IBM Books

Administration Guide

Chapter 21. High Availability Cluster Multi-Processing (HACMP) on AIX

DB2 UDB provides high availability failover support on many platforms. On AIX, DB2 UDB supports failover through the capabilities of IBM High Availability Cluster Multi-Processing (HACMP). Failover capability allows for the automatic transfer of workload from one processor to another should there be a hardware failure.

HACMP provides increased availability through clusters of processors which share resources such as disks or network access. If one processor fails then another in the cluster can substitute for the failed one.
Note:Do not use a "kill -9" against the db2start process in a high availability environment. This action is not recommended in any environment, but in particular such an action may invalidate failover recovery in your high availability environment.

There are three modes of failover support provided, a brief description of each mode and its application to DB2 follows. In each case we use the simple scenario of a two processor HACMP cluster.

Hot Standby
One processor is being actively used to run your DB2 instance and the second is in standby mode ready to take over the instance if there is an operating system or hardware failure involving the first processor.

Mutual Takeover
Both processors are either used to run separate DB2 instances, or one is use to run a DB2 instance while the other is used to run DB2 applications. If there is an operating system or hardware failure on one of the processors, the other processor takes over the tasks of the failing processor. Once the failover is complete, the remaining processor is doing the work of both processors.

Concurrent Access
Multiple processors can be used to scale to a single database instance using the DB2 Universal Database Extended Enterprise Edition product. This is done using a shared-nothing model and partitioning the data such that one or more partitions are running on each processor in the cluster. If an operating system or hardware failure occurs on one of the processors, then the other processor will take over the partitions of the failing processor. DB2 UDB Extended Enterprise Edition does not require the use of a Concurrent Resource Manager to provide redundancy. DB2 co-exists with the Concurrent Resource Manager, but does not require its capability. Redundancy is managed by using the previous two modes. The capabilities of this mode are only required by database managers with a shared architecture.

Each of the above configurations can be used to failover one or more partitions of a partitioned database. In addition, each can failover a complete instance of a single partition installation.

[ Top of Page | Previous Page | Next Page | Table of Contents | Index ]

[ DB2 List of Books | Search the DB2 Books ]