none
DHCP Failover - Migration from 2008 to 2012 RRS feed

  • Question

  • The implementation and testing of this I'm fairly happy with, I've run through it a few times while testing different scenarios. The point that I'm not entirely happy with is what I see when we flip over to the 2012 servers and a client tries to renew its lease.

    Let me try and describe what I see.

        *Before Migration*
        client1 10.45.145.194
        client2 10.45.145.193
        Scope
        10.45.145.193 - 10.45.145.252
        1hr lease
        IP Helper
        all dhcp servers (2008, Svr1, Svr2)
        Scope only active on 2008

    I then then use an export/import to setup the scopes on Svr1 (this time I'm not importing back the leases). I ensure that the server options are correct and setup the failover relationship. I check Svr2 is correct as well.

        Failover Load balanced
        50/50
        5min MCLT
        State Normal

    I then deactivate the scope on 2008 and ensure its active on Svr1/Svr2.  Now I wait and let the clients try and renew.

    What I see is that they are succesfully able to renew. The clients have kept their original IP addresses and the expiry time has been lengthened as expected. I also see from ipconfig/all that client1 is using Svr2 as its DHCP Server and client2 is using Srv1. Now this the bit I don't like, because it seems to me that if client1 is using Svr2, then to keep the scope statistics correct, it should have an IP address from the Svr2 pool, so thats 10.45.145.224 and beyond. Now I understand this is due to the hashing algorithim that is applied to the client MAC to determine which DHCP server should be servicing which client.

    If I view the Scope Stats, I see

        *Scope Stats*
        60       Total addresses    
        2  (3%)  In use             
        58 (96%) Available          
        29 (48%) Addresses Available (in this Servers Pool)
        29 (48%) Addresses Available (partner pool)
        2  (3%)  Addresses granted (this Servers pool)
        0  (0%)  Addresses granted (partner pool)

    See, that doesn't match up. This Server is using up 2 IPs, but has 29 available.

    If I release / renew on the client, it picks up the same IP address. However, if I release/reboot then the client renews with an ip address from the second half of the pool (partner), 10.45.145.224.

    Perhaps this is all by design and I'm missing something fundamental, but I'm wondering if I had 30 clients, would they have all taken up IPs from this Servers pool when some of them actually some of those clients belong to the other DHCP server. Is this this something I should be worried about? Maybe after time things would sort themselves out? Not what I would expect, and not what I have seen in practice.  I don't really want to be rebooting all my clients (different scopes for VDI, Phones, Thin clients etc.) and forcing them to Rebind. Btw, when I exported/imported the existing leases I saw the same behaviour.

    Tuesday, November 29, 2016 8:36 AM

All replies