add a make commands that start phase1

[mikedanese]

No description provided.

[errordeveloper]

LGTM, needs a rebase...

[colemickens]

Should it use something like ./$$(jq -r '.phase1.instance_prefix')/ as the staging directory rather than ./tmp? If I modify phase1.instance_prefix now and re-deploy, it will pick up the old terraform file and tear down my old cluster.

I was thinking about evaluating $$(jq -r '.phase1.instance_prefix' .config.json) in the top Makefile and passing it as an extra argument so that do [deploy|destroy] becomes do [deploy|destroy] {instance_prefix} and then do uses that extra arg as the staging dir?

[errordeveloper]

@colemickens that's a good question indeed. Terraform does a good job at figuring out when resources need to be updated vs re-created entierely, but there is probably something we should do outside of Terraform, especially to cater for non-Terraform implementations of phase 1. I think instance prefix, or cluster name would be good candidates for that. I also think that generated code in the sense we have it right now should be eligible to a life of it's own, i.e. users might need to modify and redistribute the generated code.

[errordeveloper]

I suppose you can do:

deploy destroy: .config.json
    $(MAKE) do WHAT=$@

Isn't it?

[mikedanese]

Done.

[mikedanese]

@colemickens compromise: name the tf files differently based on instance prefix which is basically cluster name. I pushed a change to all.jsonnet that does that. I'm happy to not use .tmp but I don't know what the structure should look like yet.

[colemickens]

That doesn't include the crypto assets then. I think it would be good for there to be an output folder per- deployment (aka per-instance-prefix) that contains a kubeconfig, the terraform plan, the terraform state and the crypto assets. Maybe I'll just file a separate issue.

[mikedanese]

Ouch, the crypto stuff is definitely not threadsafe right now... Ok I'm going to work on #145 next.

[mikedanese]

I'm going to leave that work for a follow up.