Recovering from a failed plex attach or synchronization operation

A plex attach operation requires that the plex be synchronized with the existing plexes in the volume. Other operations such as creating a mirror for a volume also require plex synchronization. The synchronization of the plex can be a long-running operation, depending on the size of the volume and the amount of data that needs to be synchronized.

The recovery behavior in this release has changed from previous releases of VxVM. Previously, if the plex synchronization was interrupted, VxVM dissociated the plex from the volume during volume recovery. The plex synchronization needed to be restarted manually. This behavior remains if the disk group version is less than 170.

When the disk group version is 170 and the plex resynchronization is triggered by the vxplex att, vxassist mirror, vxsnap addmir, or vxsnap reattach commands, the plex remains associated with the volume during volume recovery. VxVM detects that the synchronization was interrupted, and resumes the synchronization. If the volume has an associated DCO (version 20 or later), VxVM tracks the changes while the plex is synchronizing, If the synchronization fails due to a system crash or vxconfigd failure, VxVM resumes the synchronization from the point where it failed. The synchronization occurs in the background, so the volume is available without delay. If the volume does not have an associated DCO, but has a version 170 disk group, the synchronization restarts from the beginning.

When you create a volume and add a mirror in one operation (vxassist make nmirror=2), the synchronization of the mirror is not tracked for automatic recovery. To ensure that the synchronization resumes from the point where it failed, create the volume first and then add the mirror with the vxassist mirror command.

VxVM automatically recovers volumes in some cases. If you need to recover the volumes manually, the vxrecover command triggers the synchronization for any plexes that have a failed synchronization process. These plexes have a state of TEMP, TEMPRM, or TEMPRMSD.

In the CVM environment, if a master node crashes while a plex synchronization is in progress, the new master restarts the synchronization from the point where the master node failed, after the master recovers. The disk group must be version 170 and the volume must have a version 20 DCO. If the disk group is version 170 but no DCO is attached, the synchronization restarts from the beginning.

You can abort a plex attach operation or synchronization operation with Ctrl-C or the vxtask abort command. In this case, VxVM dissociates the plex from the volume.