Carp doesn't work after the upgrade

capitangiaco · Dec 18, 2007, 12:45 PM

Hi all

I had a setup with two pfsense 1.0.1 with carp, that worked very well.
I upgraded to 1.2 rc3.

Now when both firewalls are up after 5min of work carp loose sync and from carp status I can see that few carp ips are in master mode and other in backup mode.

I have sis nics (5 for each fw)
From the firewallA console I can see alot of warnings of the watchdog timeout.
If I turn off the firewallA the B become active and all works great whitout any problem.

Turning on device polling on firewallA seems to fix the problem, and the carp system work.
From firewallA dmesg I can see this warning:

arp_rtrequest: bad gateway 159.213.. (!AF_LINK)
I have each one for each carp ip

Must I activate the device polling also in the firewall B ?

I am very confused….

Giacomo

sullrich · Dec 18, 2007, 6:19 PM

Sounds like the newer version of FreeBSD does not like one or more nics on the machine that is reporting the watchdog timeouts. If you are not using Intel nics I highly suggest switching.

capitangiaco · Dec 19, 2007, 4:55 PM

@sullrich:

Sounds like the newer version of FreeBSD does not like one or more nics on the machine that is reporting the watchdog timeouts. If you are not using Intel nics I highly suggest switching.

why device polling fix the problem ?
freebsd guys assume that it can only works in 4.x or 5.x but not >= 6 release

Giacomo

sullrich · Dec 19, 2007, 10:07 PM

Humm.. Not really sure. That is most strange.

So you are saying that enabling polling fixed it??

Waldemar · Mar 24, 2008, 8:05 PM

Hi!

I seem to have exactly the same problem. After the upgrade to 1.2 CARP stops working after approximately 10-30 minutes. I think the problem is not CARP itself though, but the network driver. I have a setup with 2 Intel (em) NICs and several VLAN interfaces. Once it fails (it always fails on the WAN interfaces only, strangely, never on the LAN ones), I cannot ping the corresponding interface IPs from firewall A to B and vice versa. I can however ping the gateways (which are in the same subnets) from both. Luckily I found this post, and device polling indeed does solve the issue. It is very strange though. It looks like it is an OS problem rather than a pfSense problem.

Cheers,
Waldemar

sullrich · Mar 24, 2008, 8:06 PM

Try reinstalling 1.2 with the SMP kernel. Does this change the situation?