Altair RISE

A program to recognize and reward our most engaged community members

Nominate Yourself Now!

AI Hub via Docker Compose on GCP

We use Google Cloud for teaching at my institution, so I wanted to take a swing at getting the AI hub (9.10) up and running on GCP via docker-compose, but I am hitting a wall and not really sure where to turn to next.

A few notes:

- Ubuntu 20
- Docker 20
- 4 core and 16GB ram, 50 GB hard drive

Admittedly I have made minimal changes to the .env file, but because I will not have a domain, I was attempting to include the server's public IP as http://<IP_here>.

I am attaching the logs here, which were created via

- docker-compose up -d rm-init-svc
- docker-compose up -d

I noticed via `docker ps` that some of the services appeared to be restarting, so I started everything back up again (via the two commands just above) and captured the logs. The services were running for a good chunk of time.

For context: My course is in the spring, and I was hoping to expand the depth of the content by having students build and schedule processes in the cloud on GCP.

Find more posts tagged with

AI Studio

Accepted answers

All comments

btibert

Hoping to bubble this up. I understand that there are templates that can be used on other Cloud providers, but we teach with Google Cloud because of the very generous free credits that we can provide our students. As such, I need to stay inside GC and can try to coach my students on a docker install, but its not clear to me the best way to do this.

jpuente

Hi!
Just to make sure, you have to change you IP in three places:

PUBLIC_DOMAIN=

PUBLIC_URL=

SSO_PUBLIC_URL=

But I guess you've probably already done that. Otherwise, what I see in the logs are multiple permission denied errors copying to /rapidminer/.env. Could you please check permissions there? This might be what triggers all other errors.

btibert

Thanks for your note! It's been a minute since I attempted this, but I do recall that I wasn't sure how to configure the URL bits, especially the SSO. For context, I was attempting to spin this up on a VM in Google Cloud. I have access to a public URL, and used that as the PUBLIC_URL and DOMAIN, but I may have left the SSO blank, as that is not applicable for my setup or teaching need. In terms of the security in the logs, thanks for picking up on this. Outside of some of the confusion on the URL setup, next time I attempt to get this running, I will run the docker commands with sudo.

jpuente

One additional comment. The SSO name might be a bit misleading. It's related to "single sign on" within the components of the RapidMiner platform (AI Hub, Grafana and JupyterHub mostly). I'd recommend to fill it in even if you're not going to use any single sign on with external tools, with I guess was your interpretation.