DHCP server - State Switchover Interval, Maximum Client Lead Time? What options do I need?

Eduard Martinenko 101 Reputation points
2020-08-20T11:09:29.917+00:00

I have relatively a small network with two domain controllers (windows server 2012 and windows server 2016)
I want to use a shared pool option with something like 90% - 10 % (So there is the main server and the second reserved one).
The vast majority of the addresses in my network are reserved, and
The main goal to ensure stable work even if the first servers goes down, the clients shouldn't notice anything.

Is it wise to set "Maximum Client Lead Time" to a very high number - something like 90 hours or more (so if the server goes down on the weekends - I will be in blissful ignorance). Are there any caveats to it? (Considering that the will be no new devices, and the network will stay the same).
What about "State Switchover Interval" does it even mean something if the server won't be giving out new addresses.
Could someone please elaborate on this?

Windows Server 2016
Windows Server 2016
A Microsoft server operating system that supports enterprise-level management, data storage, applications, and communications.
1,764 questions
Windows Server 2012
Windows Server 2012
A Microsoft server operating system that supports enterprise-level management, data storage, applications, and communications.
1,128 questions
Windows DHCP
Windows DHCP
Windows: A family of Microsoft operating systems that run across personal computers, tablets, laptops, phones, internet of things devices, self-contained mixed reality headsets, large collaboration screens, and other devices.DHCP: Dynamic Host Configuration Protocol (DHCP). A communications protocol that lets network administrators manage centrally and automate the assignment of Internet Protocol (IP) addresses in an organization's network.
925 questions
No comments
{count} votes

Accepted answer
  1. Candy Luo 12,451 Reputation points
    2020-08-21T08:53:12.95+00:00

    Hi ,

    >>Just suddenly server1 goes down. What will happen then?

    When DHCP1 is approached by Client 1 to lease it an address, instead of leasing an address for 8 days it leases an address to Client 1 for 10 hours (MCLT). DHCP1 then informs DHCP2 that Client 1 has leased an address. DHCP2 can add the details of Client 1 to its DHCP database.

    During a failure of DHCP1, DHCP1 is approached by Client 1 to lease it an address.DHCP1 leases an address for 10 hours (based on the MCLT). When Client 1 attempts to contact DHCP1 after 10 hours. (At 50% of MCLT Client1 will send a unicast to DHCP1 trying to renew the lease. DHCP1 will not respond, since it is down. At 87% of MCLT, the Client will do a broadcast to renew its lease.)it will fail and subsequently send out a general message for any DHCP server to respond. This will be registered by DHCP2, which will recognize that Client 1 has an address that was provided by its failed partner using the MCLT. This will cause DHCP2 to lease Client 1 the same address (192.168.1.44)for the full 8 days and add the details to its database even though it hasn't taken over running the entire scope yet.

    It will actually be 11 hours (Maximum Client Lead Time - 10 hours + Auto State Switchover Interval - 1 hour) after going into the Communication Interrupted state before DHCP2 takes over running the entire scope.

    Best Regards,

    Candy


3 additional answers

Sort by: Most helpful
  1. Candy Luo 12,451 Reputation points
    2020-08-21T03:06:03.797+00:00

    Hi ,

    Maximum Client Lead Time: Specifies the amount of time for which a DHCP lease may be renewed by either failover peer without contacting the other. It also specifies the amount of time that either DHCP server will wait in a “partner down” state before assuming control of the entire IP address range within the scope. ( default = 1 hour ).

