CODEWITHSUNDEEP
amazon-web-servicesdockernginxamazon-elastic-beanstalk
CorayThan
CorayThan

Reputation: 17825

AWS elastic beanstalk container link not working

I've made a Dockerrun.aws.json file that defines two Docker images I want to run with AWS Elastic Beanstalk. One is an nginx container with a static front-end website and API routing, and the other is the backend API. But for some reason the container linking isn't working and the nginx container can't forward requests to the API.

Whenever nginx tries to proxy_pass requests, I get this error:

2018/01/02 22:03:01 [error] 5#5: *22 connect() failed (111: Connection refused) while connecting to upstream, client: 172.31.1.70, server: localhost, request: "POST /api/auth/login HTTP/1.1", upstream: "http://172.17.0.2:5000/api/auth/login", host: "my-app-dev.us-west-2.elasticbeanstalk.com", referrer: "http://my-app-dev.us-west-2.elasticbeanstalk.com/login"

From reading this aws multi-container documentation it looks like links are just made automatically?

Here's my nginx file:

server {
    listen       4000;
    server_name  localhost;

    root   /usr/share/nginx/html;
    index  index.html index.htm;

    # Matches any request with a file extension. Adds cache headers.
    location ~ .*\..* {
        try_files $uri =404;
        expires 1y;
        access_log off;
        add_header Cache-Control "public";
    }

    # Matches api requests and redirects to the api server's port.
    location /api {
        proxy_pass http://api:5000;
    }

    # Any route that doesn't have a file extension. This helps react router route page urls using index.html.
    location / {
        try_files $uri $uri/ /index.html;
    }

}

And my Dockerrun.aws.json file:

{
  "AWSEBDockerrunVersion": "2",
  "containerDefinitions": [
    {
      "name": "proxy",
      "image": "my-app/proxy:1",
      "essential": true,
      "memory": 64,
      "portMappings": [
        {
          "hostPort": 80,
          "containerPort": 4000
        }
      ],
      "links": [
        "api"
      ]
    },
    {
      "name": "api",
      "image": "my-app/api:1",
      "essential": true,
      "memory": 128
    }
  ]
}

This same setup is working locally with docker-compose but something about how I'm trying to do it with aws is obviously not correct.

Upvotes: 1

Views: 2099

Answers (2)

CorayThan
CorayThan

Reputation: 17825

It appears that my java app was silently unresponsive due to an insufficient amount of memory. Upgraded my aws instance and changed "memory": 128 to "memory": 512 and it works. No networking issue after all.

An even better solution is changing "memory" to "memoryReservation". That way I define a minimum required memory, but it can still make use of as much of the EC2 instance's additional memory as is available.

Upvotes: 0

Vineeth Guna
Vineeth Guna

Reputation: 398

From the nginx error log above, nginx is able to resolve the upstream server endpoint as 172.17.0.2:5000 which means nginx docker container is able to resolve the ip of api docker container

The error thrown was Connection Refused which be can caused by

  • Port 5000 is not exposed by the api docker container
  • Docker resolution of api container ip address (172.17.0.2) is incorrect
  • The network on which nginx container is running is not the same as api container due to which it could not connect

For the first case you can go SSH into the EC2 instance and check the api docker container information and see whether 5000 port is exposed, if the port is not exposed then Dockerfile should be modified to expose the port

Second and third case should not happen as it is taken care by AWS

Anyways to verify second and third case, you can SSH into the EC2 instance which are running the docker containers and then connect to the nginx container and do a telnet 172.17.0.2 5000, if it does not succed there is clearly a network issue

Upvotes: 1

Related Questions