Planning for disk fencing
Disk fencing is one of the features for quarantine policies that is available in PowerHA® SystemMirror®.
Disk fencing prerequisites
To use disk fencing in PowerHA SystemMirror, your disks
must meet the following requirements:
- All disks that are managed by the storage systems must be enabled for SCSI-3 Persistent Reservation (PR). Some storage systems do not enable SCSI-3 PR capabilities by default.
- All disks must not be in use when disk fencing is enabled (the volume groups must be offline).
- The disks must not be reserved before you start PowerHA SystemMirror. You can use the storage systems software to release any disk reservations.
Critical resource group settings
To use the disk fencing option, you must identify the critical resource group
and the storage subsystem must support SCSI-3 PR and ODM reserve_policy of PR_shared. This policy is
applied to all disks in the volume groups that are configured for all resource groups.
- Reservation type
- The disk must have a persistent reservation type of Write Exclusive All Registrant (WEAR). You
can run the following clmgr commands to check if your disk supports SCSI-3
capability (WEAR - type 7h) to support PowerHA SystemMirror disk
fencing:
where disk is the name of the disk and vg is the name of the volume group.clmgr -a SCSIPR_CAPABLE query physical_volume <disk> clmgr -a SCSIPR_CAPABLE query volume_group <vg>
Note: If the command does not return Yes, you cannot use the disk for disk fencing. You must choose some other disk or quarantine policy. - reserve_policy
- Defines whether a reservation method is running on the disk. The reserve_policy of PR_shared is the required policy that applies the shared host method for the disk. To view the attributes for the disk, run the lsattr -Rl <diskname> -a reserve_policy command.
To use the disk fencing feature, you must specify a critical resource group. The critical
resource group that you specify must meet the following criteria:
- The critical resource group cannot be added as a child in any parent_child, start_after, or stop_after, dependency relationship.
- The critical resource group must have all nodes in the cluster as participating nodes.
- The critical resource group must have the startup policy as Online On First Available Node.
The reserve_policy for a disk of a volume group that has cluster services that are running and
disk fencing is enabled, must use the following settings:
Option | Value |
---|---|
Configured Reserve Policy | PR_shared |
Effective Reserve Policy | PR_shared |
Reservation Status | SCSI PR reservation (Write Exclusive All Registrant) |
EMC storage and SCSI-3 PR
EMC disks do not support SCSI-3 PR capabilities by default. If you attempt to configure disk fencing on EMC storage without enabling SCSI-3 PR, an error occurs.
The following is an example of steps to enable SCSI-3 PR on some EMC storage devices:
Note: Before
you complete the following steps, each disk must not be in use on any nodes in the cluster and the
volume groups must not be in a varyon state. The following commands are part of the EMC software
packages that are installed on an AIX® LPAR.
- Identify the IDs for the storage devices and disks in the EMC storage subsystem by running the
following
command:
powermt display dev=hdiskpowerX Pseudo name=hdiskpowerX Symmetrix ID=000194900568 Logical device ID=0036 Device WWN=6000097000019490056853303030xxxx state=alive; policy=SymmOpt; queued-IOs=0
- Enable the SCSI-3 PR capabilities in the EMC storage subsystem by running the following
command:
symconfigure -sid 000194900568 -cmd "set device 0036 attribute=SCSI3_persist_reserv;" commit -v -noprompt
- Complete the following steps to rediscover the storage disks in the AIX operating system:
- Remove the storage devices or disks by running the
rmdev -R -d -l hdiskpowerX
command, wherehdiskpowerX
is the name of the disk. - Run the cfgmgr command to discover the storage devices or disks.
- Remove the storage devices or disks by running the
- Verify that the SCSI-3 PR capabilities are enabled for the storage devices or disks by running
the following
command:
/usr/symcli/bin/symdev -sid 000194900568 show 0036 | grep SCSI SCSI-3 Persistent Reserve: Enabled
For more information about configuring EMC for SCSI-3 PR, see the EMC Support website.
Hitachi storage and SCSI-3 PR
Hitachi disks do not support SCSI-3 PR capabilities by default. You must manually enable SCSI-3 PR capabilities for each disk that is part of a volume group that is managed by PowerHA SystemMirror. To enable the SCSI-3 PR capabilities, in the Hitachi Storage Navigator you must enable the Host Mode Options (HMHO) 2 and 72.