Technical Considerations for a Serviceguard Cluster that Spans Multiple IP Subnets, July 2009

remoteport=4200

nodes=rac1vip1.hp.com:4200, rac1vip2.hp.com:4200, \

rac2vip1.hp.com:4200, rac2vip2.hp.com:4200

The database nodes publish FAN events which are received by subscribers who then act on it. The

easiest way to use (subscribe to) FAN events is to use an integrated Oracle client with your

application. These are some of the integrated clients: OCI Session Pools, Connection Manager

(CMAN) session pools, JDBC connection pools and Oracle NET Listener.

TAF and FCF provide fast failover of database connections by allowing you to configure FAN-

integrated JDBC clients to automatically subscribe to FAN HA events and react to service, instance,

and database UP and DOWN events. TAF and FCF are mutually exclusive.

The “Best Practices for Client Failover in Data Guard Configurations for Highly Available Oracle

Databases” white paper is a good source for additional information on the topic of client failover to

Oracle databases, available at:

http://www.oracle.com/technology/deploy/availability/pdf/MAA_WP_10gR2_ClientFailoverBestPr

actices.pdf.

Several other useful documents are available at Oracle’s Maximum Availability Architecture website:

http://www.oracle.com/technology/deploy/availability/htdocs/maa.htm.

Applications that do not support Oracle FAN can still be configured for efficient failover by using

timeouts and application retries. For those applications, Oracle recommends to reduce the TCP

keepalive timeout for the client application server to release database connections in the event of a

database node crash.

Note:

The TCP keepalive timeout is a system wide parameter. Setting it on a

host will affect all processes communicating over TCP on that host.

If FAN is properly configured on all participating systems in the SADTA SGeRAC cluster clients will

receive the same type of events from the Oracle sub-clusters in which the RAC database is currently

running, just like in a local SGeRAC cluster.

Oracle RAC specific network configuration

While it is required that the Serviceguard heartbeat and the user/package network – in the Oracle

RAC case this is the network with the Oracle VIP addresses – need to be fully routed between the two

sites, the Oracle clusterware heartbeat network and Oracle RAC cache fusion network do not need to

be routed between the sites. This implies that the Oracle clusterware heartbeat is configured on a

network different from the Serviceguard heartbeat network Please see the DTS and SGeRAC manual

for details.

NFS in a cross-subnet cluster

Network File System (NFS) enables systems to access files from any location on the network,

transparently. An NFS server makes a directory available to other hosts on the network by sharing the

directory. An NFS client accesses the shared directories on the NFS server by mounting the

directories. HP offers a Serviceguard NFS Toolkit which allows a group of HP-UX systems to act as a

cluster of NFS servers. The nodes in the Serviceguard cluster are connected to a shared pool of

storage, allowing any of the cluster nodes to act as the NFS server for a given shared file system.