The UX of all the other hashi products I've used have been absolutely sublime compared to the industry equivalents. Nomad compared to kubernetes is a particularly stark contrast.
Which tool is that?
On a related note, Vault has a really excellent API. A joy.
Strongly disagree: suppose we have secrets mounted at my/secrets, and we want to read a secret top/secret, which is represented as my/secrets/top/secret path in vault. However, the only way to access it via API is to read _all_ mount points, and match them with the path to split path to mount point and secret. vault cli itself follows the same logic: https://github.com/hashicorp/vault/blob/main/command/kv_help...
Second that. Very steep learning curve for some use cases that could be accomplished in a much easier way. Another kubernetes in disguise.
Used right it's the only way to remain sane on any non-trivial set of infra where you don't end up hand-rolling a half-baked version of Terragrunt.
Getting to this point was a learning process though. Probably six solid years of investment now.
I second that Terraform needs to work at least on dynamically specifying providerblocks. This is where people usually resort to terragrunt.
Workspaces and state layering are in my experience hard on the novices and unfortunately they turn to terragrunt.
This is the initial comment about Terragrunt, which doesn't say that
I work with them daily, and have done so for 4 years now. It still sucks.
Unfortunately it's (Terraform, not TFE/TFC) the best in its class, and quite frankly I'm not smart to solve its issues.
We're striving to build a product that's both highly customizable and has a good UX, improving on many of the pain points. The migration path is straightforward and it'd be great to hear what you think about it!
[0]: https://spacelift.io
Disclaimer: Software Engineer at Spacelift
Plenty of people probably never run into those problems. But if you do, it's pretty terrible.
To be fair, there isn't really anything better. Pulumi and CDK address some of the limitations by using a "real" programming language, but introduce different issues. And bugs in providers are probably unavoidable, and are partly the fault of poor APIs from cloud providers.
Most of these problems I encounter aren't even down to hashi themselves. The GCP provider consumes from google's magic modules repo, which is often out of sync with the GCP API. Or in plenty of cases, it is in sync, but the API does not expose features in the same way as the console.
Who the fuck are they talking about? Clearly not HashiCorp. The most popular tool in HashiCorp's ecosystem that wasn't written by them was written by a consulting firm who hated HashiCorp's UX. Hashi is like Amazon: shitty UX, but features that capture a market early, and so they remain a dominant player despite shitty UX and unnecessary overcomplicated bullshit.
There are only two things that you need to succeed with open source. 1) solve a problem nobody else has solved yet and drive massive adoption. 2) make it so that eventually people realize they can't get any farther without paying for enterprise features, and their lock-in and over-confidence from the free product will make them convince their bosses to pay you.