Instance Lifecycle¶

This page summarizes how an instance behaves over time.

Lifecycle at a Glance¶

An instance goes through these stages:

Created - You start a new instance from the dashboard.
Running - The instance is available for your work.
Paused - The instance is stopped to save resources (manual or automatic).
Resumed - You continue working from the same instance.
Deleted - The instance is removed when no longer needed.

Pause and Resume¶

You can pause and resume an instance from the instance card.

Pause stops the running workload and frees active compute resources.
Resume starts the same instance again so you can continue your work.

Pausing an instance kills all currently running processes.

Instances are automatically paused after not being accessed for a configured amount of time.

Training Hold¶

Use Training Hold when your instance must stay running for longer tasks. Training hold keeps a running instance active for the selected duration.

To set a training hold, click the clock icon on the instance card.

Deletion¶

Delete instances you no longer need.

Manual delete is always available to the owner (and admins).
Paused instances will also be deleted automatically after a configured retention period.

Deletion is final for data stored only inside that instance.

Data Persistence¶

Data in /workspace/shared is persistent and shared.

When an instance is paused and resumed, most temporary data inside the instance is preserved.

Typically preserved on pause/resume:

software installed with apt (and therefore packages installed through rosdep),
changes in user/system directories,
shell history (for example .bash_history).

This is because the platform keeps persistent storage for common system and user paths (for example /usr, /etc, /var, /root, /home, /workspace, /opt).

What is not preserved is the running process state itself: applications and jobs must be started again after resume.