However, I realize I ended up forgetting the id at the end and copied the machine image from my dashboard that was listed at the registry link for the -i flag:
And it updated, but I noticed now the registry.fly.io… part has an @sha256 hash attached to it and the machines now attempt to start but then report that they can’t find a good candidate, and our service is unreachable. I could use some help rectifying this, please.
Oh thank you, good chance it could be this, I am using Elixir/Phoenix. I can check in about an hour — I had to step out to pick up my little one.
This does appear to be what’s happening, I’m seeing the same error [info] Protocol 'inet6_tcp': invalid node name: ... (only I didn’t explicitly set a RELEASE_NODE in my fly.toml, just the RELEASE_COOKIE).
After further inspection, I realize now that my machines reset to an older configuration whenever I deploy a new release. I’ve noticed that they seem to reset back to this setting performance-cpu-2x@16384MB and no longer list the sha256 hash in the machine_image.
The change to fly.toml didn’t seem to work when it has the machine image listed with the sha256 hash. I’m going to revert back to the older machine settings and stick to updating like this: fly machine update id -vm-cpus … to avoid changing the machine’s “image” configuration.