Note
To support hot plugging, Swarm requires control of all of the volumes (
disk.volumes = all).- If the Health Processor is actively scanning a drive when it is removed, I/O errors are recorded in the log. These errors are expected and do not indicate a problem.
- When a drive is removed, volume recovery (FVR) and erasure coding recovery (ECR) are both triggered, which includes creating new replicas or erasure set segments for objects that were stored on that drive.
- If you insert that drive in the same node or in a different node in the cluster, both recovery processes stop. There is a temporary state of over-replication because the returned volume has replicas or segments that were already recreated elsewhere. In time, the excess replicas or segments are deleted.
- If you insert a non-Swarm drive in a node, Swarm recognizes, formats, and mounts the drive as a new volume.
- If you insert a Swarm-formatted drive, either into the same node or into a different node, it continues to function as a volume without loss of data.
- If you insert a Swarm-formatted drive that was previously retired, the volume remains retired. No manual configuration or intervention is required.
- Messages display in logs and in the Swarm Admin Console to indicate that a drive was inserted or removed.
- If you are inserting multiple drives into the same server chassis, you need to wait 2 minutes between drive insertions to ensure the new drives are evenly distributed across the multiple nodes that may be running on the chassis.
Caution
When adding or relocating volumes to a node, ensure that the node has enough RAM to handle them. If not, it might be unable to mount some of the volumes.
Warning
Do not move Swarm drives between disk array controller types after they have been formatted by Swarm. Each controller reports available drive space to Swarm that is matched with the controller. For example, many controllers claim the last section of the drive, reducing the total available space. If you switch your drives with another controller, the new controller may claim additional drive space that is not reported to Swarm, so Swarm may try to write data to non-existing space, generating I/O errors.