Sample Configuration with HP Serviceguard Extension for RAG and Oracle Real Application Clusters 11g release 2 using Cluster File System
7
Virtual IP address (VIP) HA
This section describes high availability for VIP.
SG local LAN failover mechanism—preferred choice
For client public network HA in SGeRAC configurations, the preferred network HA method is to use
SG primary and standby. SG monitors the redundant network, so additional auto-port aggregation
(APA) software is not required. This mechanism is fastest to detect and recover from network failure. It
can use a single standby LAN as a standby for multiple primary LANs. It allows failover between
different speed network interface cards (NICs) in the same family. SG network sensors use NICs in the
remote node to monitor the health of NICs in the local node.
When the client network is configured with SG local LAN failover, SG performs the local LAN failover
and OC configures the VIP after the SG local LAN failover. Because OC performs monitoring and
manages the VIP address, client connectivity may be unavailable until OC detects the outage and
configures the VIP address on the local node.
Note: To know how to make VIP and Oracle Single Client Access Name (SCAN) highly available
using SG Local LAN failover, refer to article ID: 296874.1, “Configuring the HP-UX Operating System
for the Oracle 10g VIP” at
http://support.oracle.com
Local LAN failover using APA
(Oracle support account required). This article ID
has been updated for Oracle 11g R2.
When APA is used with bonded NICs, APA provides traffic distribution and load balancing capability
among multiple physical NICs or links. Load balancing is a desirable benefit for configurations where
a single interface is not sufficient to handle the network traffic. When a physical NIC or link fails, APA
provides HA by distributing traffic across the remaining NICs or links. APA monitors the NICs on each
system on which it is installed. One virtual link is presented to OC (APA network load balancing is
transparent to OC). APA requires use of the same type of NICs. Because all APA network connections
on a system go through the same switch, a switch failure means outage of the client network.
When APA is used in HA failover mode, a LAN Monitor is used. It provides HA network access with
an active link and a standby link. If a link fails, the LAN Monitor will automatically migrate the data
flow from the primary link to one of the standby links in the failover group. It monitors only the NICs
on the system on which it is installed. One virtual network link is presented to OC, and the physical
failover is transparent to OC because the same virtual network link remains available. The LAN
Monitor requires one standby NIC for each active NIC, and all NICs must be of the same type. SG
supports the LAN Monitor mode of APA.
Beginning with HP SG Storage Management Suite A.03.00, APA is supported for CFS interconnects.
Remote failover
OC fails-over VIP addresses to a surviving node when there is a catastrophic failure—for example, a
node failure or a network failure.
Network HA for cluster communication
SG, OC, and each RAC instance maintain communication with peers on other nodes. When
communication is broken due to a network partition or a node failure, each of these components must
reform its membership and eject non-members.
In CFS and CVM (5.0.1) configurations, GAB and LLT also use the cluster interconnect for peer-to-peer
communications.