How do the controllers respond to a processor fault?

Every device (for Chameleon substitute the word 'module' for the word 'device') contains separate circuitry to detect a processor fault. This circuitry consists of a watchdog timer that the processor must continually service at a minimum 1.5 seconds rate. In the event of a processor fault, the circuitry will attempt to reset the entire device. Any I/O in the device goes to its power-off state (i.e. the same state they would be in if there was no power applied to the device) during reset. After the reset operation the I/O remains in the power off state while the processor attempts to reboot. The device's I/O are not enabled by the processor unless specific checks are passed. If after a period of time the processor is not rebooted, the electronics will again attempt to reset itself. This will continue to happen until the processor recovers or is replaced.

Note that on Chameleon, processors in the remaining modules in the node will continue to function and can detect the failed processor going off-line.