File system management

Once vdisk sets are defined and created, the mmvdisk filesystem create command will take the vdisk sets and run the core IBM Storage Scale mmcrfs command with an appropriate vdisk NSD stanza file and appropriate options based on the vdisk set attributes.

Custom mmcrfs options unrelated to the vdisk set attributes may be supplied at file system creation using the mmvdisk filesystem create --mmcrfs option. All command line parameters after --mmcrfs are passed to the IBM Storage Scale mmcrfs command.
  • The mmcrfs option -j scatter is always used since -j cluster offers no benefit with vdisk NSDs. If -j cluster is used with the --mmcrfs option, it is considered an error.
  • The mmcrfs default replication settings for both data (-r) and metadata (-m) are set to 1.
  • The mmcrfs maximum replication settings for both data (-R) and metadata (-M) are set to 2.
If custom default or maximum replication settings are used with the --mmcrfs option, mmvdisk will notice them and make sure that the file system failure groups for the vdisk NSDs can be set appropriately. For example, if a default replication of 3 is used in a file system where the vdisk set has only two members, it will be considered an error.

The mmvdisk filesystem command uses failure group numbers 1 and 2 when the maximum replication is 2, and failure group numbers 1, 2, and 3 when the maximum replication is 3.

Node classes represent the single point of failure for vdisk set file system NSD failure groups. If the vdisk sets of the file system span sufficient node classes to match or exceed the maximum replication, all the vdisk NSDs in a node class will be assigned the same failure group.

If the vdisk sets do not span sufficient node classes to match the maximum replication, failure groups will be assigned based on log groups within a scale-out or shared recovery group, or based on the individual recovery groups in a recovery group pair. While this does not offer true single point of failure protection, it does allow the file system to replicate, and if recovery groups in sufficient node classes are added to the file system later, the mmvdisk filesystem change command can be used to reassign the failure groups based properly on the node classes.

If a file system has sufficient node classes to match or exceed the maximum replication, custom failure group numbers can be assigned to the file system's vdisk NSDs on a node class basis. This can be useful if multiple recovery groups share a single point of failure that mmvdisk is unaware of, such as being in the same frame or on the same network switch.

For example, if node classes NC01 and NC02 are in one frame, and node classes NC03 and NC04 are in another, the vdisk NSDs in NC01 and NC02 can be assigned the failure group number 12 and the vdisk NSDs in NC03 and NC04 can be assigned the failure group number 34:

mmvdisk filesystem create --file-system fs1 \
                            --vdisk-set TwoFramesFourNCs \
                            --failure-groups NC01=12,NC02=12,NC03=34,NC04=34
In this example, the maximum replication is 2 for both data and metadata, and two custom failure group numbers are used.

For more information and restrictions on custom failure group assignments, see the mmvdisk filesystem command.