Nodes

The compute nodes this engine runs VMs on. Nodes are stateless: they PXE-boot from the engine, run what they are given, and keep nothing that matters across a reboot - which is what makes replacing one a matter of racking another.

Memory

How much of the node's memory the VMs on it have been given, against how much it has. The engine places a new VM by these figures, so a node reading full takes no more work until something moves off it.

Storage

How much of the node's own disk is used. Only VMs whose images are on node storage count here - images on a shared pool cost the node nothing.

Load

What the node is actually doing, as reported by the node itself. High load with low memory use means the VMs on it are working hard rather than that the node is full.

Terminal

A shell on the node, in the browser. The Terminal button in a node's pane opens it over the page; the arrow in its bar moves it to a tab of its own, which is the way to have several open at once.

You are logged in as the account the engine itself uses to reach its nodes, over the engine's own connection to them — nothing has to be opened to the outside, and there is no password to know. Closing the terminal, or the tab, ends the session and whatever was running in it.

• Only an administrator who is not read-only may open one, and each one is written to the engine's log with who opened it.
• Escape, Tab and Ctrl+C go to the shell. To copy what you have selected use Ctrl+Shift+C, and to paste Ctrl+Shift+V (on a Mac, Cmd+C and Cmd+V).
• On a phone or tablet a row of extra keys appears under the terminal - Esc, Tab, Ctrl, Alt, the arrows and a few more that a touch keyboard lacks. Ctrl and Alt apply to the next key you press: tap Ctrl, then c. The keyboard button in the bar shows or hides the row anywhere.
• If the connection drops, or you type exit, Reconnect starts a new session. The old one is gone: for work that has to outlive the browser, run it under tmux or screen on the node.
• There is no button for a node that is asleep or shut down.

Serial console

The node's serial console, through its management controller — Intel AMT or IPMI. It works whatever the node itself is doing, so it is the way in when a node does not answer, and the way to watch one boot. The Serial console button is there for a node that has reported a controller address (Out-of-band IP in its pane).

It behaves like the terminal, with these differences:

• A controller carries one serial session. Opening the console takes it from whoever had it.
• What you see is whatever the node writes to its serial port, which may be nothing until you press Enter, and may ask you to log in.
• To leave, close it — or type Ctrl+] on AMT, or ~. at the start of a line on IPMI.

The engine signs in to an AMT controller with AMT_PASSWD from its configuration, and to an IPMI controller as IPMI_USER with IPMI_PASSWD (ADMIN and ADMIN unless set).

Node status

A node that is up is available for VMs. One that is down has stopped reporting - it may be off, rebooting, or unable to reach the engine. VMs on a node that goes down are not moved automatically.