Feature

Update the BMC without touching your nodes

Upstream power-cycles all four compute modules to patch the management controller. Here their rails are never touched, measured on a live cluster.

The argument, and the measurements behind it →

0modules power-cycled by a BMC update

0cluster nodes lost, measured

48 sfor the board to come back

See it in the interface →The Firmware tab, on a board that took this upgrade. Do it on your board →The upgrade from stock, step by step, and the way back. If the new image is bad →It boots on trial and takes itself back. What you are updating to →A longterm kernel, and the faults taken out.

The four compute modules, powered and counted in uptime. A BMC upgrade leaves every one of them exactly like this.