 1-8: 40/100GbE Intra-Cluster/HA Ports, int e1/1-8
  • Port 9: 10GbE breakout Intra-Cluster Ports, int e1/9/1-4
  • Ports 10: 25GbE breakout Intra-Cluster/HA Ports, int e1/10/1-4
  • Ports 11-22: First HA-pair Controller and Shelf Storage Ports, int e1/11-22
  • Ports 23-34: Second HA-pair Controller and Shelf Storage Ports, int e1/23-34
  • Ports 35-36: Intra-Cluster ISL Ports, int e1/35-36

Modified:

  • Ports 1-6: 40/100GbE Intra-Cluster/HA Ports, int e1/1-6
  • Ports 7-10: 10GbE breakout Intra-Cluster Ports, int e1/7-10/1-4
  • Ports 11-22: First HA-pair Controller and Shelf Storage Ports, int e1/11-22
  • Ports 23-34: Second HA-pair Controller and Shelf Storage Ports, int e1/23-34
  • Ports 35-36: Intra-Cluster ISL Ports, int e1/35-36

Second step would be to attach the existing NS224 as well as the A400's Storage Ports to the switches.

What do you think? Good idea or very bad idea? Alternatively we could buy X1148A NICs for all A400's and re-cable the existing shelfs from e0c/e0d to e0c/e5b. This should be supported non disruptiv, right?

Regards,
Hiro

blissful elbow
#

I would bet I could make it work. My thoughts:
Disable auto revert on cluster lifs.
Migrate all ports to one switch
On the other switch, wipe it clean and reboot. Do setup. Apply new RCF.

Now, reattach cables per new RCF. you can do a show cdp neighbors to make sure all the nodes connect.

I would then revert the cluster lifs home. Verify. Then migrate all the cluster lifs to switch 2.

Repeat process on switch 1 with wipe/setup/rcf.

#

Oh, and when complete, make sure to modify the lifs to auto revert again

blissful elbow
#

Then follow the documented procedure to convert a node with attached storage to switched storage