Ha Storage Configuration - HP StorageWorks 8000 - NAS Overview

HA Storage Configuration

This section provides suggestions for organizing your storage for optimizing HA features.
On a direct-attach NAS system without HA features, the simplest way to configure the storage
is to create a single LUN for the entire array and create one volume group with multiple file
volumes beneath it. However, in an HA cluster with two NAS servers, each volume group can
be active on only one NAS server, or
failover packages (see "Failover Packages" on page 8), you must create at least two volume
groups so each node can access some of your storage.
In addition, with SAN configurations storage generally resides on several different storage
devices within the SAN. LUNs are created on each array using the SAN management
software. To ensure that a volume group does not become unavailable due to hardware
problems on a specific storage device, the safest configuration is to create a separate volume
group for each LUN.
Note When a volume group includes LUNs from multiple arrays, if any one array has
a hardware failure, the entire volume group will be unavailable. By creating a
separate volume group for each LUN, you are ensuring that only that portion of
the storage residing on the failed array will be unavailable until the array is
restored. All other storage will remain unaffected.
When determining how many LUNs and volume groups to create, consider the following:
A volume group can be active on only one node at a time.
I
A volume group can be assigned to only one failover package.
I
File volumes can't be split between packages or nodes.
I
Multiple volume groups can be assigned to the same package.
I
A package can only run on one node at a time.
I
The HA cluster is limited to 30 packages total (both nodes combined).
I
A package is the smallest unit of failover.
I
A volume group is the smallest unit of storage that can be failed over.
I
If a package contains multiple volume groups and a single volume group, file volume or
I
even share fails, the entire package will be failed over.
More packages means a higher degree of granularity to balance the packages between the
I
nodes and detect failures.
More packages also results in higher complexity in the cluster configuration and requires
I
more virtual IP addresses to keep track of.
The packages should be divided between the cluster nodes such that the storage access
I
load (from client systems) will be balanced. This may or may not correspond to balanced
storage capacity between the nodes.
The illustrations on the following pages depict three common storage configurations:
One LUN and one volume group (for active/passive failover).
I
Multiple LUNs, volume groups and packages (most common for SAN configurations).
I
node , at a time. Therefore, if you are using active/active
Network and Storage Planning 9