Supervisor log
The agents that need attention now: failed, blocked, needs input, lost, retrying or paused. One row for each agent, with the newest first, 30 rows at most.
How to read it
- Each row: "… ago · project · reason", then " is ".
- It is a list of the agents in each state at this time, not a history. An agent shows once, at its last change. When the agent leaves the state, its row goes.
- For admins, the panel also shows the entries of
~/.cache/fleet-status/actions.jsonon the server's box, if that file exists.
When it looks wrong
| You see | It means | Do this |
|---|---|---|
| "no supervisor actions yet" | No agent needs attention | Nothing |
<name> is failed |
It crashed 4 times, or met its deadline or budget | Read the reason. See a crash. Resume it when the cause is fixed. |
<name> is blocked |
The agent said it cannot continue | Read the reason and send what it needs |
<name> is needs input |
A question, or a turn with no done |
Answer, or send a message, in workers |
<name> is lost |
Its node is silent | See a machine dies |
<name> is paused |
A usage limit, no login, or held | See a usage limit and a stuck worker |
The filter changes the agent rows, not the file entries. An error of the panel does not show when
the list is empty. Code: fleet/internal/api/state.go, fleet/internal/dash/actions.go.