HPE Aruba Networking & ProVision-based
1832791 Members
3038 Online
110045 Solutions
New Discussion

Re: stp priority vs cost

 
davidbiel1
New Member

stp priority vs cost

Need some help with stp config,

 

i have two sites (1 prod, 1 dev) set up stretched layer 2 with an ATT gigaman because of nonroutable iscsi traffic. We have a new vdi project that started last week and have this traffic at our prod site going to a solid state san. I noticed the iscsi traffic flow is going to both sites across the gigman and causing latency to the san. It should be staying local to prod site, even local to the same switch and passing right to the san.

 

The STP config that was set up here has the main core switch as root and the main switch across the gigaman at the dev site as the next lowest priority and then back across the gigman to core switch 2 and then back across again and so on between 4 core 6600s. If this traffic is on the same vlan (ISCSI traffic from VDI esx host to san) could it still be being sent across the gigaman because of stp priority? I checked everything else and the proper ports are being blocked and the host and virtual connect that the host connects to are all configured correctly. this switch priority is the only thing i can see that might force it across the gigaman. And of course it cant be default gateways or anything right? because iscsi doesnt route. Thanks for the input.

2 REPLIES 2
Richard Brodie_1
Honored Contributor

Re: stp priority vs cost

iSCSI runs over TCP, so should route fine; maybe you're thinking of FCoE.

 

Is the traffic from the production site going both ways over the WAN link? If so, that's a routing problem for sure.

paulgear
Esteemed Contributor

Re: stp priority vs cost

Richard is absolutely right about iSCSI - it is fully routable - but if you're running a stretched L2 segment over your gigaman, then routing is not likely to be the issue.

I'm not sure what the best practice recommendation is, but i can't think of a reason why you would want to alternate your STP priorities across sites.  If your central site is connected to your remote site via a single logical L2 link, then it shouldn't make much difference.

 

If you can post your configs and perhaps a topology diagram, we might be able to make more sense of it.

Regards,
Paul