Occasionally, it is necessary to transfer instances from one Compute node to another Compute node in the cloud for purposes like as maintenance, workload rebalancing, or replacing a malfunctioning node.
Performing maintenance on compute nodes: If there is a need to temporarily deactivate a Compute node, such as for hardware maintenance, repair, kernel upgrades, or software updates, it is possible to transfer the instances operating on that Compute node to another Compute node.
Malfunctioning computational unit: In the event of an impending failure of a Compute node, it is possible to transfer instances from the failing node to a functioning one in order to service or replace the failed node.
Malfunctioning compute nodes: If a Compute node has already had a failure, you have the option to evacuate the instances. To restore instances on another Compute node after a failure, you can recreate them using the same name, network addresses, and any other allocated resources that were assigned to the instances before to the failure.
Workload redistribution: To re-balance the workload, you have the option to transfer one or many instances to a different Compute node. One can consolidate instances on a Compute node to save power, relocate instances to a Compute node that is physically closer to other networked resources to decrease latency, or disperse instances across Compute nodes to prevent hot spots and enhance resiliency.
The admin configures all compute nodes to enable secure migration. Each Compute node necessitates a shared SSH key to provide users on each host access to other Compute nodes throughout the migration process.
Types of Migration
Cold migration refers to the process of moving a system or application from one physical or virtual environment to another without interrupting its operation or requiring any downtime.
Cold migration, also known as non-live migration, refers to the process of shutting down a currently active instance before transferring it from the source Compute node to the destination Compute node.
Cold migration necessitates a certain amount of downtime for the instance. The moved instance retains access to the identical volumes and IP addresses.
In order to perform a cold migration, it is necessary for both the source and destination Compute nodes to be operational.
Live migration refers to the process of transferring a running virtual machine from one physical server to another without interrupting its operation. Live migration is the process of transferring an instance from one Compute node to another without the need to turn it off, while ensuring that the state remains consistent.
Live migration an instance typically results in minimal or negligible downtime. Live migration is not always possible in some scenarios. In order to perform live migration, it is necessary for both the source and destination Compute nodes to be in an operational state.
If you require to transfer instances due to the failure of the source Compute node, you can perform an evacuation of the instances.
Occasionally, the process of transferring instances may come with extra limitations. Migration constraints commonly occur when block migration, configuration disks, or instances accessing physical hardware on the Compute node are involved.
Both the source and destination Compute nodes must possess identical CPU architectures. For certain instances that utilize CPU host pass through, it is necessary for the CPU of both the source and destination Compute node to be identical.
The destination Compute node must possess an ample amount of accessible RAM. Memory over-subscription can result in migration failure.
The process of migrating instances that utilize disks stored locally on a Compute node is considerably more time-consuming compared to transferring volume-backed instances that utilize shared storage.
Posts
No comments:
Post a Comment