Maintenance and migrations
A migration occurs when a Compute Instance is moved from one physical host machine to another. This can happen within the same data center or across data centers. Migrations across data centers can be configured by customers in Cloud Manager. All other types of migrations must be configured by a Linode administrator.
Types of migrations
There are three types of migrations that can occur: cold, live, or warm.
Cold migrations
During a cold migration, your Compute Instance is automatically shut down (if it's not powered off already), moved to another physical host machine, and then returned to its previous state (booted or powered down) once it has fully migrated.
- Advantages: Can occur within the same data center and across data centers, generally the most available type of migration option
- Disadvantages: Requires the most downtime of all migration types
Live migrations
During a live migration, your Compute Instance remains up and running while it is moved from one host to another within the same data center.
- Advantages: No downtime
- Disadvantages: Not always an option depending on the scenario (i.e. host hardware compatibility, plan resizing, etc.)
Live migrations are not supported for Compute Instances in distributed compute regions.
Warm migrations
During a warm migration, your Compute Instance remains up and running while it is synced from one host to another within the same data center. Once synced, the Compute Instance is automatically powered down and is booted back up on the new host to complete the migration process.
- Advantages: Significantly less downtime than a cold migration, alternative to live migrations if they are not available
- Disadvantages: Requires a reboot to complete the process
When migrations occur
Compute Instances can be migrated between hosts within the same data center and across data centers, but cannot be migrated between core and distributed compute regions. Some of the reason for these migrations are detailed below. The type of migration that occurs (warm, live, or cold) depends on the situation as well as the host hardware on the origin and destination machine.
Within the same data center:
- Scheduled or emergency maintenance
- By customer request via Support ticket while troubleshooting performance issues
- During the resize process
Migrations within the same data center can be achieved via any migration type (cold, live, and warm).
Across data centers:
- Customer initiated for personal or business reasons (see our guide on Initiating a cross data center migration)
- By customer request via Support ticket
- When an old data center is retired. If this is the case, you will be notified ahead of any anticipated migrations or maintenance.
Migrations across data centers can only be achieved using cold migrations.
Troubleshooting migrations and FAQs
What if my cold migration fails?
Should a cold migration fail for any reason, the Support team is notified and will configure a new cold migration.
Why would a live migration not be available?
In order for a live migration to occur, a host compatible with the host your Compute Instance currently resides on must be available. Not all host hardware is alike or available to accept a live migration, so live migrations are not always a viable option.
What could cause a warm migration to fail?
- Your Compute Instance is not configured to respect ACPI shutdowns:
- This applies to warm migrations initiated by customers, including warm migrations taking place during the warm resize process.
- This also applies if you are running a custom distribution (i.e. Windows) or unsupported disk image that is not configured for ACPI shutdowns.
- If a Linode administrator cancels the warm migration.
- If the sources Compute Instance stops responding, is shut down before the cross-host sync, or if the process is disturbed in any way.
What should I do if my warm migration fails?
- If your warm migration fails to complete after an early initiation is triggered via Cloud Manager, the Support team is notified. There are two options that are used at the discretion of the Support team.
- A new warm migration can be configured to be performed at a scheduled time.
- A cold migration can be configured. You can then initiate the cold migration via Cloud Manager or allow it to proceed as otherwise scheduled.
What should I do if my warm resize fails?
- If a warm resize fails, you can either proceed with a cold resize (recommended) or troubleshoot your Compute Instance's configuration to respect ACPI shutdowns.
What if my cross data center migration fails?
- Should a cross data center migration fail, you may reattempt the migration. If it fails again, open a Support ticket for further assistance.
Next steps
If you want to learn more about migrations or need additional troubleshooting tips, see the following resources for additional information:
Updated about 23 hours ago