The instances in your environment access ephemeral disk storage. When you request that the Compute service start an instance, the service attaches ephemeral disk storage from the hypervisor storage or shared storage system. That ephemeral disk only exists as long as the instance runs. When you terminate the instance, the ephemeral storage reserved for that instance becomes available again on the hypervisor or shared storage subsystem for optimal performance.
As this storage is temporary, you should not rely on these disks to maintain long-term data or data that you must retain in case of failure, for example, stopping and starting an instance, hardware failure, or terminating an instance. For long-term data retention, use persistent storage. Ephemeral storage is useful for basic runtime requirements, such as storing an instance’s operating system.
You can augment your ephemeral storage with persistent storage by attaching a volume. These volumes can be detached and re-attached to instances when you need them, and can only be accessed through the instance they are attached to. See Working with Block Storage for more information on working with persistent storage.
Each instance must have at least one root disk that contains the bootloader and core operating system files. It may have optional ephemeral disks depending on the flavor selected when you created the instance. See Managing Flavors Using the Command-line Interface for more information.
In many environments, ephemeral disks are stored on the host’s local disks, for production environments it is recommended that hosts use a shared storage subsystem instead. A shared storage subsystem allows for instance migration between hosts, which is useful when you need to perform maintenance on a host system. Using a shared storage subsystem also allows you to recover instances when a host goes offline. You can evacuate the instance to another host and boot it up again. See Migrating Instances for more information.