Failover DHCP server support

sullrich · Nov 4, 2005, 12:48 AM

So, who's using this option? And how well does it work for you?

billm · Nov 5, 2005, 4:46 AM

So, who's using this option? And how well does it work for you?

I was…I never had a machine down long enough to actually test it well though, but I did continue getting DHCP, so it's working.

--Bill

Britton.Kerr · Nov 10, 2005, 9:12 AM

Use it everyday. But I bug you guys enough about it on the ticket system…..
I plan to have eventually 500 devices running behind...I can't afford any downtime

With the continuing issue from ticket 627 about the hostnames would it not be easier to have the the DHCP lease tables be sync'd from the master to the backup like pfsync but not have the backup give out leases if it is not the carp master? I mean I've set my leases to expire like every 10 hours or something so its not that life or death usually when a failover happens. I could live with the time it would take the carp process activating the listener service and then shutting down after the master comes back up.

Am I making any sense? What ever machine is master at the time pushes to the other box.

But let me say that if we can ever get the hostname problem fixed and static leases sync'd over I'll be content.....

sullrich · Nov 10, 2005, 4:15 PM

Have you tested the latest fix that I listed in the ticket yet?

Britton.Kerr · Nov 10, 2005, 6:58 PM

Yes.. and I attached a screenshot. Both are now after another update running 1.14 of the file. I don't understand the sync process enough to help. Cause one lease file has the info in it and the other doesn't…. I thought whatever was in the area where the lease is was supposed to be sent to the other box... which file has the logic for the lease syncing?

sullrich · Nov 10, 2005, 7:00 PM

@Britton.Kerr:

Yes.. and I attached a screenshot. Both are now after another update running 1.14 of the file. I don't understand the sync process enough to help. Cause one lease file has the info in it and the other doesn't…. I thought whatever was in the area where the lease is was supposed to be sent to the other box... which file has the logic for the lease syncing?

Syncing of DHCP information is handled strictly by DHCPD. I'm not sure if this can easily be fixed without hacking DHCPD, which I don't plan on doing.

billm · Nov 11, 2005, 4:45 AM

@sullrich:

@Britton.Kerr:

Yes.. and I attached a screenshot. Both are now after another update running 1.14 of the file. I don't understand the sync process enough to help. Cause one lease file has the info in it and the other doesn't…. I thought whatever was in the area where the lease is was supposed to be sent to the other box... which file has the logic for the lease syncing?

Syncing of DHCP information is handled strictly by DHCPD. I'm not sure if this can easily be fixed without hacking DHCPD, which I don't plan on doing.

Ditto…I never could replicate this issue, but I probably don't have near the hosts needed and haven't rebooted enough. Once all my gear is back online I'll poke some more - this is obviously not a show stopper bug, but more of a nuisance than anything. If I remember correctly, you did delete the lease files on both computers so the leases files should be clean (the best bet though is to bounce both firewalls simultaneously). I know you were an early adopter of this and had some wierdness in the lease files from back then.

--Bill

Britton.Kerr · Nov 15, 2005, 4:24 PM

I have deleted lease files several times and even experimented with deny duplicates and both end up with duplicates so I am just giving up and going static as much as possible to avoid confusion…..

JanZ · Jan 13, 2006, 2:10 PM

@sullrich:

So, who's using this option? And how well does it work for you?

Hello.

I have 2 pfsense nodes (1.0BETA1), they are using CARP on WAN and on LAN interface (sucessfully).

I tried to configure DHCP failover, but with no success. In web configuration I enabled DHCP on LAN interface and in Failover server entered real IP of the other server.

So I did on the other server.

As thing refused to work, I had a lok into /var/etc/dhcpd.conf and surprisingly found out, that both nodes tries to be master server (which is obviously wrong).

I edited manually dhcpd.conf on host pf2 (put it in secondary mode), like described in http://www.lithodyne.net/docs/dhcp/dhcp-4.html and things got better.

Anything, that I'm doing wrong? Why hosts are not aware of each other? CARP is working well…

Thnx, /jan

sullrich · Jan 13, 2006, 5:32 PM

You used the REAL IP instead of the CARP IP?

JanZ · Jan 16, 2006, 11:48 AM

@sullrich:

You used the REAL IP instead of the CARP IP?

Sorry, thing got better… CARP was working fine, but synchronisation config was broken...

Now it is OK.

/jan