Trunking on HP Procurve 2510G-24

[smoothrivers]

I have 2 HP PC 2510G-24 that I am using strickly for communications from VM Hosts to VM storage units.

Ports 1-20 on both switches are configured for VLAN VMiScsi (10.110.110.xxx.
Port 23 is on the default VLan and plugged into my main network.
(172.xxx.xxx.xxx)
POrt 24 is linking the two switches togeter.
I have a trunk (not LACP) set up on ports 20-21, and the trunk is a member of VMiSCSI VLAN on both switches, to provide some redundancy.

I'm sure I missed something, because when ever I plug in ports 20-21 on switch 1 into ports 20-21 on switch 2, I get packet storms and all connections drop.

What did I miss?

Thanks in advance.

 

[cajuntank]

Ok, a couple of things. You say you have ports 1-20 configured in VLAN VMiSCSI, but then you say you have a trunk group created using ports 20-21... If your goal is to have ports 20-21 in a trunk group and using that trunk group to connect to the other switch:
Then you need to make sure both 20 and 21 are untagged in VLAN 1,
Create your trunk group first using both ports (20 and 21) before you assign that trunk into a VLAN,
Tag (not untag) that new Trunk Group you created (i.e... Trunk 1) into VLAN VMiSCSI,
Rinse and repeat on other switch.*

*Note- you will need to make sure you create the same VLANs on both switches using the same VLAN ID so things get tagged correctly.

 

[cajuntank]

Also, forgot to mention, you can turn on Spanning-Tree to ensure you don't create "loops" which is what's causing your broadcast issue since it appears your trunks weren't created correctly, so for example:

spanning-tree
spanning-tree force-version RSTP-operation

But if you create your trunk correctly, then you would not have had that issue even without STP turned on.
STP would put the redundant port into a blocked state pending an issue with the other port while putting multiple ports in a trunk allows all of the ports to operate for redundancy and additional throughput.


On another note, you didn't mention about what iSCSI storage you were connecting to, so check your storage documentation as normally best practice is to enable flow-control on those iSCSI connected ports. That switch also supports jumbo frames, so that too might be of benefit to you.

 

[jkupski]

If I'm, reading this right (the bit about redundancy stands out) you're trying to split the trunk across both switches. Correct?

If so, this won't work. The 2520 doesn't support distributed trunking (which is a fairly new feature for procurve, available only on the yl/zl switches), and it doesn't support real stacking, so splitting a trunk just isn't possible.

Remove the ports from the trunk, and let the host manage the failover and you should be fine.

 

[VinceWhirlwind]

No, I think he's saying his trunk involves ports 20 & 21 on switch1 connected ports 20 & 21 on switch2.

He'd forgotten to enable spanning-tree. I'm sure he's fixed it by now.