DHCP Redundancy

[boomam]

Hi.

 

We've got a second server here setup to do AD & DNS redundancy, and ive been trying to do the same with DHCP.

After some reading around, i concluded that what i need to do was:

 

Have a range covering a certain amount.

Have server A exempt half one half of that range. And server B exempt the other half.

have the same reservations on both servers.

All sorted.

 

Problem is, that it doesnt work.

 

For example.

Server A has a range of 172.17.5.10 to 172.6.254.

Exclusion of 172.17.6.1 to 172.17.6.254

 

And

 

Server B has a range of 172.17.5.10 to 172.6.254.

Exclusion of 172.17.5.10 to 172.17.5.254

 

Both have the same reservations. gateways, ect;

 

Doesnt appear to work though.

Theoretically if i disable DHCP on Server A and renew an address, B should pick it up. Doesnt happen though.

 

Ideas?

 

Thanks in advance all. .

[Geoff]

Windows DHCP failed miserably in this regard for me also. I'm using ISC DHCPd with failover on Linux machines now instead.

 

Failover with ISC DHCP

[boomam]

Tbh, i'd prefer to get it working on windows.

Its only a temporary thing anyway, the networks getting nuked in the summer. .

[Geoff]

It took me 12 hours of not getting it to work on Windows. It took me 2 hours of setting up on *nix. Was a no brainer.

[Michael]

It is a good question - why isn't redundancy on DHCP straight forward? What about Windows Server 2008; does this support DHCP redundancy?

[strawberry]

it wont, server 2 will know that the ip address is valid and so wont attempt to renew it. i think once it's expired then server 2 will dish it out an ip from its range.

[boomam]

So when server 1 has no more addresses to hand out, server 2 will pick up the rest?

Thats a bit messed up, as theres no redundancy there?

[RabbieBurns]

It took me 12 hours of not getting it to work on Windows. It took me 2 hours of setting up on *nix. Was a no brainer.

 

Aye ive got it setup like that as well. Took me longer to figure out how to get the two servers times in sync for dhcp failvoer to work than it did to configure dhcp failover ;p

[Geoff]

NTP is one of the first things I install on new *nix servers.

[RabbieBurns]

it was the timezone I had set wrong on one of them. Had it set for just GMT rather than the london zone, so it was exactly 1 hour out.

[bossman]

@Boomam:

 

Would this be of any help:

http://www.lanarchitect.net/Articles/DHCP/index.htm

 

or this:

DHCP-Redundancy for free

[strawberry]

So when server 1 has no more addresses to hand out, server 2 will pick up the rest?

Thats a bit messed up, as theres no redundancy there?

 

nooo... Pc1 will try to renew it's ip, server2 will look at its request and dismiss it as valid. When pc1's ip expires it then looks for a new ip, which it will receive from server2.

[boomam]

@Boomam:

 

Would this be of any help:

http://www.lanarchitect.net/Articles/DHCP/index.htm

 

or this:

DHCP-Redundancy for free

The first link is basically describing what im doing, having one server serve one half of a scope and the other server serving the other half.

Not too keen on using 3rd party software to do something that afaik, should work anyway?

 

nooo... Pc1 will try to renew it's ip, server2 will look at its request and dismiss it as valid. When pc1's ip expires it then looks for a new ip, which it will receive from server2.

So even if ive switched off DHCP on server 1 it wont work?

 

Thats a bit rubbish, as surely the entire point of redundancy is to provide that facility when the 'main' part isnt functioning.

[strawberry]

kinda works for me, It'll ensure that you always have a valid working ip address and if server 1 comes backonline before the end of the lease it can pick up where it's left off.

[Diello]

We've got it working on Windows here - but instead of excluding half the range, each DHCP server is only set to serve half the range each. We find it load balances perfectly.

[boomam]

We've got it working on Windows here - but instead of excluding half the range, each DHCP server is only set to serve half the range each. We find it load balances perfectly.

Can you describe in detail how you have it setup please?

[RabbieBurns]

Well when the time came that the failover was needed - it failed!

 

Got it set up like it is in the guide : Failover with ISC DHCP

 

But today I get in and one of the DHCP servers is down. No problem, the failover should be doing the job, but it wasnt.. in the /var/log/messages was about peer holding all the leases or something. Ill find the exact message later.

 

What could I have done wrong ?

[Geoff]

Are your DHCP servers authoritative? What are your min/max lease time settings?

 

Also check your failover is ok. Is the slave acking the updates from the master server correctly?

[RabbieBurns]

I copied the confs from that howto verbatim pretty much only changing the IPs etc, which has max set at 30 mins. They seem to be talking to each other fine, I cant get on just now to copy from the logs but Ill double check when I get a chance.

