Hi,
I have two nodes with vrrp.When I restart the main node, switching to the backup node is without downtime. If I turn off the ethernet interface on the master node or disconnect the ethernet interface, there will be a failure of a few seconds because it can’t send advert
My vrrp log:
VRRP: Netlink reports eth2.2008 down
VRRP: (VMinternal) Entering FAULT STATE
VRRP: (VMinternal) Cant send advert to 172.32.255.6 (Network is unreachable)
VRRP: (VMinternal) Cant send advert to 172.32.255.6 (Network is unreachable)
VRRP: (VMinternal) sent 0 priority
VRRP: VRRP_Group(HA) Syncing instances to FAULT state
VRRP: (ILO) Entering FAULT STATE
VRRP: (ILO) Cant send advert to 172.32.255.6 (Network is unreachable)
VRRP: (Management) Entering FAULT STATE
VRRP: (Management) Cant send advert to 172.32.255.6 (Network is unreachable)
VRRP: (Pubsim) Entering FAULT STATE
VRRP: (Pubsim) Cant send advert to 172.32.255.6 (Network is unreachable)
VRRP: (VMOTION) Entering FAULT STATE
VRRP: (VMOTION) Cant send advert to 172.32.255.6 (Network is unreachable)
VRRP: (Wan) Entering FAULT STATE
vrrp communication is a direct connection between two ethernet interfaces
What could be the problem? Version Vyos 1.3RC3
Thank’s
Zdenek
A failure of a few seconds is expected. Of course it can’t send adverts, its interface is down, so those errors you see are expected.
What the primary router reports (I am assuming those logs are from your master router, you’re not clear) isn’t important though, it’s what your backup router reports that’s important.
With a standard VRRP of 1 second keepalives, 3 keepalives missed means it will take usually ~3 seconds to fail over. After 3 seconds, the backup router sends out the GARP and it takes over the VRRP mastership and everything keeps working.
Now why shutting down the primary router gracefully fails over quicker:
The primary router will stop sending out VRRP adverts BUT its interface is still up and working (and able to accept packets). So it won’t be sending VRRP keepalives, but it’s still functional. It probably takes a few seconds to shutdown, which is just enough time for the backup to have seen no keepalives while it was shutting down, send out the GARP and take over. An almost seemless transition.
All it all, it sounds like you’re seeing VRRP behave exactly as I’d expect it to. The only way I know to make VRRP faster is on Junos (which I have experience with) where you can have say, 100ms VRRP keepalives and failover with say 300ms. I don’t think Vyos supports “fast” VRRP though, at least not 1.2.7 which I am using.
I hope this helps and again, welcome to the Vyos forums and good luck with your Vyos adventures!