GHOSTS on Amazon ECS

[msprack]

Currently working on getting the GHOSTS Docker file started on Amazon's Elastic Container Service. Having some difficulties and about to stop for a couple of days to focus on other things so I'm writing this post for three reasons:

1. So I can remember what I was doing when I come back to this.
2. In case someone else has already done this and can save me the trouble.
3. So I can save someone else some headaches if they start on the task before I've figured it out.

With that in mind, here's what I've done so far:

1. Adapted the ECS tutorial to create a cluster and task for the provided image.
2. Edited docker-compose.yml to comment out all of the volume stanzas due to a rather annoying error that stems from AWS not allowing source paths to be specified with FARGATE deployments.
3. Tested the api/home endpoint, got a response code 500. Traffic capture on the windows client shows errors for all attempted endpoints as well.

Currently working on troubleshooting number three. Based on the following log file, it would appear that errors are occurring due to the SQL client being unable to resolve ghosts-postgres from the connection string. I briefly played with using ECS service discovery to work around this without success.

2021-09-19 20:29:59 System.Net.Internals.SocketExceptionFactory+ExtendedSocketException (00000005, 0xFFFDFFFF): Name does not resolve
2021-09-19 20:29:59 at System.Net.Dns.GetHostEntryOrAddressesCore(String hostName, Boolean justAddresses)
2021-09-19 20:29:59 at System.Net.Dns.GetHostAddresses(String hostNameOrAddress)
2021-09-19 20:29:59 at Npgsql.NpgsqlConnector.Connect(NpgsqlTimeout timeout)
2021-09-19 20:29:59 at Npgsql.NpgsqlConnector.RawOpen(NpgsqlTimeout timeout, Boolean async, CancellationToken cancellationToken)
2021-09-19 20:29:59 at Npgsql.NpgsqlConnector.Open(NpgsqlTimeout timeout, Boolean async, CancellationToken cancellationToken)
2021-09-19 20:29:59 at Npgsql.NpgsqlConnection.<>c__DisplayClass32_0.<<Open>g__OpenLong|0>d.MoveNext()
2021-09-19 20:29:59 --- End of stack trace from previous location ---
2021-09-19 20:29:59 at Npgsql.NpgsqlConnection.Open()

[sei-dupdyke]

Although the default docker-compose file assumes the API and Postgres containers running on the same host, I've not personally tested it in AWS, and I can't think of what would be causing a connection issue. As it is, it should be docker that is handling DNS inside of docker and between containers. Perhaps doing docker exec -it ghosts-api and then looking at what the container can see will help us debug a bit more?

[msprack]

Did some digging and it looks like aws ecs' fargate will ignore the hostname directives, so dns is no good in this situation. It also seems like as long as all of the containers are in the same "task" they'll end up running on the same host. Adding "127.0.0.1 ghosts-postgres" to the api container's /etc/hosts file seems to have worked so I figure the easiest thing for me would be to deploy a modified docker image of the ghosts-api with a modified application.json to connect to 127.0.0.1 for the database