A machine idled out
The machine suspended after a quiet spell and its state was saved. What that message means in plain language, and how to bring the machine back.
- What you're seeing
- A notification says the machine idled out and its state was saved — the wording is quoted below. The terminal in that machine's channel is quiet, and nothing runs until you ask for something.
- What it means
- The machine suspended after a stretch with no work on it. Its state is saved, and it resumes when work arrives. This is the machine behaving as designed, not a failure.
- What to do
- Give it work. Type a command in the terminal, or ask an agent in the channel for something that needs the shell. The machine resumes with the state it suspended on: your processes, your files and your open shell are as you left them. If a command gets no response at all, read the channel — a machine that cannot resume posts there — and tell an admin the machine name.
- If you do nothing
- The machine stays suspended and its state is kept. Minutes or weeks later, the first command wakes it to the same state, so there is no window you have to catch.
The message, in plain language
The Notifications panel — Inbox, at the bottom of the channel list, opens it — writes this one in the product’s own voice:
dev-server-01 idled out, so we banked the coals.
Read it as: the machine suspended after idling, its state is saved, and it resumes when work arrives.
Why this is not the same as losing the link
Two machine states read alike in a notification and mean opposite things.
Suspended is deliberate. Nothing dropped, nothing is retrying, and the machine is waiting for a reason to wake. It costs you a moment on the next command and nothing else.
A dropped link is not deliberate: Crusible is retrying and the outcome is not settled yet. That one has its own page, A machine shows “Lost the link” . If the notification says retrying, you are reading the other case.
Waking it
Anything that needs the machine wakes it. A command in the terminal is the most direct route. Asking an agent to take a task in that channel works the same way, because the agent needs the shell to do it.
The first command after a suspension takes a moment longer than the ones after it. That is the machine coming back, not a stall.
When to involve an admin
A machine that does not resume is a compute-provider matter, and compute providers live in the Account group of Settings, which an admin owns. Give them the machine name from the notification and the channel that wraps it.
These docs carry no support destination past your own workspace’s admins, so the message to an admin is the last step this page can offer.
Idling itself is not something to fix. A machine that suspends when nobody is using it is the arrangement that lets a workspace keep machines standing for work that arrives at odd hours.