Periodic loss of external access to the site

Questions / Help
1

Hiya! I have an app running on fly.io, and every now and then (twice in the last day but usually once every couple of days) I can’t reach it. The dashboard shows everything is fine, and fly status too:

> fly status -a grislygrotto
Update available 0.1.21 -> v0.1.23.
Run "fly.exe version upgrade" to upgrade.
App
  Name     = grislygrotto
  Owner    = personal
  Hostname = grislygrotto.fly.dev
  Image    = grislygrotto:deployment-01H15XGN834A5GWXA0KM2S221S
  Platform = machines

Machines
PROCESS ID              VERSION REGION  STATE   CHECKS  LAST UPDATED
app     918577d1a49ee8  17      syd     started         2023-05-31T10:11:05Z

However when I browse to the site it just hangs and then eventually fails with (in Firefox) a PR_CONNECT_RESET_ERROR.

Notably, if I try and use fly open, I get:

> fly open -a grislygrotto
Update available 0.1.21 -> v0.1.23.
Run "fly.exe version upgrade" to upgrade.
Oops, something went wrong! Could you try that again?

So far, I’ve fixed this by restarting the app. But I would like to find the root cause.

2

fly open doesnt error when I run it from the app directory, where it can access fly.toml, so might just be a bug in that command (running with debug output doesnt show any errors). However when the page opens it still times out with connection reset errors.

Also I can still open a shell connection to the container without issues. If I install curl in there and curl localhost:3000 the site responds without issue, so it looks like some sort of fly.io networking problem

3

Should also note that the issue appears on both my custom domain and grislygrotto.fly.dev

4

It has failed again this morning. Setting up a task to reset the site every morning could be an idea, but seems a bit silly.

5

I have a suspicion this might be IPv6 related - my site was incorrectly bound just on 0.0.0.0:3000. Have changed that to [::]:3000 (which should capture both v4 and v6) and will see if that fixes it

1 Like
6

Yeah that seems to have fixed it. Good work everybody, real collaborative effort, woo!

1 Like
7

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.