Raspberry Pi 3D print rackmount
Sometimes a tech hobby starts as a small experiment and grows to become a mess. We need a way to organize the equipment and keep it managable.
Centralizing multiple singleboard computers
Some of the problems that get worse the more singleboard computers are available:
Ethernet cables
Power cables that are getting tangled constantly
SD cards that you are constantly swapping out
Boards without electro static protection
Adding more cables over time causing more tangling
One to three pis doesn't create too much of a problem, but as the group grows four and beyond it begins to become unmanagable.
Keeping them all in one linear group makes a big difference. While there are rack options available to purchase they tend to not be able to grow.
So we need to have an option that is cheap, flexible, and supports tinkering, so let's print one.
Version 1
This is the top rack in the picture above.
The first Raspberry Pi rack mount case was great. It was:
Expandable: you could add on more mounts as needed (up to a limit that would fit in 2U racks)
Stable: Four metal rods kept the mounts solidly connected
Maintainable: each mount had a sliding tray that each Pi was screwed to which allowed one to slide out to bring to ones desk to reimage or for maintenance/troubleshooting
But there was a problem, the mounts allowed for SD cards, but not SSD drives. For most use cases this isn't an issue, but after a while of working with Kubernetes, we wanted to change to an highly Available installation which would burn through SD card reliability and required SSDs.
This led us to look to the Version 2 model.
Version 2
This is the bottom rack in the picture above.
The goals of the prevous rack mount were still in play, but the new one needed to have mounting for SSD drives.
But there's still some problems:
The USB to Sata connector cables are thick and not flexible, and stress on the cable snapped the plastic of the tray.
The new format uses only two metal bars instead of four so it's more flimsy than the other version.
We also ran into issues with the 3d printer. The heated print bed ended up breaking, which is a challenge to diagnose when some prints would work just fine and in others there would be unexpected warping. It's a problem that most others would not have issues with.
Other considerations
In both situations, powering the Pis with power over ethernet would have been great, but PoE often adds $200+ to a switch and would require PoE hats for each of the Pis.
Taking the individual computer approach to home labbing allowed adding on new machines over time, had I known this hobby would grow to this point, it would have been better to invest in a larger server and leverage virtualization like Proxmox to allow spinning up and down of nodes.
The individual computer approach does mean dealing with physical networking and limitations of smaller nodes, this created scenarios (including networking) that need solving that may not otherwise be run into.
Recent lab experimentation (since 2020)
Lab descriptions and details currently not available online, please reach out to discuss.
