Understanding and Designing Serviceguard Disaster Recovery Architectures

Key points that set an Extended Distance Cluster apart from a “normal” Serviceguard cluster:

• It is an architecture where the cluster nodes and storage are split evenly across two different

independent locations. The main purpose is to protect against failure scenarios in which one

of the two locations fail entirely—as opposed to an individual node or disk failure.

• There is no minimum distance between the locations. It can be in two adjacent rooms with

dedicated power sources and separated by a “fire wall.”

• The maximum distance is dependent on the volume manager and number of nodes, but cannot

exceed 100 kilometers.

• The data replication is done by volume manager mirroring.

With VERITAS Cluster Volume Manager (CVM) or VERITAS Cluster File System (CFS), available in

some of the Serviceguard Storage Management Suite bundles, you can create Extended Cluster

configurations of 2, 4, 6, 8, 10, 12, 14 or 16 nodes, for distances of up to 100 kilometers. For

Extended Cluster for RAC (also known as EC RAC) you can create cluster configurations of 2, 4,

6, 8, 10, 12, 14 or 16 nodes, for distances of up to 10 kilometers, and for 2 or 4 nodes for

distances of up to 100 kilometers.

Using Mirrordisk/UX and Shared LVM with Serviceguard Extension for RAC (SGeRAC), you can

create a 2 or 4 node Extended Cluster for RAC for distances of up to 100 kilometers. Because

RAC uses the network for lock passing (Oracle’s Cache Fusion feature), link distance and latency

may impact application performance. Make sure that you involve Oracle to review the impact on

RAC performance.

NOTE: Ensure that the patches, PHSS_40783 (for 11i v2) or PHSS_40784 (for 11i v3) or later,

are installed on all nodes if LVM Mirroring is used with Serviceguard A.11.19.

Extended Distance Cluster configurations support two or three data center solutions using

Mirrordisk/UX or VERITAS VxVM mirroring and Fibre Channel disk arrays for data replication

between the data centers.

See Table 3 (page 49) for the maximum supported distances between data centers for Extended

Distance Cluster configurations on HP-UX 11i v1.

Table 3 Extended Clusters Support for HP-UX 11i v1

Notes for Extended Cluster SupportVolume ManagerRevisionProduct

Supports up to 16 nodes for distances up to 100 KM.LVMA.11.16Serviceguard

3.5 only. Supports up to 16 nodes for distances up to 100 KM.VxVM

Supports up to 2 nodes for distances up to 100 KM. Oracle 9.2 or 10gR2.SLVMA.11.16SGeRAC

3.5 only. Supports 2, 4, 6 or 8 nodes for distances up to 10 KM (more

than 4 nodes requires patch), or 2 nodes for distances up to 100 KM.

Oracle 9.2 only.

CVM

LVM and VxVM support in SGeRAC is the same as that provided by the equivalent Serviceguard

revision.

See Table 4 (page 49) for the maximum supported distances between data centers for Extended

Distance Cluster configurations on HP-UX 11i v2.

Table 4 Extended Clusters support for HP-UX 11i v2

Notes for Extended Cluster SupportVolume ManagerRevisionProduct

Supports up to 16 nodes for distances up to 100 KM.LVMA.11.16Serviceguard

3.5 only. Supports up to 16 nodes for distances up to 100 KM.VxVM

Supports up to 16 nodes for distances up to 100 KM.LVMA.11.17 –

A.11.19