Following infrastructure#1747 `altssh.gitlab.com` became unavailable. I have found the root cause in the old `lb10` and `lb11` not being able to load the new configuration. The fix was to simply update `haproxy.cfg` with the new nodes. The outage ended today at 07:22 UTC. This is what we need to avoid this from happening again in the future: - Review our monitoring for altssh. - Move the service to a new set of load balancers in ARM. - Delete everything from the old platform to avoid confusion. - As a more general thing, don't leave services/nodes half migrated.