Set up experiment around mapping with open machine learning

[pantierra]

We need two similar setups to map empty databases with the Tasking Manager. Such stacks can potentially being created by using the OSM Seed stack. One stack should use iD and the other one RapiD. They can be linked in the Tasking Manager as custom editors.

• Functional iD editor pointing to an empty and self-hosted OSM database
• Functional RapiD editor pointing to an empty and self-hosted OSM database
• Both editors can be pointed to, through the TM's custom editor field

ping @bopercival-hot @dakotabenjamin @ramyaragupathy @willemarcel

[dakotabenjamin]

We can either setup on a single machine (EC2) using docker-compose or on a Kubernetes cluster. It's not clear what the infrastructure requirements will be from the OSM Seed stack docs but we can compare differences and our own needs here.

Using Kubernetes on AWS has a base cost of $0.10/hr and then each pod (ec2/container/Fargate) has certain base costs. It may take some time to architect an optimized stack if they haven't been already. It is however possible to run some pods but not all, if we only need certain resources.

OTOH, a single EC2 instance may need to be quite large to handle the workload of an entire OSM stack. I am currently reaching out to some developers to try to get a better idea on the differences in workload requirements.

[geohacker]

Tagging @batpad @Rub21. Happy to talk through anything on the osm-seed front.

[dakotabenjamin]

I have almost everything working for osm-seed. You can now:

1. Go to the URL and create an account (Will share privately until we setup https and a dns record)
2. Email isn't working yet so for now we have to manually confirm emails in the db
3. Once I have edited your user status, you can access iD at port 8080 (not the "Edit" button on the osm page)
   I have managed to create and edit a way without problem.

What's left:

• Fix the email server. Amazon doesn't let you send emails from an EC2, we have to use Simple Email Service (like TM). I have this configured in the .env but need to troubleshoot why it isn't working.
• It doesn't seem like the planet data is showing up in iD. I need to check the config that we are processing the data, and that we have proper access rights to the s3://hotosm-seed bucket
• I need to setup https and a url. Not sure the best way to go about the encryption. Any thoughts @geohacker ?

[geohacker]

Once I have edited your user status, you can access iD at port 8080 (not the "Edit" button on the osm page)

We have a fix for running iD as part of the rails app instead of a separate container. This should land in master in a few days.

Fix the email server. Amazon doesn't let you send emails from an EC2, we have to use Simple Email Service (like TM). I have this configured in the .env but need to troubleshoot why it isn't working.

For OpenHistoricalMap, we use SES and it works really well.

It doesn't seem like the planet data is showing up in iD. I need to check the config that we are processing the data, and that we have proper access rights to the s3://hotosm-seed bucket

That's weird. Did you use https://github.com/developmentseed/osm-seed/tree/develop/populate-apidb to populate the api database with a planetfile?

I need to setup https and a url. Not sure the best way to go about the encryption.

Do you use Route53? I'd recommend that. You can get an SSL and setup a loadbalancer to route requests to the ec2.