cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1126
Views
15
Helpful
10
Replies

What is the order of DHCP server selection with multiple relays used?

I have an interesting problem with one of my sites that they call me about every 1.5 months that they "cannot connect"

I have deemed this to be an issue with getting DHCP leases, but the consistency and debugging has be a bit confused. I wiped out a bunch of non standard DHCP relay addresses that we no longer use, but this brings up the question of what does the router use for the DHCP server as it sends out its unicast to the DHCP relay agents?

Is the DHCP server that is the first to respond that has an active pool whom dishes out the lease? Or does the selection and assignment process happen in some other fashion in the background?

I am using Infoblox as my IPAM with DHCP servers on grid members in InfoBlox if that helps at all. I am sure it is related, but understanding the flow of who knows how to answer when setup in multiples will help me troubleshoot this forward. 

10 Replies 10

If all DHCP in same subnet then you can use DHCP farm IP instead of single DHCP IP. 
I will double check to find Lab for you, If not find I will run small lab

marce1000
VIP
VIP

 

 - We  use a single relay per subnet , (the infoblox) , avoids all confusing . let alone having conflicts when pools are 'addressed' and or addresses would no longer be unique if different pools can hand out addresses.

 M.



-- Each morning when I wake up and look into the mirror I always say ' Why am I so brilliant ? '
    When the mirror will then always repond to me with ' The only thing that exceeds your brilliance is your beauty! '

The DHCP servers are not all in the same subnet. They are across multiple hosts and multiple DC's for failover. I typically try to use 2 for failover purposes, but still need to figure how how the DCHP server decides which will enter the query. I guess that is more on the server end and not the router end that has the relays?

After thinking about this more, I think maybe this is better to ask Infoblox support on how the software determines which grid member will answer the lease requests?

One Q, 
if both serve same subnet 
i.e. same subnet 
then Server Two can give same IP to client that other Clients get before from Server One 
you need to divide the subnet between two Server. 

They do not duplicate leases when in failover mode. I have done split scopes already in other IPAM tools, but currently we are using in failover mode and load balancing based on certain criteria and response times. 

So I guess now that I dug deeper, I kind of answered my own question on how to determine which responds, which puts me back to the drawing board trying to figure out why this 1 subnet out of tens of thousands gives me a problem every 1.5 months

I run small lab 
the Relay Router try first with first Server IP add under the interface, 
if client not get IP from DHCP the relay Router try second server IP. 

So can you double check and ping both Server using the interface IP as source of ping. 
I think the relay can not connect to one server 

I tried pings when this happens. The router can reach the relays fine. 

I look at the DHCP servers and see it dishing out the leases, but the end devices not getting them. I turn on DHCP server debugs and can see the requests going out, but no response with lease coming back. 

The more I think about this, it smells like a rogue DHCP server maybe getting turned on once every 1.5 months. It isn't handing out leases though as they get APIPA addresses and nothing valid. But my logic is, if something local has DHCP services turned on, that would be the first to respond and take precidence over a remote dhcp relay?

When you check ping you must use as source ip of interface you config ip helper under it.

I did source the interface. The connectivity to the servers is there. 

May be this is not full answer You but I think it help you to detect the issue with your network, 
during My study of ISE, the ISE must get copy of DHCP packet, so we add IP add of ISE with ip helper, 
what that meaning that the Relay send copy of DHCP to all IP list in helper IP command. 
hhhhhkkk.png
AND to make me so sure, I do small lab R3,R6,R9,R7,R8 all have dhcp pool mhm and all exclude all IP in pool.
Now the only active server is R4 
R2 which is Relay must try server one by one 
BUT NO 
relay send dhcp copy to all Server 
R2 send dhcp packet to all router R3-R9
and only R4 reply and this reply is forward from relay to client. 

opopopop.png
so check the DHCP server, by
check each DHCP server if you see one DHCP server not have free IP then it OK 
if DHCP server have no lease IP then this Server must be deep check why it not give IP to client. 
the issue may come from relay IP add to dhcp request or Op add from relay or SW to dhcp request make DHCP refuse assign IP to client. 

Review Cisco Networking for a $25 gift card