Carp strange behavior

[Damien Fleuriot]

On 18 Mar 2013, at 22:22, Rafael Ganascim <rganascim at gmail.com> wrote:

> Hi list,
> 
> I have multiple FreeBSD firewalls with carp working well. I have no problem
> and the vast majority of firewalls works perfectly.
> 
> But now, I'm with problems with a simple firewall cluster with carp that
> the state randomly goes to MASTER and randomly returns to BACKUP.
> 
> Looking to the L1/L2 tests, I have no rx/tx erros, buffers miss, in/out
> drops , etc. The physical conection between the firewalls looks good.
> 
> Monitoring the interfaces/buffers/mbufs/virtual memory with netstat,
> vmstat.... no errors was found.
> 
> Using tcpdump, I can see that in the exact moment of the state change, the
> currently master's firewall stop sending multicasts to the 224.0.0.18
> during some seconds and the state change occurs.
> 
> The system:
> # uname -a
> FreeBSD fw-cj-01 9.1-RELEASE FreeBSD 9.1-RELEASE #0: Thu Feb 28 13:18:41
> BRT 2013     root at fw-new-01:/usr/obj/usr/src/sys/DEDICr9v1CoreX64  amd64
> 
> 
> Now, how can I debug why carp stops to send multicast packets?
> 

Lots of things to be said here.

First, how do you know carp stops sending packets ?
Might not be the case.

Second, triple check that the VHID is not already used somewhere else.

Third, any firewalling in place ?
If so, disable it, check for better results.

Fourth, netstat -m -p carp

Fifth, raise advbase on both boxes and see if that helps.

Sixth, what's the frequency of these role swaps ?

Seventh, what do you get in dmesg ?