The Management → Hardware screen shows the machines hosting your factories: how much of their CPU, memory and disk space is in use, what the load is and how long they have been up. It answers the question “is there enough capacity, and is it time to expand” — you can see which machine is already at its limit and which is idling.
CPU and memory figures are computed over containers: MultiAgent adds up the numbers of all the machine’s running containers, the way the Docker panel does. So they show exactly what your agents eat and match the Docker panel’s figures. The operating system’s own numbers are always higher: they include the kernel, system programs and the data the system keeps in memory to speed up file work — your agents have nothing to do with that. Disks, load average and uptime, on the contrary, come from the machine itself — they describe the hardware, not the containers.
The Machine List
The list carries one row per machine: the address (clicking it opens that machine’s page), chips for the factories living on it (each chip opens that factory’s page), and three columns — the CPU share, the memory share and the space used across all disks, in percent. For memory and for disks, the percentage is followed by the volume in small type: used of total.
The columns fill in as each machine answers, independently of one another: while there is no answer, a gray dash stands there, not a spinner — so waiting never looks like a frozen page. If a machine did not answer at all, a retry button appears in its CPU column, with the dashes beside it. The list refreshes itself every 15 seconds, and a failed re-measure does not erase the previous numbers: the last known ones stay on screen. When the browser tab is hidden, no polling happens — the machines get no extra checks.
The Machine Detail Page
Clicking the address opens one machine’s page: CPU cores one by one, memory and swap, every disk with its used and total space, load average, uptime and the list of the factories living on it. Below sit three charts: CPU load, memory and disks.
The page also refreshes every 15 seconds, adding one new point to the charts. Points accumulate only in the browser, while the page is open: nothing is stored on the server, and nothing remains after the page is closed. If another measure fails, the last known numbers stay on screen and an error message appears — you see both what there was and that the connection is gone right now.
Watching a Selected Fleet
When a fleet is selected in the header, the list narrows to the machines hosting that fleet’s factories (including the ones assigned to it), and the chips and the machine page keep only its factories. One glance then shows the hardware behind one line of work. More about fleets — on the Management page.
Jumping in from a Factory
From the factory page, the “Hardware stats” icon in its header — right beside the load figures — leads to the hardware: one click opens the page of the machine that factory lives on. The machines themselves and their connection are described on the Servers page.