1810G redundant uplinks to 2x 2910AL

[Adrian_C]

Hi,

I have 2 ProCurve 2910al-24G's with 10GBe interconnect modules (CX4), and two 1810G-24's sitting on my desk.

These will eventually be installed in our server room, with the two 2910's becoming the core switches (servers will have one NIC plugged into each with NIC teaming for redundancy).

The 1810G's will be the perimeter or client switches that workstations etc will connect to. I have seen a previous setup with the same model switches, that had a 2 port trunk defined on the 1810G's with one each of the ports uplinked to different 2910 core switches. Only one of the ports was ever active at one time. This did provide redundancy incase a 2910 core switch or the cabling failed, the 1810G would fail over to the next port connected to the other 2910 core swithc. BUT this process took about 60 seconds.

I have been experimenting with the equipment here to see if there is another way of doing this, and this is what I have found. I'm after any input on this setup, to see if it may cause any issues in the future.

I have enabled LACP Passive on 2 ports on each of my 2910 core switches (ports 23 and 24). I have also enabled Spanning Tree protocol on both 2910 core switches.

I have then created 2x 1 port trunks on each of my 1810G perimeter switches (called CoreSw1 and CoreSw2). I have one trunk port from each perimeter switch going to one of my core 2910 switches, and the other trunk port going to the 2nd 2910 core switch.

I have tested this, and it appears to be working fine. If I kill the power to one of my core 2910 switches, I lose 2 packets in a constant ping, or loss of connectivity of about 6 seconds, much less that the inital 60.

This scenario is working fine in my test environment, but wanted some others input before it goes into production to see if there may be any shortfalls, that I may discover when the switches are under load.

I have attached two images, one showing the trunk status from one of the 1810G perimeter switches, and a network diagram showing how it is setup.

Any comments are appreicated. 

[Arimo]

Hi Adrian

With this equipment your only choice is block the redundant links with Spanning Tree.

I'd like to point out one thing. "Trunk" in HP terminology refers only to link aggregation. One-port trunks serve no purpose. On the convergence time... you are using RSTP, not STP, right?

[Adrian_C]

Thanks

[Adrian_C]

Thanks for your help!

It appears I was really over complicating things with the trunks.

I had seen someone else setup redundant uplinks using the trunking before on the same hardware, but it didn't really seem right.

I have now removed all trunking from my switch configs and enabled RSTP on the two core 2910's. And this is working as I would expect.

Do you think I'll have any issues with this down the track?

If I kill the active link SRTP takes about 6 seconds to re route traffic through the other core switch. Is this adjustable at all?

Thanks for your time.

Adrian