HP CloudSystem Matrix How-To Guide: ESXi Cluster Provisioning

36
Appendix 4: Configuring storage in Matrix
Before Matrix Operating Environment infrastructure orchestration can allocate storage, it must be configured in
logical server storage pools in Matrix OE visualization. A logical server storage pool entry (SPE) represents storage
that has been carved from a storage array (for example, HP Enterprise Virtual Array or HP 3PAR), presented to a
host-initiator port, and zoned in a SAN fabric on the SAN switches for connectivity with the blade. Creating an SPE
is detailed in the HP Matrix Operating Environment 7.0 Logical Server Management User Guide at
www.hp.com/go/matrixoe/docs.
How an SPE is defined determines how IO uses the storage for provisioning. Storage for physical provisioning can
be defined in one of four ways:
 Predefined single-volume SPEs
 Predefined multivolume SPEs
 Predefined SPM-managed static SPEs
 SPM-managed autocarved SPEs
Predefined single-volume storage pool entries
A single-volume SPE defines a single volume that is presented to one (or more for shared storage) host initiator port
WWNs. Adding multiple single-volume-SPEs to a blade requires N_Port ID virtualization (NPIV) support in the Fibre
Channel (FC) host bust adapter (HBA), allowing multiple WWNs on each host port. IO can use any predefined SPE
that meets the required criteria (Virtual Connect Domain Group, size, RAID level, or tags) for VM cluster provisioning
or disk add operations on the cluster.
Adding multiple SPEs to an FC port in Virtual Connect requires that the blade first be deactivated to add WWNs to
the NPIV configuration. This operation is disruptive to the VM host and consequently the cluster. For VM clusters, all
VMs must first be powered down before a new SPE can be added to the cluster.
Adding storage to an IO VM Cluster service by adding a predefined single-volume SPE requires taking the following
steps:
 Verify that an SPE has been defined and is available to be used. The request will pause and wait for this to
be put in place if it is missing.
 Power down all VMs running on the cluster.
NOTE: This can be mitigated as follows:
a. Add an external VM host to the cluster.
b. Migrate the VMs to the externally added VM host.
c. When the operation is complete, move the VMs back, and then remove the external host from the cluster.
 Perform the IO disk add.
 Power on VMs.
The disk add operation will power down all blades known to be in the IO defined VM cluster, add the disk to each
host, and then power on the blades. The 10_ConfigureVMHostDataStore_group_end OO workflow attached to the
Add Data Disk End actions causes the VM host to rescan the SCSI bus and creates a data store for the volume in
vCenter.
Predefined multivolume storage pool entries
A predefined SPE can be defined with more than one volume. Each volume in the SPE is manually designated as
"Ready." After a volume is marked as "Ready," it is assumed to be properly carved from the SAN storage array,
presented to a host-initiator port, and zoned in a SAN fabric on the SAN switches for connectivity with the blade.
Because adding a volume does not introduce a new SPE to the blade, no Virtual Connect profile deactivation is
required (or the associated reboot). After the storage becomes available to the blade, the OS can rescan the SCSI
bus and begin using the new volumes.