    Auto State Switchover Interval: When selected, specifies the amount of time that elapses before a DHCP Server is automatically transitioned to a “partner down” state when network communication is interrupted to a DHCP Server. If this option is unchecked, an administrator must manually transition the status of a DHCP Server into a “partner down” state using the DHCP Management console or PowerShell. ( when checked, the default = 60 minutes )

    The short explanation for MCLT is that having a too-small value will cause performance problems in normal operation (but will indeed buy you a quicker recovery if one of the pair fails). Conversely, having a value that is too large will improve performance but means that you have a longer delay at failover. And this is more detail on why that is:

    https://tools.ietf.org/html/draft-ietf-dhc-failover-12#section-5.15

    You can also refer to the following article:

    https://www.itprotoday.com/windows-78/dhcp-failover-windows-server-2012

    Please Note: Since the web site is not hosted by Microsoft, the link may change without notice. Microsoft does not guarantee the accuracy of this information.

    ---Please Accept as answer if the reply is helpful---

    Best Regards,

    Candy

    No comments

  2. Eduard Martinenko 101 Reputation points
    2020-08-21T07:19:48.88+00:00

    I still don't have an understanding of how exactly "Maximum Client Lead Time" works.
    I have even done some testing by in my environment everything just works which is suspicious and I wan to have a better understanding before implement this on the production server.

    The first question - am I right that "Maximum Client Lead Time" is completely irrelevant if the IP address is reserved for this particular mac address. So the client PC will get the same reserved IP address from either of the servers anyway? Even if one of them is down and "Maximum Client Lead Time" is expired?

    To give my questions some context I will explain what my test environment is.
    I have two servers in "load balance mode" let's say 50 - 50. (I suppose they divide the zone of authority like - 192.168.1.1 -> 192.168.1.128; 192.168.1.129 -> 192.168.1.256). let's name them server1 and server2 respectively.
    Maximum Client Lead Time - 10 hours.
    Auto State Switchover Interval - 1 hour.
    The global maximum lease time is 8 days. (it is in the scope's options).

    My client (PC1) gets an IP address from DHCP server1 - 192.168.1.44 (lease time is 8 days).
    Just suddenly server1 goes down. What will happen then?

    Am I right that if the PC1 stays online for 8 days straight, the IP address will stay the 192.168.1.44 for all this time? And after this, it will try to automatically renew the IP address from the server2. And it will get an address in the range that starts from 192.168.1.129.

    What will happen if the PC1 is restarted just after the server1 has gone down? I suppose it will try to renew its address, in this case, it must try to reach server2. What address will it get? Will it continue to use 192.168.1.44 (that has been given by server1) for 10 hours (Maximum Client Lead Time). Or will it use it for 8 days? What will happen after "Auto State Switchover Interval" has expired - does it mean that the server2 will start to give out scope form 192.168.1.129 only for new clients or if the PC1 released the previous address?
    What will change after "Maximum Client Lead Time" has been expired? Will it no longer try to prolong the IP address lease after 8 hours? Or will it try to prolong all addresses for 8 days?

    The short explanation for MCLT is that having a too-small value will cause performance problems in normal operation (but will indeed buy you a quicker recovery if one of the pair fails).

    Why is that? How is MCLT can cause performance problems? I thought (maybe wrongly) that "Auto State Switchover Interval" can cause performance problems because it will send a lot of messages to ask the second server about its status.

    And how the low MCLT value could help to recover faster? Does it only apply to newly added devices? So if the client already was given an address before server1 had crashed, it really won't matter?

    Is there a big difference in these options for “Hot Standby” and ”Load balance“ modes?

    No comments

  3. Candy Luo 12,451 Reputation points
    2020-08-24T08:34:03.793+00:00

    Yes. Client is basically impossible to detect downtime.

    However, as I said before, if MCLT has a too-small value will cause performance problems (but will indeed buy you a quicker recovery if one of the pair fails). But if MCLT has a value that is too large will improve performance but means that you have a longer delay at failover.

    >>Why is that? How is MCLT can cause performance problems?

    The downside of a short MCLT value is that the initial lease interval that will be offered to every new DHCP client will be short, which will cause increased traffic as those clients will need to send in their first renew in a half of a short MCLT time.

    This is why it affects performance.

    Hope this can help you understand better.

    Best Regards,

    Candy

    No comments