kensuio-oss / dcos-up Goto Github PK

Got this working with latest terraform and latest stable DC/OS. I had to switch to use on demand instances vs spot instances and a few networking changes. Thanks a bunch for putting this out.

What do you think is involved in being able to support adding agent nodes onto a running cluster ? autoscaling groups perhaps ? How can this project be extended ? I'm happy to contribute with some direction.

I'm hitting the following issue, which I am currently suspecting is due to using a higher version of terraform:

bash$  terraform plan
Error loading config: Error loading /Users/cwelton/dcos-up/dcos.tf: Error reading config for aws_instance[dcos_bootstrap]: Invalid dot index found: 'var.provisioner.key_name'. Values in maps and lists can be referenced using square bracket indexing, like: 'var.mymap["key"]' or 'var.mylist[1]'. in:

${var.provisioner.key_name}

I found one trouble after starting script "dcos_install.sh" the consul was stopped. It is because dcos_install.sh do restart systemd-journald from the script. To fix this I recommended change one string in the consul.service in the provision script "setup-consul.sh":

from
Restart=on-failure
to
Restart=always

I spent much time for debug this I hope it will help someone else ...

Hi,

Since today (Sun 31 July 2016) it seems DC/OS version 1.8 is downloaded in the early release and now the Kafka tests work a lot easier. I have documented in a low-level fashion how I made it work on my Mac here https://github.com/petervandenabeele/dcos-up/blob/master/howto.md .

I did not submit this as a PR, as it may not be really suitable for long term inclusion (and is also hacky and not complete).

This connection:

  connection {
    type = "ssh"
    user = "${lookup(var.provisioner,"username")}"
    private_key = "${file("keys/${lookup(var.provisioner,"key_name")}.pem")}"
    agent = false
  }

seemed to work, but this connection:

  connection {
    user = "${lookup(var.provisioner,"username")}"
    key_file = "${path.module}/keys/${lookup(var.provisioner,"key_name")}.pem"
  }

did not and returned the error

* aws_spot_instance_request.dcos_master_node: unknown 'connection' argument "key_file"

until I replaced the line starting "key_file" with the line from the first connection starting "private_key". I am using Terraform v0.9.5.

FYI, in my fork, I added a version that uses Spot Instances in https://github.com/petervandenabeele/dcos-up/commit/bc88e99a5f4d0e57337e3177693fecefbc3a9393 (that commit would probably not cleanly merge anymore here)

This is quite a lot cheaper when using this for experimenting and testing (presuming 0.05 per m3.xlarge node now vs. 0.25 USD per m3.xlarge node for On-Demand Instances).