That’s how clouds try to lock you in, by making you use a custom tool that is different for the sake of being different.
If you use standard tools you don’t have this problem.
Containers running on VMs is standard.
A mesh of microservices that depend on cloud queues and managed services is not.
One argument against standard containers is saving dev time. You can still save dev time by using standard open source software. How many different ways are there to implement a queue or a load balancer?
If you really need access to some proprietary technology then by all means use the cloud that offers it. Eg if your customer demands GPT4.5, then go with Azure.
But if you need something standard, don’t get caught in the trap.
I am an older guy that was building kubernetes clusters before eks, aks, gke. So I used terraform to build shit out to make it happen. Azure was 5x the code just to be different. You can try to blame terraform but if you used MS custom tooling it was no different.
What about the way Terraform is a 3rd class citizen on Azure? And there are multiple ever-changing ways of doing everything, major parameters aren't supported, etc. It just makes it more difficult to deal with.
If you use standard tools you don’t have this problem.
Containers running on VMs is standard.
A mesh of microservices that depend on cloud queues and managed services is not.
One argument against standard containers is saving dev time. You can still save dev time by using standard open source software. How many different ways are there to implement a queue or a load balancer?
If you really need access to some proprietary technology then by all means use the cloud that offers it. Eg if your customer demands GPT4.5, then go with Azure.
But if you need something standard, don’t get caught in the trap.