I'm a new sysadmin at a large company and I need to upgrade our way of setting up around 5 bare metal servers and about 30 VMs—mostly RHEL, with a few Windows servers. Currently, we're using a somewhat homemade method that involves manually creating kickstart files from various templates and PXE booting the VMs from a central server. We also have some custom gradle plugins that help us manage artifacts and build RPMs, but we're looking for something simpler.
Ideally, we want to create a final artifact from our source that we can install on the boot server and use to kick off installations for all machines via PXE, preferably with minimal manual intervention. I'm considering tools like Ansible, Foreman, Packer, RHEL Image Builder, MaaS, and Terraform and would love to get opinions on the best approach.
Here's what I'm thinking so far:
1. Use Ansible for machine configuration
2. Use Foreman for deployment and monitoring
3. Use Packer to create deployable images (though I'm unsure if it can create images for bare metal servers)
I'm unsure if I'm on the right track due to my limited experience and would appreciate any suggestions or tips. Thanks!
3 Answers
You might want to check out Foreman/Katello or Cobbler; they're great for your needs and can simplify many of your provisioning tasks.
Switching to Ansible with a golden image setup really streamlined our processes as well. Just remember, don't aim for perfection right away. It's more effective to standardize your setup step by step to prevent overwhelming yourself.
For Linux, I recommend not creating golden images. Instead, provision blank VMs and configure them with Ansible. From my experience, using Terraform for infrastructure really simplifies things, and most hypervisors support cloud-init for basic provisioning, which means you won't have to rely on PXE booting. You'd set up your VMs with a minimal OS and necessary packages, including SSH for management. With a management node running Ansible, you can then configure your VMs as needed easily. Basically, maintaining your Ansible configuration becomes your main task, with adding new VMs being straightforward. Plus, you can skip the PXE setup by using cloud-init with Terraform.
I totally see where you're coming from! Those bare metal machines are a different beast, and I'm also trying to figure that part out.
Thanks for the info! I get that for the VMs, but how do we adapt this for the 5 bare metal machines that will be handling our software? I still feel like I'll need PXE for these. Can you clarify that part?