Top
PRIMECLUSTERGlobal Disk Services Configuration and AdministrationGuide 4.5
FUJITSU Software

D.9 sdxfix - Restore a failed object

SYNOPSIS

sdxfix -C -c class 
sdxfix -D -c class -d disk [-e online] [-x NoRdchk]
sdxfix -V -c class {-g group|-d disk} -v volume

DESCRIPTION

Use sdxfix to restore failed objects (excluding shadow objects). In such a case, you need to restore integrity from backup data or check by executing the "fsck" command. The "sdxfix" command can be executed with superuser access privileges only.


PRIMARY OPTIONS

You can use one of the following options.

-C

Class
Restores closed class to a normal status on the current node.

This command can restore class when the class includes a configuration database normally accessible and:

  • less than 3 disks in ENABLE status and 1 or more disks normally accessible

  • three to 5 disks in ENABLE status and 2 or more disks normally accessible

  • six or more disks in ENABLE status and 3 or more disks normally accessible

After restoration is complete, objects within the class are restored to the previous status. However, if class is a local class, volumes that were in STOP status before the class closure will be in ACTIVE status. Additionally, if class is a shared class, volumes that were in ACTIVE status before the class closure will be in STOP status.


-D

Disk
Restores the state of disk disk that detected an I/O error.

Reads all disk area, and if there's no problem, then clear the error state. Response from the command needs some time depending on the disk size because the entire disk area is read.
To restore the object status without reading the disk area, specify the sub option -x NoRdchk.

If there's a volume in the highest-level group to which disk belongs, or disk has a single volume, the volume should be stopped or inactive (STOP or INVALID) on all nodes (except when -e online is specified).

When disk is connected to a switch group, restoration fails. To clear an I/O error in a disk connected to a switch group, use the sdxswap -O command to make the disk exchangeable and then use the sdxswap -I command to make the disk useable.


-V

Volume
Restores a slice with invalid data (INVALID) or a not-ready slice (NOUSE) specified by a set of disk and volume or by a set of group and volume to the STOP status to restore the volume with invalid data (INVALID) to the STOP status.

The volume should be stopped or inactive (STOP or INVALID) on all nodes. The slice state specified in the combination of disk and volume, or group and volume should be INVALID or NOUSE.

Reads the entire slice specified in the combination of disk and volume, or group and volume, and if there's no problem, changes the state to STOP, and then, changes the state of stopped slice to INVALID.


SUB OPTIONS

You can use the following sub-options.

-c class

Specify a name of the class to which the object belongs.


-d disk (when using -D)

Specify a name of the disk.


-d disk (when using -V)

When volume is a mirror volume, specify a name of the disk that is connected to the mirror group to which the volume belongs. This disk should have the INVALID mirror slice that needs to be restored to STOP.

Specify a single disk name when volume is a single volume.

When volume is a switch volume, specify a disk name of the active disk connected to a switch group that includes the volume into disk. Do not set an inactive disk name to disk.


-e online (when using -D)

Restores the object even when the highest level group to which disk belongs or the single disk specified by disk includes an active volume.


-g group (when using -V)

When volume is a mirror volume, specify a name of lower-level group that is connected to the mirror group to which the volume belongs. This group should have the INVALID mirror slice that needs to be restored to STOP.

Specify a name of the highest-level stripe group when volume is a stripe volume.

Specify a name of the highest-level concatenation group when volume belongs to the highest-level concatenation group.


-v volume (when using -V)

Specify a name of the volume.


-x NoRdchk (when using -D)

Does not perform a read check in the disk area.

When a read check is not necessary, for example, if a disk enters the I/O error state because the path has been removed, you can shorten the recovery process time.


Use this option only when a read check is clearly unnecessary.

RETURNED VALUE

When it is normally terminated, "0" is returned. Otherwise, a non-zero value is returned.