As discovered on #1410, the oscillator for the tech port PHY on the front IO board does not always work. We worked around this by power cycling the front IO board when monorail detects a problem (see https://github.com/oxidecomputer/hubris/blob/master/task/monorail-server/src/bsp/sidecar_bcd.rs#L270-L291). Something we've seen in the field a small number of times is this loop running infinitely. While we of course need to get to the bottom of that, a step in the right direction in the near term would be to let the rest of the management network operate while noting the tech ports are down. Ideally we'd have an ereport or some structured way of discovering the switch was in this state as well.
As discovered on #1410, the oscillator for the tech port PHY on the front IO board does not always work. We worked around this by power cycling the front IO board when
monoraildetects a problem (see https://github.com/oxidecomputer/hubris/blob/master/task/monorail-server/src/bsp/sidecar_bcd.rs#L270-L291). Something we've seen in the field a small number of times is this loop running infinitely. While we of course need to get to the bottom of that, a step in the right direction in the near term would be to let the rest of the management network operate while noting the tech ports are down. Ideally we'd have an ereport or some structured way of discovering the switch was in this state as well.