Open Telemetry
Metrics and log aggregation sound great, but if those are each separate solutions, and sentry is another separate solution, then we have to go to a lot of unrelated services to try and cobble together a complete view. An OpenTelemetry initiative would be awesome. You’ve already got grafana and prometheus, so some of the usual suspects are already in place.
GitOps
The other one would be to come up with a clear solution for gitops. Right now, you have a terraform provider that doesn’t offer enough platform coverage. For example, I can deploy an app w/ an existing docker image, but I can’t deploy one where the image is built at deploy time. Also, once I use terraform, I can’t use flyctl. So deploying updates once the infrastructure is in place doesn’t work either. There are others limitations. As side projects go, it’s phenomenal. But side project is what it looks like at the moment. (fly.toml is unnecessary for app deployment?)
Flyctl is slick, but it’s procedural, and isn’t what’s powering the terraform provider. So it’s… really only suited to small projects with 1 environment, or folks who aren’t doing gitops. You have to have fly.tomls, which are a pain because they’re targeted at specific environments and would be in conflict w/ terraform anyway.
Most concerning is the announcement that flyctl is where you were planning to move all of the “brains” of your deployment was concerning. (Flyctl Versions, Autoupdating, and the CLI Apocalypse) Because unless this becomes the brains of your terraform provider, you’re working on numerous, seemingly independent deployment strategies. (terraform provider → graphql, flyctl → graphql, and graphql directly). And a gitops story will be left out here.
So at the moment, I’ve used terraform for 3rd party things, like rabbit. But I have to stick w/ flyctl for my code, and to have a CD solution
Managed Postgres
I just read the barman post which sounds like a great step. Will look into that. But again, its baked it into flyctl, not the api. Regardless, something fully managed there would be phenomenal.