fly launch: database has non-existing domain - :nxdomain

Questions / Help Build debugging
,
1

Hi I’m taking an existing phoenix app that works on digitalocean.

I configured it for three HA databases and fly launch appears to boot and configure them correctly:

Setting secrets on app markably-app-com-db...
Provisioning 1 of 3 machines with image flyio/postgres-flex:15.6@sha256:92917e5770cef6666dddbf17061c2f95b9e19b8155be9a8ce8c35e09e5381167
Waiting for machine to start...
Machine e2867529c996e8 is created
Provisioning 2 of 3 machines with image flyio/postgres-flex:15.6@sha256:92917e5770cef6666dddbf17061c2f95b9e19b8155be9a8ce8c35e09e5381167
Waiting for machine to start...
Machine 90801ed0a13168 is created
Provisioning 3 of 3 machines with image flyio/postgres-flex:15.6@sha256:92917e5770cef6666dddbf17061c2f95b9e19b8155be9a8ce8c35e09e5381167
Waiting for machine to start...
Machine 6e825d07b33078 is created
==> Monitoring health checks
  Waiting for e2867529c996e8 to become healthy (started, 3/3)
  Waiting for 90801ed0a13168 to become healthy (started, 3/3)
  Waiting for 6e825d07b33078 to become healthy (started, 3/3)

Postgres cluster markably-app-com-db created
  Username:    postgres
  Password:    asdf
  Hostname:    markably-app-com-db.internal
  Flycast:     fdaa:0:3adb:0:1::5
  Proxy port:  5432
  Postgres port:  5433
  Connection string: postgres://postgres:asdf@markably-app-com-db.flycast:5432

after the docker image successfully build and when it comes time to run the release, I get this error:


-------
 ✖ release_command failed
-------
Error release_command failed running on machine 9185e7d4f22718 with exit code 1.
Check its logs: here's the last 100 lines below, or run 'fly logs -i 9185e7d4f22718':
  Pulling container image registry.fly.io/markably-app-com:deployment-01J0MWC03GFYKTEC3PA1MZ9Z8S
  Successfully prepared image registry.fly.io/markably-app-com:deployment-01J0MWC03GFYKTEC3PA1MZ9Z8S (756.391441ms)
  Configuring firecracker
  2024-06-18T05:16:45.366841514 [01J0MWDACPP9HSEVMES2H278HJ:main] Running Firecracker v1.7.0
  [    0.038467] PCI: Fatal: No config space access function found
   INFO Starting init (commit: f7402432)...
   INFO Preparing to run: `/app/bin/migrate` as nobody
   INFO [fly api proxy] listening at /.fly/api
  2024/06/18 05:16:45 INFO SSH listening listen_address=[fdaa:0:3adb:a7b:2c9:1953:67e5:2]:22 dns_server=[fdaa::3]:53
  Machine created and started in 1.918s
  05:16:47.384 [error] Postgrex.Protocol (#PID<0.167.0>) failed to connect: ** (DBConnection.ConnectionError) tcp connect (markably-app-com-db.flycast:5432): non-existing domain - :nxdomain
  05:16:47.384 [error] Postgrex.Protocol (#PID<0.166.0>) failed to connect: ** (DBConnection.ConnectionError) tcp connect (markably-app-com-db.flycast:5432): non-existing domain - :nxdomain
  05:16:48.616 [error] Postgrex.Protocol (#PID<0.167.0>) failed to connect: ** (DBConnection.ConnectionError) tcp connect (markably-app-com-db.flycast:5432): non-existing domain - :nxdomain

How do I debug this issue?

Regards,
Michael

2

Added elixir

3

I’m also hitting this issue on an already-created db. Suddenly it stopped working. It refuses to connect and I’ve changed nothing with my DATABASE_URL since initializing

5

Are you still seeing this? I got sidetracked with something else and have not made a re-attempt yet.

6

There’s so many posts on this similar issue. Typical advice is to make sure you’re setting your IPV6 settings correctly. However I’m still hitting it despite following the advice. Frustrating…

8

Deploying a different app and I have no issue.

9

Can you look to see if there are different config options set in both apps?

Out of the box, Phoenix/Elixir doesn’t enable IPv6 support, but we turn it on for you when you run fly launch in a Phoenix app, but this might not always happen if the files we’re looking for already exist.

Specifically, rel/env.sh.eex should set these two env vars (among others):

export ERL_AFLAGS="-proto_dist inet6_tcp"
export ECTO_IPV6="true"

If those aren’t set, then Ecto won’t know how to connect to a database inside your organization’s private IPv6 network.

There may be other areas where config needs to be set – here’s the default runtime repo config from a simple generated Phoenix 1.7.14 app.

Sadly, if one of these things isn’t configured correctly, you’ll get connection issues.

1 Like
10

For a different app, after getting this error, I called fly deploy a second time. After running fly deploy a second time it seemed to not generate this error and the app deployed successfully.

11

I compared my app with a new Phoenix app, and it seems that the lack of the dns_cluster library could be the issue here.

However, I did try to follow the guide for legacy apps at Legacy pre v1.6.3 Phoenix App · Fly Docs , especially the section for IPv6 config, which still didn’t work.

Adding dns_cluster worked in the end.

Maybe the issue is that I already upgraded the Phoenix version of my app to > 1.7 previously, but didn’t add dns_cluster when doing so.