Lifecycle
create -> running -> suspended -> running -> ... -> removed| Action | What happens | E2B SDK | CLI |
|---|---|---|---|
| Create | Starts a new machine. | Sandbox.create() |
rookbox new |
| Suspend | Saves memory and state and releases resources. | beta_pause() |
rookbox pause |
| Resume | Comes back exactly where it stopped. | Sandbox.connect(id) |
rookbox resume |
| Remove | Deletes the machine and its disk. | kill() |
rookbox kill |
| Fork | Copies the disk, not memory. | Dashboard only |
In the E2B SDK, pause is Rookbox suspend and resume is restore. A suspended machine uses no CPU or memory.
Timeout
Section titled “Timeout”Every machine created through the SDK has a timeout.
sandbox = Sandbox.create(timeout=15 * 60) # secondssandbox.set_timeout(30 * 60) # extendconst sandbox = await Sandbox.create({ timeoutMs: 15 * 60_000 });await sandbox.setTimeout(30 * 60_000);- Without
timeout, the beta default applies: 30 minutes. - Minimum 60 s, maximum 24 h. Values outside that are adjusted.
- When the timeout ends, the machine is removed. Auto-suspend is off unless you ask for it.
- To ask for auto-suspend at the end of the timeout, use
Sandbox.beta_create(auto_pause=True)in Python.
If you do not need the state, let the timeout clean up. If you do, call beta_pause() before the timeout ends.
The Dashboard has Fork on each machine. It creates a new machine from the original’s disk. Memory does not come along: running processes do not continue. Fork is not available through the SDK.