Fork me on GitHub
#biff
<
2024-06-03
>
Andrew Berisha13:06:31

hi all! I'm having trouble deploying to http://fly.io. I found this repo but I'm not sure what to do with the work I did following the tutorial. I launched and deployed the Dockerfile using the flyctl cli but my VM's just stop after a few minutes @foo I found this repo that you made https://github.com/jacobobryant/biff-docker2 but I'm struggling on how to apply it

Jacob O'Bryant15:06:07

Hi--can you see if fly has any deployment/runtime logs? the Dockerfile that comes with new projects should work on Fly; I've deployed it there before. Let's see... I think the default RAM for fly apps is 500m, you'll need to change it to 1gb. other than that I don't think you'll need to do anything special to get the app running. you'd want to set up fly postgres for persistence after you've got the app running.

Jacob O'Bryant15:06:24

besides that, I'll also note that every time I've tried Fly I've ran into weird bugs... last time I got the app running and could log in, but after that, trying to go to the app in another tab always had some kind of networking/DNS error. 🤷

Jacob O'Bryant15:06:02

so let me know how it goes! I've been sticking with digitalocean for now but have been watching Fly with interest from the sidelines

chromalchemy21:06:24

In electric there have been some http://fly.io deploy demos, and I successfully followed one. But I was warned to stick to a more classical host (DO) because of some anomalies that others experienced.

1
Jacob O'Bryant21:06:26

yeah--I think Fly is the future, just not so sure about the present ha ha

🚀 1
1
Jacob O'Bryant21:06:01

I think traditional VMs are a bit better fit for Biff anyway, at least in the early solo developer stage where you only run on a single machine. you can deploy faster + with zero downtime via clj -M:dev soft-deploy, and you don't have to set up checkpointing for XTDB (without checkpointing, if you're deploying with containers, you'll have to wait for XTDB to re-index the entire transaction log whenever you do a new deploy--but with DO/regular VMs, the index persists between deploys).

chromalchemy21:06:01

Did you see this recent tutorial? https://bogoyavlensky.com/blog/deploying-full-stack-clojure-app-with-kamal/ Looks like Kamal is a deployment config library to make a heroku-like experience, portable to different cloud hosts. https://kamal-deploy.org/ > Feed the config file a list of IP addresses with vanilla Ubuntu servers that have seen no prep beyond an added SSH key and you’ll be running in literally minutes. overkill for Biff?

Jacob O'Bryant21:06:22

I hadn't read the article, though I see Kamal is still using Docker, which yeah, I think would be overkill for Biff--`server-setup.sh` works pretty well IMO, and the JVM already has good dependency management/isolation built-in.

👍 1
Andrew Berisha15:06:33

great! thank you for all the messages! I got a little busy with work. I'll go through and try them out, thank you!!

🙌 1