Administrators can insert a disk into a running node as long as the server hardware supports this function. Replacing failed disks (hot swapping) or adding additional disks (hot plugging) is supported without a server reboot.

Swarm recognizes, formats, and mounts the disk as a new volume when inserting a new unformatted disk. The disk continues to function as a volume without data loss when inserting a Swarm-formatted disk into the same node or a different node. The volume remains retired if the formatted disk was previously retired.

No manual configuration or intervention is needed. Messages are displayed in logs and in the Swarm Admin Console to indicate a disk was inserted or removed.

Requirements for Hot Plugging

• The configuration option disk.volumes must be set to all.

• JBOD/pass-through mode must be supported and enabled to use a disk with a RAID controller. Contact DataCore support for details.

• Disks must not be configured in RAID.

• Any virtual machines housing Swarm storage nodes must enable disk UUIDs (set disk.EnableUUID=TRUE). 

Guidelines for Hot Swapping

• Count the total disks (status OK) across all node processes and verify it equals the value before pulling any disks to determine if hot swapping succeeded. The disk may not be assigned to the same node process handling it before being moved.

• Expect a disk that is plugged back in to least show up in the first node process on the machine if SNMP is slow to update. The disk adds algorithm attempts to keep the volume assignments balanced across node processes.

• Check the disk identification lights: The disk identification light is automatically enabled if a disk cannot mount when hot-plugged into a system or fails at boot time.

• Expect to see "noise" in the syslog about failed volume recovery (FVR) starting when pulling a live, good disk (not retired or disabled due to error count). Look for these announcements:

  • FVR has completed or been cancelled on the hot-swapped disk.

  • The hot-swapped volumeID is mounted and recognized by the assigned node process.
    
    

    Open