Hello all,
we have a fairly simple topology with two 6509 as core switches acting respectively as STP root for odd/even vlans. RSTP is running through our campus and our switches two connections, one to each individual 6509.
Recently we found out that when STP converges for a cluster interconnect vlan, the clustering of the nodes is broken, apparently because it is extremely sensitive (even though the cluster timers have been changed, so as to be more robust in case of topology changes). Browsing through the internet I came across a lot of vendor articles stating it is best to disable STP for cluster interconnect vlans.
Has any one any thoughts on this?
The cluster nodes are vms on different ESXi servers, which aren't connected to the same physical switch. Spanning-tree portfast trunk is enabled on all ports connecting to ESXi.
If I were to disable STP for that specific vlan, do I need to do it on all devices where vlan x (cluster interconnect vlan) exists? I mean, do I need to disable it on my core switches and all switches where ESXi servers exist, or only on the leaf switches? I am really afraid that no matter what, a loop will occur.
Has anyone gone through and disabled STP?
The other thing I was thinking of, was cross-connecting the switches, where this is possible, allowing only the cluster interconnect vlan, through the connection, and removing it from the uplinks to the core/distribution layer. This way the switch/switches with the cluster nodes, would be immune to topology changes (on vlan x). But what happens when the switch cross-connections is not applicable?
Thank you in advance,
Katerina
