It still seems to me that the behavior
of the non-lease-granting failover peer,
when responding to a renew request,
when it has (presumably) no knowledge
of a new lease, is not optimal.
What seems to be happening is that
the failure peer receives a renew request
before it has been updated by its peer
about a new lease grant.
It responds by ack'ing back the full lease
time of 24 hours, rather than the MCLT
time of 1 hour.
Also, although its response to the initial
discover is delayed by 3 seconds,
as its peer is selected by the HBA,
its ack of the renew is not delayed.
And since its network latency is less
than that of the lease-granting peer,
its lease-extending ack gets back
to the local switch before the MCLT - x
ack of the lease-granting peer -- and
causes a mismatch between the switch
and the attached client, resulting in
a network disconnect after one hour.
I'm just saying.
Mark
Sent from my iPhone
_______________________________________________
dhcp-users mailing list
[hidden email]
https://lists.isc.org/mailman/listinfo/dhcp-users
Locked
