Hooray! I got it working.
Listening to music over VPN, through RDP, you only hear the sound cut out for maybe 0.5 seconds after a failover now.

Was just an issue with the way I was testing my Public IP's offline. (Spoofing them internally + double nat lol)

@pete-s
It works.
You can choose to monitor a carp-interface, in which case haproxy will only run on the node that is 'master'.. Otherwise it will run on both, and perform health-checks from both.
If you have stick-tables you want to keep synced youl need to have it running on both nodes, and the sync configuration could be a bit tricky (not really supported in the webgui though with some manual advanced texts there is almost nothing thats impossible ;) ..),

Anyhow if you are only requiring that haproxy runs with the same config on both nodes that should be easy.. Just make sure to enable the 'config sync' checkbox only on the master node.

p.s. in any case, all active connections will 'break'.. and need to be re-established when a failover happens.. it wont transfer the tcp-connection-states

@jimp @bbrendon

hello guys,

i have this issue !! do you have any suggestions ??
Exactly how you describe it... in the middle of the night , random hours, can happen also with low traffic and no traffic.

How do you fix that ?

Interesting,

00:00:00.294065 IP xx.xx.xx.19 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 0, authtype none, intvl 5s, length 36
00:00:00.715554 IP xx.xx.xx..19 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 0, authtype none, intvl 1s, length 36

If i change the frequency, looks like the deamon stays alive an starts a new one and i see the old and new times, so maybe a prio0 stays broadcasting also during a failover?

00:00:00.574783 IP xx.xx.xx..19 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 240, authtype none, intvl 1s, length 36
00:00:01.965796 IP xx.xx.xx..19 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 240, authtype none, intvl 1s, length 36
00:00:01.363018 IP xx.xx.xx..19 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 0, authtype none, intvl 1s, length 36
00:00:00.583826 IP xx.xx.xx..19 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 240, authtype none, intvl 1s, length 36
00:00:00.283089 IP xx.xx.xx..19 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 0, authtype none, intvl 1s, length 36
00:00:00.845702 IP xx.xx.xx..19 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 0, authtype none, intvl 1s, length 36
00:00:00.827872 IP xx.xx.xx..19 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 240, authtype none, intvl 1s, length 36

Thats what its doing, so the backup is not responding... (this is when maintenance mode is clicked)

It survived also the CARP Maintenance and the upgrade of both units, without the Port Forward NAT.
The only issue is that in this way OpenVPN Client (to a VPN Service) binded to WAN interface will start on both nodes because both will have connectivity. Solution is to bind to a real CARP VIP like LAN and it correctly starts only on the node where LAN is MASTER.

admin please close this thread.

minimum 3 IP's for CARB.

Thanks everyone for support.

I have just realized that I had the mtu set to 9000 everywhere except on the routers. Setting it to 9000 on the routers solved the problem...

I'd get an ISP that is willing to do things right. Just sayin'.

I'd pcap and see exactly what's happening. Maybe they have something silly like an inability to ARP for more than X IP addresses per MAC address or something.

It is almost certainly NOT the pfSense software.

@jimp said in DHCP Server wrong IP CARP:
On the primary, add the secondary interface address in each active DHCP server tab.

Yup, did that! Thanks both of you - I didn't realize it was supposed to be that way, and although I've read the HA docs a number of times I guess I missed the part where DHCP server address would show as the actual IP of the box, not the VIP.

Thank you!
Cheers
Tiwing

OK forget it. I was messing around and disabled, then re-enabled guest interface on both primary and secondary and problem is now gone. I can't explain why, but ..... woohooo!

mods, you can close or outright delete this thread!

cheers

We are having L3 switches and I can't find anything about that... I'm thinking more and more that our pfSense appliance is having a hard time with the traffic!

Thank you, I'm going to look at all this documentation and I'll come back to tell you the solution to the problem or more specific questions for more specific help.

CARP is designed for a router failing or router interface going down, so the IP addresses will switch to the backup router. The CARP IP on WAN isn't going to work on both ISPs without some sort of SD-WAN arrangement.

@PiBa said in HAProxy forwarding to NGINX Seafile:

How to do it, basically still follow the instructions and don't do what do you don't need.?

If you don't want to offload, leave the ssl checkbox on the frontend 'off' and choose for mode 'ssl/https'..
As your then not using offloading, also leave the 'Encrypt-SSL' checkbox on the backend server 'off', (but do check the SSL-Checks checkbox..).

Should be pretty easy.. give it a try, and if it doesn't work show the config as you have made, and tell what the stats page looks like, is the server green and if not what does lastchk column say?.
I'm not inclined to write a step-by-step guide tailored to a specific user. As that would be more work for me, and less of a learning experience for you.. Really.. if it doesn't work first time you can try again for no additional fee

Thank you with your help I got it done. Unfortunately I don't understand most of it because it is completely new territory for me. here it is difficult to say what you need and what you don't. the side can be reached from the outside and everything should go :) THANKS again

It might be your switch doing it and not the firewall, check for and disable things like multicast storm control to rule that out.

Also you could set advbase higher on the VIPs so that it takes longer to trigger a failover. If you increase advbase to 1 that would take 1 second + skew to switch. Or use QoS to limit the initial burst to a lower speed.