I have a local application server. I’m running about 10 applications on it. The workflow is simple:
- define a docker-compose file
- set ‘restart: always’
docker-compose up -d- (let watchtowner update container as needed)
This has worked with every single application I’ve ever used. Even something complicationed like airflow, and certainly mlfow. It seems the wandb design pattern is non-standard and the only entrypoint is the command wandb server start. I thought I might be able to put this as the command in my docker compose, but this command proceeds to download a docker container . Is it possible to do this very standard workflow with wandb?
Hi @djkl9r4, just wanted to check if you followed the instructions from our docs? That’s the way we recommend so wondering if you would have any problems with it?
Thanks for the reply luis.
Yes I did follow the instructions. The problem I run into is when I want to add this to my application server and make modifications (backup artifacts to mounted docker volume, use a docker network to share traffic with another container, run this as part of as a larger application using docker-compose). Mostly I would like to create IaC and setup my containers to be configured as needed and updated automatically. The problem is after step 6. I’d like to add a volume for backups, change the network, hook up with my sql install put it in a docker-compose.yml, and I don’t see any instructions on how to do this.
Hi @djkl9r4, thanks for sharing this details. I checked internally and we don’t have official docs for that process so our recommendation would be to rely on the formal process, since other processes haven’t been fully tested and may potentially have issues, so we don’t ensure that’s an option