test veritas logo



vxrecover - perform volume recovery operations


vxrecover [-bcElmnpsvV] [-d defaults_file] [-g diskgroup] [-o options] [-P count] [-t tasktag] [volume | volume_set | medianame...]
vxrecover -H


The vxrecover utility performs plex attach, RAID-5 subdisk recovery, and resynchronize operations for the specified volumes, for all volumes under the specified volume sets, or for volumes residing on the disks specified by medianame. If no medianame, volume or volume_set is specified, the operation applies to all volumes, or to all volumes in the specified disk group. If -s is specified, disabled volumes are started. With -s and -n, volumes are started, but no other recovery takes place.


-b Performs recovery operations in the background. With this option, vxrecover runs in the background to attach stale plexes and subdisks, and to resynchronize mirrored volumes and RAID-5 parity. If this is used with -s, volumes are started before recovery begins in the background.
-c Operates on cluster-sharable disk groups. This option is used by the cluster reconfiguration utility to effect recovery on volumes in shared disk groups after the failure of a node in the cluster. Do not use -c as a command-line option. Available only if the Veritas Volume Manager (VxVM) cluster feature is enabled and a license is installed.
-d defaults_file
  Specifies the absolute pathname of a defaults file to use instead of /etc/default/vxrecover. The defaults file can be used to define new values for the I/O region size and delay time when reattaching plexes. The following attributes may be defined in this file:
  Specifies the size (in standard length units as defined in vxintro(1M)) of the regions that are used to perform recovery operations. Specifying a larger size typically causes the operation to complete sooner, but with greater impact on other processes using the volume. The default I/O size is 2048 sectors (1 megabyte).
  Allows the impact on system performance of copy operations to be reduced. Plex recovery operations usually consist of a number of short operations on small regions of the volume. This attribute inserts a delay between the recovery of each such region. A specific delay is specified as a number of milliseconds. The default value is 250 milliseconds.
  Limits the number of concurrently active recovery tasks. The default maximum value is 10. This attribute can be used to control the amount of resources taken up by recovery tasks.
  Turns on distribution of recovery I/Os across nodes in the cluster. The distribute is valid only for cluster-sharable disk groups.
exclnodelist=nodename1[:nodename2:. . .]
  If the distribute keyword is specified with this keyword, it can be used to exclude certain cluster nodes from taking part in the recovery I/O distribution. The node names in this list should be in the form displayed in the output of the vxclustadm nidmap command.
The following settings are suggested for a clustered environment with a large number of mirrored volumes:

iosize=4m slow=1000 distribute
corresponding to an I/O size of 4 megabytes, and an I/O delay of 1000 milliseconds with recovery I/O distribution enabled.
-E Starts disabled volumes or plexes even when they are in the EMPTY state. This is useful for starting up volumes restored by the vxmake utility when specified along with the -s option.
-g diskgroup Limits operation of the command to the given disk group, as specified by disk group ID or disk group name. If no volume or medianame is specified, all disks in this disk group are recovered, otherwise, the volume and medianame are evaluated relative to the given disk group.
Without the -g option, if no volume and medianame is specified, all volumes in all imported disk groups are recovered; otherwise, the disk group for each medianame specified is determined based on name uniqueness within all disk groups.
-H Provides full help for the vxrecover utility.
-l Operates only on private disk groups. The auto-reattach daemon (vxattachd) uses this option to perform recovery on volumes in private disk groups after any detached disks are reattached to the disk group.
-m Recognizes and enables volumes that were moved as a result of a disk group move, split or join operation.
-n Does not perform any recovery operations. If used with -s, volumes are started, but no other actions are taken. If used with -p, vxrecover only prints a list of startable volumes.
-o options Passes the given option arguments to the -o options for the vxplex att and vxvol start operations generated by vxrecover. You can specify an option argument in the form

to restrict the set of commands that the -o option is applied to. Defined prefixes are:
o vol applies to all invocations of the vxvol utility (volume starts, mirror resynchronizations, RAID-5 partity rebuilds, and RAID-5 subdisk recoveries)
o plex applies to all invocations of the vxplex utility (used only for attaching plexes)
o attach applies specifically to plex attach operations
o start applies specifically to volume start operations
o recover applies to subdisk recoveries
o resync applies to mirror resynchronization and RAID-5 parity recovery.
The available options for -o include the following:
o distribute turns on distribution of recovery I/Os across nodes in the cluster. The distribute option is valid only for cluster-sharable disk groups.
o exclnodelist=nodename1[:nodename2:. . .] can be specified with the distribute option to exclude certain cluster nodes from taking part in the recovery I/O distribution. The node names in this list should be in the form displayed in the output of the vxclustadm nidmap command.
-p Prints the list of selected volumes that are startable. For each startable volume, a line is printed containing the volume name, the disk group ID of the volume, the volume’s usage type, and a list of state flags pertaining to mirrors of the volume. State flags include:
o kdetach indicates that one of the mirrors was detached by an I/O failure
o stale means that one of the mirrors needs recovery, but that the recovery is related to an administrative operation, not an I/O failure
o stopped prints if neither kdetach nor stale is appropriate for the volume.
-P If multiple volumes use the same underlying disk, recovery for such volumes is serialized by default. That is, the volumes are recovered one after the other. If the disk arrays are capable of handling parallel I/Os efficiently, use the -P option to specify the maximum number of volumes to be recovered in parallel.
-s Starts disabled volumes that are selected by the operation. Volumes are started before any other recovery actions are taken. Volumes are started with the -o delayrecover start option. This requests that any operations that can be delayed in starting a volume are delayed. That is, only those operations necessary to make a volume available for use occur. Other operations, such as mirror resynchronization, attaching of stale plexes and subdisks, and recovery of stale RAID-5 parity, are typically delayed.
-t tasktag If any tasks are registered to track the progress of the operation, marks them with the tag tasktag. This option is also passed to any utilities called to perform recoveries, so all tasks registered by any utility to perform the recoveries are also tagged with tasktag.
-v Displays information about each task started by vxrecover. For recovery operations (as opposed to start operations), prints a completion status when each task completes.
-V Displays each command executed by vxrecover. The ouput generated is more verbose than the job descriptions printed with the -v option.


To recover, in the background, any detached subdisks or plexes that resulted from replacement of a specified disk, use the command:

vxrecover -b medianame

To monitor the operations, use the command:

vxrecover -v medianame


/etc/default/vxplex Standard defaults file for vxplex that determines whether FastResync is used when attaching plexes. See vxplex(1M) for details.
  Standard defaults file for vxrecover that is used to set the I/O size and I/O delay values for recovery operations.


vxassist(1M), vxcache(1M), vxintro(1M), vxmake(1M), vxplex(1M), vxrelayout(1M), vxvol(1M), vxvset(1M)

VxVM 7.3.1 vxrecover(1M)