Hi
I am running a failover pair
default-lease-time 43200;
dynamic-bootp-lease-length 43200;
failover peer "failover" {
max-response-delay 20;
max-unacked-updates 10;
load balance max seconds 3;
mclt 300;
split 128;
}
I have a client that will repeatedly do a DHCPDISCOVER at 50% lease time (rather than a renew), even when it already has an address
in 4-3-6 this is how the flow goes :
===
* client does discover
* server offers lease of 5 mins
after 2.5 mins :
* client does discover
* server pings previous IP it handed out to client
It gets a response (client is trying to do discover as a way of renewing lease and has not released the old address)
* server marks previous address as abandoned and ignores client
* Client does discover
* server offers new address with lease time of 5 mins (MCLT)
Then we loop round and start again
===
I happened to have an old device at the back of a cupboard with 4-1-1-p1 on it so just tried against that
The result is the same, except it does not mark leases as abandoned in the leases file so they get back into pool quicker
And also on the 4th DHCPDISCOVER the client does it gets a 12 hour lease for the same address offered on 3rd DHCPDISCOVER.
Client packets are the same against both servers, client never sends it's address in the DHCPDISOCVER packet
I know this client is behaving badly but I am not sure what he expected dhcpd behaviour is, should it just keep
letting the client eat through addresses or should it treat it like a renew as the old version does (eventually on 4th go) ?
I know the server will start to use the addresses that were abandoned when it runs out of new leases but
would rather they didn't all get eaten up in the first place. I am approaching the device vendor too but I
don't hold out much hope there.
Thanks
Dave