Skip to main content

Read a node's status

When tasks don't start, or start slowly, the node they run on is the first thing to check. Its row gives you a quick answer; its details tell you why.

Who can do this: anyone who can view nodes.

Read the list​

The Machines tab: the New node button and two online nodes with their load bars, S/R/A counts and Enabled ticks

  1. In the sidebar, select Directory, then the Machines tab.
  2. Find the node's row. In a long list, type its name in Search and press Enter, or select a column's heading to sort by it.
  3. Read the dot before its name: green means the node is online.
  4. Check Enabled: a ✓ means the node takes new tasks.
  5. Read the three bars under Load: C for CPU, M for memory and D for disk. A bar turns yellow, then red, as it fills.
  6. Read S/R/A: slots, runners and logged-in accounts on the node. Hover the numbers for the words.
  7. Check the clock column, the last time the node reported. A node reporting normally shows now or a few seconds.

Find out what's wrong​

A node's Overview tab: Last Seen, OS, CLIs, Slots, Running tasks, Held and Accounts, with the Enabled, Drain and More buttons

  1. Select the node's name. Its details open beside the list on the Overview tab.
  2. Read anything at the top. A healthy node shows nothing there; otherwise each finding and issue is named. Hover an issue for its remedy.
  3. Check Last Seen and Accounts. When the node has stopped sending a report, Last Seen says Some reports are late; hover it for which report and when it last arrived. Accounts counts the AI accounts with a working sign-in on this node. A node with none shows None, with why when it's online — for example not logged in or no account placed here — and can't run tasks.
  4. Check Slots, Running tasks and Held: how many worktrees the node has, how many tasks are running there, and, when a task holds a slot, which tasks hold them, each linked.
  5. Under Metrics, pick 1h, 6h, 24h or 7d to see how CPU, memory and disk have moved over time. Dashed lines mark the warning and critical levels, each labelled with its percentage.
  6. To dig further, open Troubleshooting at the bottom: when the node last sent each kind of report, and any setting that differs from what was set for it.

What to do next​