DHCP and ESRP

[UpTheCity]

We have 2 Black Diamonds, one live the other resilient.
All edge switches and stacks have a live connection to the
live BD and a resilent connection to the resilent BD.
All connected VLANS are running ESRP.
The BDs are interconnected using a point-to-point VLAN.ESRP
is not enabled on this VLAN
This setup appeared to be working as would be expected,ie
if the live link of an edge switch fails the VLAN on the live BD goes into standby and the BD resilient becomes the master and traffic continues to be passed
However we recently discovered a problem. A live link to a user area was lost and the resilient connection kicked in.
All mail, file and print operations continued to work without a problem.
If , however, the ip address of a workstation was released and then renewed the renewal would fail. As soon as the live link was reconnected the release\renew process worked fine.
Any ideas on this would be much appreciated.

Thanks.

 

[ourrob]

I've seen this myself on ESRP, and vrrp, i think its one of two things

dhcp relies on the address of the router it passes through to understand which scope/pool to assign the address out of, its possible its getting confused over the router address, usually you have to specify the router address for each scope/pool.

Alternatively, if you've got a couple of resilient paths, which are equal cost, you can find the traffic from the dhcp request being sent one way, and returning a different way. ( this usually happens if you have a pair of resilient links that are downstream ). One fix, is to cost one link slightly higher than the other

 

[UpTheCity]

ourrob,

Thanks for the reply. I think your first suggestion is more relevant. In this case the request will go thru the Resilient BD, then through the Live BD and then to the core edge switch which the DHCP server is connected to. When the reply is sent back I'm suspecting it's not getting passed through to the Resilient BD but see's the originating router's address ,as it's directly connected, on the Live BD and drops it as the VLAN is down.

Shouldn't the Live BD be smart enough to know it
needs to switch the reply over the point-to-pointy Vlan to the Resilient BD?

 

[UpTheCity]

Believe we discovered what the problem was.
The reply from the DHCP server was getting confused as
the live BD and the resilient BD have the same logical MAC address when ESRP is enabled on the VLAN. The reply appeared to be dropped at the live BD instead of been passed through to the resilient BD.
To rectify the problem we configured bootprelay on the edge switches, which connected to both the live and resilient
BDs, and added the required DHCP server addresses.
This time when the live connection was pulled the DHCP request presented the MAC and IP address of edge switch, connected to the resilient BD, and so the reply got back safely.

If there is an Extreme guru out there is thinks I shouldn't
have needed to do this please step forward.