Per-server: the readings
Overview. The control panel read over the live login in one round trip and refreshed quietly while on screen: load, memory, disk, what the box is, what is listening, who is on it, what is eating the CPU, the disks, the templates the box remembers (read from /var/lib/lens on the server), and the pending updates with Upgrade (all, or a ticked subset by name) and a signed Restart. On a database member the engine's state and its role in its set are read beside the vitals and change what the dangerous buttons do.
Monitor. The box right now: every number a rate between two samples, six minutes of history, every core at once, network mirrored on one axis, hover charts, what is eating it, and a spike log — a number crossing a line is a mark on the timeline with the processes that were running at that moment. Runs only while looked at; nothing polls thirty-six boxes in the background and nothing was installed.
About this server. Everything Lens knows how to ask a Linux box, asked at once: the Find out catalog's 209 curated reads (the ones needing no input) run as one script per group over one channel, filling in a section at a time. The whole reading is copyable as text for a ticket or a handover note.
Update badge, Check for updates, Update all. Every server's row on the map and in the tree carries what it is waiting on ("N updates · S security", "restart required") from a reading taken the moment a shell goes live, with Update and Restart as verbs on the row. A group's Check for updates refreshes package lists on every box (connecting the ones without a shell, after asking) and Update all sends one job per box. A member of a replica set, a config set or a router does not get Update/Restart on its row: its updates go through the Databases section, where the safe walk does secondaries first and the primary stepped down last.