Currently our 3 registrar servers are not configured for fail-over at all. When we have a planned/unplanned outage we rely on the proxies dispatcher set to route the message to an available registrar. We also shorten the registration interval to take into account that the nat pings wont be available for those endpoints that don't re-register on an available registrar, although, this method isn't completely foolproof.

Once the "failed" registrar is brought back into service dmq_usrloc replicates state back to it when it comes back and we are in service again.

On the other-hand, we do have an ha setup for the proxies involving corosync/pacemaker where there are floating addresses assigned to each kamailio instance, but for the registrars we don't run the ha setup.

Would it be possible to have dmq_usrloc updated to reflect what is currently done in shared-db mode? Shared-db mode currently has some flaws in the scenarios identified in previous posts above, but, at least, for this particular issue, it should, hopefully, resolve the current issue at hand, whereby we have one registrar servicing the nat keep alives for a particular endpoint.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or mute the thread.