#Cannot reach my deployed Phoenix server on AWS ECS

40 messages · Page 1 of 1 (latest)

sleek peak
#

So, what I have:

  • Successfuly deployed elixir server on AWS ECS (I can see logs, attached)
  • Runtime config as
  host = System.get_env("PHX_HOST") || "example.com"
  port = String.to_integer(System.get_env("PORT") || "4000")

  config :platform, :dns_cluster_query, System.get_env("DNS_CLUSTER_QUERY")

  config :platform, PlatformWeb.Endpoint,
    url: [host: host, port: 443, scheme: "https"],
    http: [
      ip: {0, 0, 0, 0},
      port: port
    ],
    secret_key_base: secret_key_base

When I ssh to another machine in the same private network, I cannot reach my phoenix app via curl

curl -i -H "Accept: application/json" -H "Content-Type: application/json" -X GET http://10.0.2.173:4000
curl: (28) Failed to connect to 10.0.2.173 port 4000: Connection timed out

It basically hangs for 1 minute and then connection timeout.

I'm pretty sure something is wrong here, maybe with ports, but would be happy to hear some ideas.

image.png
kind sun
#

Your AWS hasn't got PHX_HOST set and likely SYSTEM_PORT. It's trying to load example.com, looks like?... Hopefully someone better versed in Phoenix can help but that's just at a glance from me.

sleek peak
#

As I understood https://example.com is just there to generate links inside phoenix app. I don't have any UI/html yet, only one health-check json endpoint

kind sun
#

Try setting it to the server's address, might work.

peak berry
#

Also, how are you starting it? As a release? Are you using the ./server call or something else?

#

You might be missing the server: true config

sleek peak
#

I auto-generated docker file from Phoenix deployment guides, and build image from it

#

Should I maybe expose here in dockerfile 4000 port? 🤔

peak berry
#

No, but have you eliminated the possibility that your machine you curled from can't access the machine that Phoenix is hosted on?

sleek peak
#

yes, I can access other machines in the ECS cluster that are in same network

#

Only phoenix one that is deployed is hanging

dim kestrel
#

And you're sure the environment is correct? You shouldn't be getting logs with example.com.

sleek peak
sleek peak
peak berry
#

Yeah, it's definitely work if you got that log line

#

Dumb question: there's no way for you to check whether or not the Phoenix app is up from the machine that's running it right?

#

I would imagine it would be kinda pointless, you would be going into the container itself.

sleek peak
#

Well, the logs showing that it is booted 🤷‍♂️ Machine records all stdout, so if it would crash I assume it will be in logs

peak berry
#

The fact that it times out tells me that it isn't a problem with Phoenix though

sleek peak
#

Fun fact, the same exact image I run locally and everything works

peak berry
#

if you haven't changed anything, and just did a simple endpoint, it should have logged something.

sleek peak
peak berry
#

Yeah, which is why I can't help to suspect there's something off with the network possibily

#

The request has disappeared to the AWS clouds, never to be seen again

dim kestrel
sleek peak
#

But still accessible under localhost/0.0.0.0:4000

kind sun
#

EC2 might be relevant? Don't know AWS but the idea is the same.

sleek peak
#

I use same terraform script/config as I use for other services, so 4000 port that I specified in config should be open

kind sun
#

Does your Dockerfile have any EXPOSE mentions?

sleek peak
#

Just added 30 min back, but still no success

image.png
kind sun
#

Maybe exposing both will help, I'm just spitballing.

sleek peak
#

Finally

image.png
#

Sorry, was my bad. As we though some small mistake in network configurations

#

@kind sun you was right, apparently for inbound rules in my claster I was allowing only ports above 8000 😅

image.png
#

So 4000 was rejected by security group all the time

kind sun
#

I have a beer in hand, mate. Congratulations!