Thank you for your help.
Indeed it's working as expected when this option is set.
I will check on keepalived side if I can improve the option
description on the manpage.
On Thu, 21 Oct 2021 at 20:47, Julian Anastasov <ja@xxxxxx> wrote:
> On Thu, 21 Oct 2021, Alexis Vachette wrote:
> > Hello everyone,
> > If this is not the correct list to address my issue feel free to tell me.
> > We are heavily using Netfilter/IPVS with maglev scheduler algorithm
> > and it's working great on most use case.
> > The only case when we face issue is when inhibit_on_failure option is
> > enabled.
> > Here is our setup:
> > - 2 LVS director in active/passive mode
> > - 2 backend node
> > Exposed service is SSH using port 2222.
> > You can find below the LVS state when issue is occurring:
> > root@dev-lvstest-01 ~ ❱ ipvsadm -L -n
> > IP Virtual Server version 1.2.1 (size=1048576)
> > Prot LocalAddress:Port Scheduler Flags
> > -> RemoteAddress:Port Forward Weight ActiveConn InActConn
> > TCP 172.16.42.170:2222 mh (mh-port)
> If inhibit_on_failure sets weight to 0 you may need
> to add the mh-fallback scheduler flag to allow traffic to
> use fallback server(s).
> > -> 172.16.42.168:2222 Route 100 0 0
> > -> 172.16.42.169:2222 Route 0 0 0
> > Here is a working example output:
> > root@dev-lvstest-01 ~ ❱ tail -f /var/log/kern.20211021 | grep
> > "172.16.42.170"
> > Oct 21 09:17:34 dev-lvstest-01 kernel: [ 75.590633] IPVS: lookup
> > service: fwm 0 TCP 172.16.42.170:2222 hit
> > Oct 21 09:17:34 dev-lvstest-01 kernel: [ 75.590637] IPVS: mh: TCP
> > 172.16.42.170:2222 - no destination available
> Julian Anastasov <ja@xxxxxx>