Changing GPFS disk states and parameters

You might find it necessary to change a disk's state if there is some indication of disk failure or if you need to restripe the file system.

Refer to Displaying GPFS disk states for a detailed description of disk states. You can change both the availability and status of a disk by using the mmchdisk command:

Change disk availability by using the mmchdisk command and the stop and start options.
Change disk status by using the mmchdisk command and the suspend and resume options.

Issue the mmchdisk command with one of the following four options to change disk state:

resume

Informs GPFS that a disk previously suspended is now available for allocating new space. Resume a disk only when you suspended it and decided not to delete or replace it. If the disk is in a stopped state, it remains stopped until you specify the start option. Otherwise, normal read and write access to the disk resumes.

start

Informs GPFS that a disk previously stopped is now accessible. GPFS does this by first changing the disk availability from down to recovering. The file system metadata is then scanned and any missing updates (replicated data that was changed while the disk was down) are repaired. If this operation is successful, the availability is then changed to up.

If the metadata scan fails, availability is set to unrecovered. This can occur when other disks remain in recovering or an I/O error occurred. Repair all disks and paths to disks. It is recommended to run mmfsck command. For more information, see mmfsck command. The metadata scan can be reinitiated later by issuing the mmchdisk start command again.

If more than one disk in the file system is down, they should all be started at the same time by using the -a option. If you start them separately and metadata is stored on any disk that remains down, the mmchdisk start command fails.

stop

Instructs GPFS to stop any attempts to access the specified disk. Use this option to inform GPFS that a disk failed or is inaccessible because of maintenance. A disk's availability remains down until it is explicitly started with the start option.

suspend

or

empty

Instructs GPFS to stop allocating space on the specified disk. Place a disk in this state before disk deletion or replacement. This is a user-initiated state that GPFS is never used without an explicit command to change disk state.

Note: A disk remains suspended until it is explicitly resumed. Restarting GPFS or rebooting nodes does not restore normal access to a suspended disk.

The empty option is similar to the suspend option. In GPFS 4.1.1 and earlier, the output of the mmlsdisk command displays the status as suspended, as shown in the following example.

For example, to suspend the hd8vsdn100 disk in the file system fs1, enter:

mmchdisk fs1 suspend -d hd8vsdn100

To confirm the change, enter:

mmlsdisk fs1 -d hd8vsdn100

The system displays information similar to:

disk         driver   sector failure holds    holds                            storage
name         type       size   group metadata data  status        availability pool
------------ -------- ------ ------- -------- ----- ------------- ------------ ------------
hd8vsdn100   nsd         512       7 yes      yes   suspended     up           system

For GPFS 4.1.1 and later, the status in the mmlsdisk command is displayed as to be emptied, as shown in the following example:

For example, to set

to be
emptied

state for gpfs1nsd disk of the file system fs1, enter:

mmchdisk fs1 empty -d gpfs1nsd

To confirm the change, enter:

mmlsdisk fs1 -d gpfs1nsd

The system displays information similar to:

disk      driver sector failure holds    holds                                    storage
name      type    size   group metadata  data  status        availability disk id pool     remarks
--------- ------ ------ ------ --------  ----  ------------- ------------ ------- -------- ---------
gpfs1nsd  nsd     512      -1   Yes      Yes   to be emptied   up             1   system
gpfs2nsd  nsd     512      -1   Yes      Yes   to be emptied   up             2   system    desc

You can also use the mmchdisk command with the change option to change the Disk Usage and Failure Group parameters for one or more disks in a GPFS file system. This can be useful in situations where, for example, a file system that contains only RAID disks is being upgraded to add conventional disks that are better suited to storing metadata. After adding the disks by using the mmadddisk command, the metadata that is stored on the RAID disks must be moved to the new disks to achieve the desired performance improvement. To accomplish this, first the mmchdisk change command would be issued to change the Disk Usage parameter for the RAID disks to dataOnly. Then, the mmrestripefs command would be used to restripe the metadata off the RAID device and onto the conventional disks.

For example, to specify that metadata should no longer be stored on disk hd8vsdn100, enter:

mmchdisk fs1 change -d "hd8vsdn100:::dataOnly"

To confirm the change, enter:

mmlsdisk fs1 -d hd8vsdn100

The system displays information similar to:


disk         driver   sector failure holds    holds                            storage
name         type       size   group metadata data  status        availability pool
------------ -------- ------ ------- -------- ----- ------------- ------------ ------------
hd8vsdn100   nsd         512       1 no       yes   ready         up           sp1

For more information, see the mmchdisk command and the mmlsdisk command.