so far, i’ve seen a few different ways to handle formatting disks, via:
- commissioning scripts, including a python script do do this
- curtin templates
- maas cli, which does look tedious and not at all the way you’d want to manage a fleet
- possibly some other way i’m forgetting at the moment
we’ve tried the commissioning script approach, which seemed to work but has recently been failing. i’m about to convert our layout to curtin preseed templates for our builds to see how that works. the examples in curtin docs look like a good starting point, at least.
for context, we’re looking to manage thousands of physical compute nodes via maas + terraform + atlantis to handle the workflow. the build should look the same across all these nodes, with minor variations in hardware (gpu vs. cpu compute), so doing anything with the maas cli is not viable. we need to automate the whole process.
how are people managing storage layouts? what are some best practices here? i’d love to see some patterns folks are using. or, for that matter, are you not managing storage beyond the default layout?