How host maintenance works for Amazon EC2 Dedicated Hosts

When a degradation is detected on a Dedicated Host that is enabled for host maintenance, we automatically allocate a replacement Dedicated Host in your account. The replacement Dedicated Host receives a new host ID, but retains the same attributes as the original Dedicated Host, including:

Once the replacement host has been allocated, we migrate the instances using either live migration host maintenance or reboot-based host maintenance, depending on the instance.

Once the degraded host has no more running instances, it is permanently release from your account.

Instances that support live migration host maintenance

Instances that support live migration host maintenance are automatically migrated to the replacement host within 24 hours, without stopping and restarting them. The migrated instances retain their existing attributes, including:

After the supported instance types have been automatically migrated to the replacement host, we send you email and AWS Health Dashboard notifications that include IDs of the degraded and replacement hosts, information about the instances that were automatically migrated using live migration host maintenance, and information about the remaining instances.

Instances that require reboot-based host maintenance

Instances that require reboot-based host maintenance are scheduled for instance reboot scheduled events for 14 days from the date of the notification. You can continue to access your instances on the degraded Dedicated Host before the scheduled event.

Amazon EC2 automatically reserves capacity on the replacement host for these instances. You can't run instances in this reserved capacity.

At the date and time of the scheduled event, the instances are automatically stopped and restarted into the reserved capacity on the replacement host. The migrated instances retain their existing attributes, including:

However, since the instances are stopped and restarted during the migration, they do not retain their memory, CPU, and networking states.

You can also manually stop and restart these instances at any time before the scheduled event to migrate them to the replacement host or to a different host. You might need to modify your instance's host affinity to restart it on a different host. If you stop an instance before the scheduled event, the reserved capacity on the replacement host is released and becomes available for use.

Unsupported instances

Some instances can't be automatically migrated to the replacement host.

Instances with instance store-backed root volumes

For these instances, we schedule instance retirement events for 28 days from the date of the notification. At the date and time of the scheduled event, the instances are permanently terminated. We recommend that you manually launch replacement instances on the replacement host and then migrate the required data to the replacement instances before the scheduled event.

Instances with EBS-backed root volumes

For these instances, we schedule instance stop events for 28 days from the date of the notification. At the date and time of the scheduled event, the instances are stopped. We recommend that you manually stop on restart the instance on the replacement host or on a different host. You might need to modify your instance's host affinity to restart it on a different host.

You can continue to access your instances on the degraded Dedicated Host before the scheduled event.

Host maintenance states

When a host becomes degraded, it enters the permanent-failure state. You can't launch instances on a Dedicated Host that is in the permanent-failure state.

Once the replacement host is allocated, it remains in the pending state until the instances that support live migration host maintenance have been automatically migrated from the degraded host, and until the scheduled events have been scheduled for the remaining instances. After this has been completed, the replacement host enters the available state.

Once the replacement host enters the available state, you can use it in the same way that you use any host in your account. However, some instance capacity on the replacement host is reserved for the instances that require reboot-based host migration. You can't launch new instances into this reserved capacity.

When the degraded host has no more running instances, it enters the released, permanent-failure state, and it is permanently released from your account.