Planning for shared storage pools

Edit online

Shared storage pools allow a set of SAN volumes on one or more supported storage controllers to be managed as a clustered storage device from any Virtual I/O Server that is in the cluster. Those Virtual I/O Servers share access to aggregated logical volumes and present this aggregated storage space as a single pool of storage, optionally divided into separate tiers for Quality of Service (QoS) needs. PowerVC can manage and allocate storage volumes from this pool as it can from Fibre Channel SAN providers.

A cluster consists of up to 24 Virtual I/O Servers with a shared storage pool that provides distributed storage access to the Virtual I/O Servers in the cluster.

For more background and setup information about shared storage pools, see Shared storage pools in the POWER8® Knowledge Center.

Before using shared storage pools

To use a shared storage pool with PowerVC, you must create it outside of PowerVC. The shared storage pool provider is automatically added to PowerVC when a host with a participating VIOS is registered, or, if the host is already registered, the next time it is synchronized with PowerVC. There is typically a 5 minute duty cycle.

To learn how to create a shared storage pool, see Creating shared storage pools.

Shared storage pools on HMC and NovaLink hosts

PowerVC supports shared storage pools for hosts that are managed by PowerVM® NovaLink or Hardware Management Console (HMC). The location where the shared storage provider service runs is determined by the first host that is brought under PowerVC management that has a VIOS that is participating in the shared storage pool cluster.
  • If the first host is NovaLink managed, then the shared storage pool storage provider service (the Cinder block volume driver process) initially runs on the NovaLink partition's operating system. The running service, the logs, and the management API that the service talks to are on that partition's operating system.
  • If the first host is HMC managed, then the Cinder storage provider service initially runs on the PowerVC management server and communicates with that host’s HMC.

Shared storage pool manager

When a shared storage pool is managed by PowerVC, one entity (host or HMC) is selected to provide the management API to service the shared storage pool. This entity is considered the manager of the shared storage pool, and its location determines where the volume driver runs. If the manager is an HMC, then the volume driver runs on the PowerVC management server. If the manager is NovaLink, then the volume driver also runs on the NovaLink host.

If the shared storage pool manager is unmanaged from PowerVC, then PowerVC tries to select a new one. If there are no more managed hosts with Virtual I/O Servers that are participating in the shared storage pool cluster, then the shared storage pool storage provider is removed from PowerVC. Additionally, the administrator can choose a new shared storage pool manager and depending on where this is, the volume driver location might change accordingly.

To change the manager of a shared storage pool, edit the provider from the storage providers page in the user interface. The system that is currently managing the provider is listed on the shared storage pool provider's details page.

Adding and removing shared storage pool providers

To change the manager of a shared storage pool, edit the provider from the storage providers page in the user interface. The system that is currently managing the provider is listed on the shared storage pool provider's details page.

If a shared storage pool is removed outside of PowerVC, for example, if the VIOS cluster is deleted, it might take up to 10 min. for the host to report it gone and for PowerVC to remove it. Maintenance tasks, such as adding a physical volume, backup, and so on, must also be done outside of PowerVC.
Notes:
  • If a shared storage pool goes into ERROR state due to any out of band operations (For example, any changes in HMC credentials), on the Storage provider page, select the shared storage pool, click Edit Connection and then click Save to bring back the shared storage pool to OK state.
  • If the account gets locked due to maximum number of retries (this can be checked by logging into VIOS), unlock the account and then use Edit Connection for the shared storage pool from the PowerVC GUI to bring back the shared storage pool to OK state.

Storage connectivity groups

A storage connectivity group is automatically created for the shared storage pool provider when the provider is added to PowerVC. This becomes the default storage connectivity group for the shared storage pool. It contains Virtual I/O Server members of the associated cluster that are also on hosts managed by PowerVC. This means that if two Virtual I/O Servers on a host are members of the cluster and a volume is attached to a virtual machine, it will have storage paths through each of the Virtual I/O Servers.

The default storage connectivity group also allows both shared storage pool volumes and NPIV-attached volumes to be attached as data volumes to virtual machines when a virtual machine is booting from a shared storage pool volume. Custom storage connectivity groups can be created for the shared storage pool that remove VIOS members (for less multipathing) and that disallow NPIV data volumes to be attached to virtual machines deployed with that storage connectivity group.

User Tiers

PowerVC can only use the default tier of storage in the shared storage pool. PowerVC clones volumes to the default tier within a shared storage pool.

Thick and thin images

Storage templates can be created for the shared storage pool that specify either thick or thin provisioned volumes. The automatically created storage template that is set as the initial default specifies thin volumes. These do not consume space in the shared storage pool until data is written to them. A thickly provisioned volume immediately consumes its capacity from the free space in the pool.

When deploying a thick image backed by a shared storage pool, the Virtual I/O Server makes a full copy of the image. The location of the copy depends on the workload on the Virtual I/O Server and how your environment is set up. The time it takes to make this full copy depends on the speed of the backing Fibre Channel device, the speed of the Fibre Channel network, the resources allocated to the Virtual I/O Server and the current workload on the Virtual I/O Server. This copy operation has lower priority on the Virtual I/O Server than I/O requests. Therefore the copy operation could take up to one or more minutes per GB to complete on a heavily loaded and undersized Virtual I/O Server.