◆ Lab history

Each machine's collector (tools/lab/live/live-collector.py) writes a record every 5 s to its own disk, so a machine that reboots loses only the time it is down; this page is rebuilt from those records every 5 minutes, and the hour view also takes new readings from the live streams while it is open. Hour: 5-second points; day: 1-minute means; week: 10-minute means (the card's dashed "peak" line is the highest value in each bucket of the firmware's peak since the card started or its stats were last reset, not a current hottest sensor: the firmware reports only the mean of its 34 shire sensors and that peak; see Requests for Roman C30 and CF14). A card's die temperature is read once a second only while nobody holds a card on that machine: amber bands mark a card in use (no reading), red bands a card off the bus, grey bands a machine that recorded nothing. Dots are the dashboard's 30-minute samples from before the records began. Dashed red line: 95 °C.