[RabbieBurns]

managed to get the server back up within half an hour. It was the primary dhcp that went down. Last couple of entries in log, primary then secondary:

 

Oct  7 20:05:44 roberts dhcpd: peer dhcp-failover: disconnected
Oct  7 20:05:44 roberts dhcpd: failover peer dhcp-failover: I move from normal to communications-interrupted
Oct  7 20:05:45 roberts dhcpd: failover peer dhcp-failover: peer moves from normal to normal
Oct  7 20:05:45 roberts dhcpd: failover peer dhcp-failover: I move from communications-interrupted to normal
Oct  7 20:05:45 roberts dhcpd: balancing pool 80de608 192.168.2/24  total 25  free 12  backup 12  lts 0  max-own (+/-)2
Oct  7 20:05:45 roberts dhcpd: balanced pool 80de608 192.168.2/24  total 25  free 12  backup 12  lts 0  max-misbal 4
Oct  7 20:05:45 roberts dhcpd: Sending updates to dhcp-failover.
Oct  7 20:07:02 roberts dhcpd: Wrote 23 leases to leases file.
Oct  7 20:07:02 roberts dhcpd: failover peer dhcp-failover: I move from normal to startup
Oct  7 20:07:02 roberts dhcpd: failover peer dhcp-failover: peer moves from normal to communications-interrupted
Oct  7 20:07:02 roberts dhcpd: failover peer dhcp-failover: I move from startup to normal
Oct  7 20:07:02 roberts dhcpd: balancing pool 80de608 192.168.2/24  total 25  free 12  backup 12  lts 0  max-own (+/-)2
Oct  7 20:07:02 roberts dhcpd: balanced pool 80de608 192.168.2/24  total 25  free 12  backup 12  lts 0  max-misbal 4
Oct  7 20:07:02 roberts dhcpd: failover peer dhcp-failover: peer moves from communications-interrupted to normal

 

Oct  7 20:05:46 oasis dhcpd: Wrote 25 leases to leases file.
Oct  7 20:05:46 oasis dhcpd: failover peer dhcp-failover: I move from normal to startup
Oct  7 20:05:46 oasis dhcpd: failover peer dhcp-failover: peer moves from normal to communications-interrupted
Oct  7 20:05:46 oasis dhcpd: failover peer dhcp-failover: I move from startup to normal
Oct  7 20:05:46 oasis dhcpd: balancing pool 69f780 192.168.2/24  total 25  free 12  backup 12  lts 0  max-own (+/-)2
Oct  7 20:05:46 oasis dhcpd: balanced pool 69f780 192.168.2/24  total 25  free 12  backup 12  lts 0  max-misbal 4
Oct  7 20:05:46 oasis dhcpd: Sending updates to dhcp-failover.
Oct  7 20:05:46 oasis dhcpd: failover peer dhcp-failover: peer moves from communications-interrupted to normal
Oct  7 20:07:03 oasis dhcpd: peer dhcp-failover: disconnected
Oct  7 20:07:03 oasis dhcpd: failover peer dhcp-failover: I move from normal to communications-interrupted
Oct  7 20:07:03 oasis dhcpd: failover peer dhcp-failover: peer moves from normal to normal
Oct  7 20:07:03 oasis dhcpd: failover peer dhcp-failover: I move from communications-interrupted to normal
Oct  7 20:07:03 oasis dhcpd: balancing pool 69f780 192.168.2/24  total 25  free 12  backup 12  lts 0  max-own (+/-)2
Oct  7 20:07:03 oasis dhcpd: balanced pool 69f780 192.168.2/24  total 25  free 12  backup 12  lts 0  max-misbal 4

[soapyfish]

I am in a situation where I need DHCP redundancy I have a large range of 2032 Ip addesses avaliable this is slighly more than twice our current number of devices.

 

I was wondering about the suitablilty of having each server provide addresses to a chunk of the total range.

 

for example SVR1 would provide IP's for the first 1016 addresses this would cover the entire network at the moment. Then to have SVR2 service any remaining requests in needed.

 

This way if SVR1 fails for any reason I would still be able to provide DHCP to the clients.

[boomam]

Theres multiple ways to to do it.

Personally, although it requires more effort, i'd split it into smaller chunks.

So server 1 did the first 255, server 2 the second 255, server 1 the third 255, and so on.

[soapyfish]

Why would you up them in smaller chunks are there performace issues ?

I do have 4 servers I could stretch the service over if it is worth while.

 

Currently I do have one DHCP server providing IP's for the entire range.

[boomam]

Why would you up them in smaller chunks are there performace issues ?

I do have 4 servers I could stretch the service over if it is worth while.

 

Currently I do have one DHCP server providing IP's for the entire range.

Not as far as im aware.

Just for redundancys sake tbh.

Theres alot of schools of thought on how to do it, none are more correct than another tbh.

As long as you have some sort of redundancy, then all is good tbh. lol. .