This message was deleted.
# rke2
a
This message was deleted.
b
This depends on what you are using on prem, is this some kind of on prem cloud provider like Harvester, or do you just mean like a local docker deployment?
g
It is a non cloud bare metal host
b
So you just need to run the installer. If you are minting ISOs you could add the install to your image pipeline, just don't start rke2 until you boot the machine.
You could use the rke2 install module https://github.com/rancher/terraform-null-rke2-install if you wanted to use Terraform, but any configuration as code tool would also work. I officially support the install mod.
g
Thank you for your response. What would be the benefit of Terraform based deployment over Ansible based?
b
Terraform has a state so it can be more efficient during subsequent runs, it can also bootstrap for things you don't want to run every time.
I provision a lot of infrastructure with Terraform, so while Hashicorp recommends using some kind of config as code tool to manage provisioning software on the infrastructure I find it much easier and less complex to use heavily restricted bash scripts that run on the infrastructure for software provisioning. I use Terraform provisioners to execute and deploy the bash scripts. To me, adding another tool (like Ansible) is adding a whole lot of eccentricity (and the need to know that tool) to an already complex topic. I try to keep the number of different tools I use to a minimum so that things are easier to understand and troubleshoot. The added benefit to bash scripts is that it tells readers exactly what was run on the command line should they need to manually replicate anything or undo it.
I have gotten feedback from sysadmins about how much easier it is to troubleshoot the bash scripts that I execute over having to dig through a tools documentation on how it accomplishes the same goal. This significantly reduces the time it takes to get a pet server back up and running during an outage.
I know that is sort of old school in the world of immutable infrastructure, but not everyone is remotely PXE booting hardware or using network storage to back up every system and I want to meet users where they are rather than the ideal situation.