HP 1810: LACP trunks only working if odd number of ports active

[aminski]

Hello,

I've configured two LACP trunks on an 1810 24Gv2, one using 4 ports connecting an NexentaStor (Solaris) system, and one using two ports connecting an Citrix Xenserver 6.5 (OpenVSwitch). I've tried do ping the storage unit from the XenServer and vice versa  without success. This is strange, we have a bunch of switches (except some edge switches they are all switches from Dell/Broadcom), storage units and XenServers runnning here, running a lot of LACP trunks, and we never had any problem. Well, I've tried a different switch, an old 2650, the same behavior, storage is not reachable. I've tried some tests. I've configured 4 ports for each trunk, and I got them working, but only if an odd number of ports is enabled for each trunk. Ping is working if I enable one port, is not working if I enable the second one, is working if the third port is enabled, and is not working after the fourth port is enabled. This is true for both trunks.

What happens here?

Thanks.

P.S. This thread has been moved from ProCurve / ProVision-Based to Web and Unmanaged. - Hp Forum Moderator

[Vince-Whirlwind]

It probably means the LACP hashing algorthim is broken.

Try to manually configure it as something other than the default.

[aminski]

Okay, I've switched from "load balancing based on IP and port of source and destination" to "load balancing based on source MAC address" on XenServer site, doesn't change anything.

[Vince-Whirlwind]

Seeing as you can replicate the behaviour on either trunk, it sounds more like the switch is at fault.

Are you sure the trunks are actually up?
Are the VLANs all available on all ports that are members of the trunk?

How much have you spent on storage and servers, compared with the cost of one of HPs most bottom-end switches? Maybe a better switch?