Leftover DNS Records

We’ve been noticing that fly is sometimes a bit slow to update our internal ipv6 dns records. For example, if I run fly status on our platform-ca-workflows-api right now it shows that there are 2 instances running but if I run fly dig for platform-ca-workflows-api.internal I see 4 AAAA records.

We have a proxy service that routes requests via these internal urls. If those urls are stale then we’ll be sending traffic to boxes running old code. Is there a way to force a DNS refresh? Do you have recommendations on how we should handle the existence of these stale records? Or is this maybe a bug?


DNS lag is a known issue, and right now there’s no way to refresh. That said, improving this situation is being worked on.

Meanwhile, if you’re running a proxy, you might want to look at the Machines API. Machines are a bit different than the current ‘standard’ Fly deployment platform. One of their upsides is getting a fixed IPv6 address, so you could use those in place of DNS.


Are both private and public IPs fixed for Machine VMs?

Right now, yes. We are debating how to handle this in the future, so just keep your eye on things when we ship fly apps on machines for real.