Comparison to `deploy-rs`

[poly2it]

Hi!

We've been evaluating options for Nix deployments, and the two contenders we're down to are morph and deploy-rs. We need to support deployments from non-committed revisions for testing and need compatibility with disko. Looking at these two options, it's not trivial to see which would be a better fit. According to the authors of deploy-rs, they wrote it after using morph, but they did not elaborate on why they were unsatisfied with it.

I'd be eager to hear if anyone is able to help us conclude which would be a better fit for us.

Thank you!

[adamtulinius]

Hi there

I've never used deploy-rs, but their readme seems to point out some very big differences:

Morph doesn't have flake support, magic rollback (even though that doesn't have to be built into the deployment tool!), or support for multi profile.

On the other hand Morph has support for health checks (both before and after deploying a server). To this day I'm still puzzled as to why none of the other deployment tools supports something like health checks out of the box; we for sure wouldn't want to manage our infrastructure without it.

Regarding disko; morph has no opinions about matters like that, and we use disko at DBC too. :)

I can't tell you what's right for your use case, but I can say that we rely on morph every day for deploying production systems, and have been doing that for years.

Good luck picking 😄

(For what it's worth: Me, personally, also have a morph alternative in the works. It will be announced later this year, but some keywords are fully parallel deployments with rules for what can and can't run in parallel, and a flexible deployment model that allows for specifying alternative deployment flows in nix. Too big changes to make to morph at this stage. Expect Morph's existing deploy format to be fully supported, along with a format close to Colmena's down the line.)