Does anyone have a work around for this issue. I've got a 3750 connected to an hp 2810 switch in a port-channel setup.
on cisco side
int range gix/x/x -x
channel-group 1 mode active
on the hp side
trunk 47-48 trk1 LACP
Ive been getting drops on the HP side with no errors on the Cisco 3750.
It appears as though the link is working but still getting drops and these equal retransmisssions?
Does anyone have any ideas or known workarounds?
What type of drops are you seeing? Input queue drops, output queue, frame errors etc?
I don't think this would have much to do with the etherchannel itself, as you are using LACP. As long as its formed, its really pretty simple. I don't see many issues with this itself.
Also how many drops are you getting....do they increment at a certain rate?
I'm getting several hundred thousand drops, slowing incrementing as traffic passes.
These are as Rx Drops on the HP switch. The drops increment about 1 per 5-10 seconds during normal traffic periods.
My next step is to force both side of the trunk to 1gig full.
Can you paste output of :
show int po <>
show run int po <>
show int <>
show run int <>
(Of the physical ports that belong to the port-channel)
show etherchannel <> detail
Is there any kind of flow control done on Hp switches ?
What happens if you make HP as LACP active and Cisco switch as LACP passive?
this is the default
I think you have hit the nail on the head as the hp uses src-dst ip pairs for load balancing!
I did more research and found that the HP only does sa/da ip load balancing and the cisco had the default setup of src-mac.
I will change the cisco to sa-da-ip load balance method tonight and let you know if that fixed it.