dhcp failover

[RabbieBurns]

OUr DHCP server locked up overnight, causing all sorts of issues this morning until it was rebooted.

So Im planning on setting up a 2nd server to run DHCP as well, and just give it a seperate range to dish out.

What things do I need to consider with the likes of printers, which Ive set reservations for etc? And is there anything I should think about with regards to length of lease times?

Edit: this is a windows network

[danrhodes]

DHCP Best Practices: Dynamic Host Configuration Protocol (DHCP)

[Hightower]

A course I was on recently suggested the best practice was to set the same range on your second DHCP server to your first.

S1: 192.168.0.1 - 192.168.0.100
S2: 192.168.0.1 - 192.168.0.100

And then create a block on each server, so S1 hands out 80% of addresses and S2 hands out 20%:

S1 Block: 192.168.0.81 - 192.168.0.100
S2 Block: 192.168.0.1 - 192.168.0.80

You can still lease out 100% of your addresses, but both servers are setup so that if one falls over you remove the block and your 100% comes from the remaining server instead of a share of the two.

The instructer said there was no 'automatic' way to do this, but by setting up your servers like this the hard work is done and when you lose a server you simply remove the block which take all of 10 seconds, and then replace the block when the server comes online again.

[Hightower]

DHCP Best Practices: Dynamic Host Configuration Protocol (DHCP)

Ah, just noticed this haha - looks like what I've just tried to explain

[box_l]

Windows 2008R2 features part V: DHCP Split-scope | A-fanatic blog

for server 2008r2

BoX

[Chuckster]

Reservations should be all right as it worked for me (I had to manually re-create them on the other server).

On both servers (Win 2003) I enabled and authorised DHCP. One server had the the range, for example, 192.168.0.1 to 192.168.0.99 and the other one had 192.168.0.100 to 192.168.0.240.

The last remaining 240 onwards was reserved for switches
x.x.x.1 - x.x.x.50 is reserved for printers and servers.

[RabbieBurns]

Cheers, will have a go at this tomorrow.

[danrhodes]

Cool let us know how it turns out!

D

[RabbieBurns]

Dont currently have any 2008R2 DC's but might be an excuse to upgrade.

In fact, does DHCP need to run on a DC or can it be a service on a server all on its own?

[danrhodes]

Dont currently have any 2008R2 DC's but might be an excuse to upgrade.

In fact, does DHCP need to run on a DC or can it be a service on a server all on its own?

No it can be on any box.

You can run it independently of your DC, and even run it on Linux if you so wish

D

[RabbieBurns]

No it can be on any box.

You can run it independently of your DC, and even run it on Linux if you so wish

D

I run ISC in failover mode on 2 gentoo boxes at home, works great

But apart from my nagios box we're a MS house at work, so will just be using that.. Ill look at the 80/20 setup though tomorrow and report back.

[danrhodes]

I run ISC in failover mode on 2 gentoo boxes at home, works great

But apart from my nagios box we're a MS house at work, so will just be using that.. Ill look at the 80/20 setup though tomorrow and report back.

Good Stuff, hope it all goes well!

D

[K.C.Leblanc]

A course I was on recently suggested the best practice was to set the same range on your second DHCP server to your first.

S1: 192.168.0.1 - 192.168.0.100
S2: 192.168.0.1 - 192.168.0.100

And then create a block on each server, so S1 hands out 80% of addresses and S2 hands out 20%:

S1 Block: 192.168.0.81 - 192.168.0.100
S2 Block: 192.168.0.1 - 192.168.0.80

You can still lease out 100% of your addresses, but both servers are setup so that if one falls over you remove the block and your 100% comes from the remaining server instead of a share of the two.

The instructer said there was no 'automatic' way to do this, but by setting up your servers like this the hard work is done and when you lose a server you simply remove the block which take all of 10 seconds, and then replace the block when the server comes online again.

I'm missing something here.

If S2 goes down you're going to reconfigure S1 to give out the whole range.

Won't you get clashes where S1 leases out an I.P. that S2 leased out before it went down.

Also what happens when S1 runs out of addresses?