I don’t see why I, the end user, should be forced to make that differentiation since Terraform is effectively useless without the providers from the big three. Statements like “you’re blaming the wrong people” is in the same spirit as what ZeroVer is lampooning.
Like yes, our internal provider was a pain point and we own that but the rest of the drama around providers was just weird. The work to move over 200 repositories through the hoops to keep them updated, especially mature services that may not have been deployed for several months, was difficult to automate and very brittle even when it was.
It broke down at scale and really no one should have been using it that wide spread before it was 1.0.
Hashicorp’s whole treatment of terraform 0.x was horrendous and constantly broke everything, all while they said it was production ready. You can blame whoever you want but the total lack of stability and easy upgrade paths and constant manual fiddling and reviewing output from ‘upgrade0.11’ type commands was ridiculous and a massive time sink for our org.
I've written more Terraform than most people outside of possibly the folks at HashiCorp themselves and probably Gruntwork. Have been using it at 10^5 scale of systems and haven't had nearly the kind of problems that you're describing.
Also the treadmill is really not any different than integrating with _ANY_ Google service. In fact I'd say it's an order of magnitude better. Google has set a standard of breaking changes without notification and if that's one of the providers you're using then I understand.
And well, if it was Azure (as it likely looks to be) the state of their public facing APIs is/was an absolute fucking mess and the preferred way to do anything in their system still seems to be using the UI. I've talked to several people at Microsoft at Azure teams responsible and there's multiple compounding problems there. For one you have 200+ engineering teams with no unified approach to exposing services. Then you've got multiple regions in their cloud that for years didn't have the same authentication system, didn't have consistent features between regions, etc.
There's very little you can lay at Hashicorp's feet for this when the underlying systems themselves have very poor automation.
And then you talk about having 200+ repos and services that haven't been deployed for months and all I can say is the consensus around the need for CI/CD is over a decade old now and infrastructure needs these things just as much as code does.
100% of my Terraform is in a CI/CD pipeline. Yes it was a lot of work to set up, but the alternative is nothing but problems. Terraform is just a tool. It's not a panacea. It will not make all of your problems go away -- it's up to the craftsman how good it is.
Like yes, our internal provider was a pain point and we own that but the rest of the drama around providers was just weird. The work to move over 200 repositories through the hoops to keep them updated, especially mature services that may not have been deployed for several months, was difficult to automate and very brittle even when it was.
It broke down at scale and really no one should have been using it that wide spread before it was 1.0.
Hashicorp’s whole treatment of terraform 0.x was horrendous and constantly broke everything, all while they said it was production ready. You can blame whoever you want but the total lack of stability and easy upgrade paths and constant manual fiddling and reviewing output from ‘upgrade0.11’ type commands was ridiculous and a massive time sink for our org.