Previous Years
For previous events, we used USB auto-configuration. We generally tried to keep the BYOC layout the same, year-to-year and put the same switches in the same location. On the occasion that something changed (public IP space, switch locations, etc.), we maintain some automation to generate switch configurations from CSV export of the master subnet allocation sheet.
Drawbacks
To put each switch in it's former location is a significant challenge. We have 2 1/2 pallets of switches. Each switch can only go in one spot, or reconfiguration is needed. At setup, this challenge is manageable. The switches are unboxed and sorted into sequential order for distribution when the tables are available. The sorting and inventory time is significant and ties up a dozen volunteers for most of a day.
Tear-down, is less manageable. We label the cardboard boxes with the switch label, which was based on location at the 2017 event. The cardboard boxes are disintegrating after 6 years. The particular box for each switch must be found, and the correct switch returned to it. This is a highly distributed, error-prone process. The switched generally end up in an arbitrary order on the pallet (due to tear-down time constraints).
We do all this to avoid connecting a cable to the console of each switch, wiping the startup-config, and loading another from USB. A process which takes up to 10 minutes per switch.
DHCP auto-configuration
The switches also support auto-configuration via DHCP/TFTP, which opens the door for the switches to be location agnostic. The dream is that we can take a switch off a pallet, connect it to a distro and it will receive the correct management address, and pull a working config based on it's location on the distro.