3 x P4300 Node Cluster with 2x 2910al Switch.-> Switch Redundancy problems ?

[Korlean]

Hi all, i'm having an issue with our Lefthand cluster setup

We have 3 * P4300 nodes setup in a RAID 10+1 setup. The network cards on the Nodes are ALB Bonded and each NIC goes to a different switch.

Flow Control is enabled on all ports (verified in the CMC) , Jumbo Frames are also enabled.

There also is a 2port Trunk between the two switches (in "Trunk" mode (not LCAP).

On the server side we have 2 NIC's configured with the HP DSM MPIO (in round robin mode) , the nics are also connected with the 2 switches. So NIC1 goes to switch 1 and NIC2 goes to switch 2.

So here is the issue : i wanted to test switch reduncancy , so when one switch goes down the other one picks up the work.

I test this by copying a big file from the SAN and watch the network transfers. Both NICS are working at around 60~70%

When i unplug one switch both NIC transfers stop, and after about a minute one NIC starts copying again... So in a way this is redundant, but for a minute the SAN is not reachable from the the server. When the second switch it back online the second network starts copying again without intervention.

So, my question is if anyone knows why this downtime happens (about a 50 ~~80 seconds) and if there is a way to get this to go down, a couple of seconds would be acceptable but 40+seconds is too long.

Does anyone have any suggestions ?

[David_Tocker]

Spanning tree? Could take a while to reform?

Is 40 seconds that bad for a 'failover' (aka worst case) scenario?

I would try killing STP first and see what happens...

You are pointing the servers at the VIP? (MPIO enabled?)