Strange cow_boy timeout error

Since upgraded to Debian 10, I am facing a very strange error on DB connection.

from time to time, I would see an error

[error] #PID<0.21892.0> running AuroraWeb.Endpoint (cowboy_protocol) terminated
Server: localhost:80 (http)
Request: POST /graphql
** (exit) exited in:<0.468.0>, {:checkout, #Reference<0.688527683.3513253890.14916>, true, 15000}, 5000)
    ** (EXIT) time out

Which I have noticed they are all from Elixir.DBConnection.Connection:init/1.
If I kill the process in :observe.start, of course the process will get restarted and it works as usual.

How do I get debugging started? There is really not much helpful info in the error message.

Please list your Erlang and Elixir versions. If you can, use the latest version of the Ecto library as well – I suspect you are getting DB request timeouts.

I am using

  • Elixir (1.6.5)
  • Erlang/OTP 20

I will try to upgrade Ecto to v3 and see if the problem goes away.