Not all virtual IPs in moving to Master node in a 2 node cluster

Started by Sonderbar, February 01, 2024, 02:08:37 PM

Previous topic - Next topic

I did enable it.
How can I see if it gets better?

Is it dmesg lines like the following that showed you the interface reset you were talking about?

carp: 30@vlan0.30: INIT -> BACKUP (initialization complete)

Or maybe on opnsense: SYSTEM: LOG FILES: GENERAL should I look for "interface down" ??

Just to make sure, do you use an untagged vlan between both opnsenses? I could see this kind of mixup of master and backup once on trunks that could communicate via their native vlan ID. Once I only used tagged vlans and had the (untagged) vlan parent interfaces unassigned in the opnsenses, this kind of behavior stopped.

Its just a guess additionally to all the other troubleshooting done here so far.
Hardware:
DEC740

No. Each of my servers have only 2 physical eth ports:
- a WAN port connected to my ISP box.
- a LAN port connected to the main switch. This port is tagged for all existing VLAN.
The pflink connecting the 2 firewall is one of those VLAN (90).

My switch has the following configuration on the LAN port of my firewalls (same for backup and master):
- PVID: 1
- VLAN tag: 10,20,30,40,90 (=all vlans except 1)

On OPNSENSE I did disable LAN interface and created CARP, VIP for all VLAN except 90. And obviously VLAN 1 do not exist in my opnsense conf.


@mimugmail: you are a magician!
I don't wan't to conclude too early but I can't wait longer to tell you that it's been 48h without any of those entries in my opnsense logs:

2024-02-28T21:31:06 Notice kernel <6>carp: 1@re1: INIT -> BACKUP (initialization complete)
2024-02-28T21:31:04 Notice kernel <6>carp: 30@vlan0.30: INIT -> BACKUP (initialization complete)

I guess that's the "inteface reset" you were talking about.
So if I understand correctly it's a huge stability improvement that may impact positively our subject.
I'll keep you posted.