Web and Unmanaged
1752765 Members
4970 Online
108789 Solutions
New Discussion

Who Me Too'd this topic

T_1_6
Regular Advisor

1810-24G LACP/Trunking problem

Hi,

I hope someone can help.

I have two ESXi boxes each with an NC360T running MPIO with round robin to load balance accross the two NICs.
These are connected to a server via an 1810-24G running an iSCSI target which has an NC365T Quad Nic in it which is trunked to the switch.

Now, back in August this was working great with an LACP team setup with the Network config utility so that all four links from the ESXi hosts would load balance when writing (incoming) to the Iscsi Target.

With a reinstall of the iSCSI server, and a firmware upgrade to the 1810 switch, it now only EVER uses three out of the four links when uploading to the Quad NIC.

When I pull out cables and reconnect them, the switch briefly load balaces acrross all 4 NICS as it should, but after a few seconds goes back to just using three out of the 4 links.

 

 Now, when I pull the four cables in the trunk and connect them to a configured 3com switch, it load balances the incoming traffic accross ALL four links perfectly. 

Each Esxi host having 2 x 1gb connections uploads to two links on the iSCSI target. When two hosts are writing to the target, all four links are used as you would expect.

With the HP switch, one host will upload using two links, but when you upload with the other host, it will only ever use one link.

Does anyone have any idea why the 1810-24G will NOT work correctly for Recieve load balancing?!


Its driving me INSANE, as it used to work.

Now, if I disable the trunk completely and just have the links teamed, it will load balance accross all four, but the speeds are terrible, and there is no redundancy like with an LACP trunk.

LACP works as such, but ONLY with three out  of four links being used for uploads/writes to the server.

 

I did go back to the 1.17 firmware, but this made no difference.

 

Who Me Too'd this topic