LVM Online Disk Replacement (LVM OLR)

Replacing an LVM disk in a Serviceguard cluster volume group

Replacing LVM disks in volume groups in a Serviceguard cluster follows a similar

procedure to the one described earlier except that if the volume group is shared, the

commands to attach and detach the disks and paths must be executed individually on each

of the cluster nodes sharing the volume group.

Special care is required when performing a Serviceguard rolling upgrade. Disks that are

part of a shared volume group on a Serviceguard cluster should not be replaced online until

after all the nodes on the cluster have the LVM Online Disk Replacement patches installed.

The LVM OLR feature provides new cluster broadcast messages that are sent from the

cluster server to the client nodes when disks are reattached so that all the nodes recognize

that the replaced PV has stale user data on it. Any nodes not updated with these patches

will crash when the replaced disk is recovered by the cluster server. This is necessary to

prevent the potential data corruption that could result due to the nodes in the cluster having

inconsistent maps of stale or fresh data on the replaced disk. See the section entitled

Determining whether LVM OLR is available on the system for a description of how to

ensure the LVM OLR is present on each cluster node.