Linux-HA and DRBD overview

High-Availability Linux (also called Linux-HA) provides the failover capabilities from a primary or active IBM® Netezza® host to a secondary or standby Netezza host. The main cluster management daemon in the Linux-HA solution is called Heartbeat. Heartbeat watches the hosts and manages the communication and status checks of services. Each service is a resource. Netezza groups the Netezza-specific services into the nps resource group. When Heartbeat detects problems that imply a host failure condition or loss of service to the Netezza users, Heartbeat can initiate a failover to the standby host. For details about Linux-HA and its terms and operations, see the documentation at http://www.linux-ha.org.

Distributed Replicated Block Device (DRBD) is a block device driver that mirrors the content of block devices (hard disks, partitions, and logical volumes) between the hosts. Netezza uses the DRBD replication only on the /nz and /export/home partitions. As new data is written to the /nz partition and the /export/home partition on the primary host, the DRBD software automatically makes the same changes to the /nz and /export/home partition of the standby host.

The Netezza implementation uses DRBD in a synchronous mode, which is a tightly coupled mirroring system. When a block is written, the active host does not record the write as complete until both the active and the standby hosts successfully write the block. The active host must receive an acknowledgement from the standby host that it also completed the write. Synchronous mirroring (DRBD protocol C) is most often used in HA environments that want the highest possible assurance of no lost transactions if the active node fails over to the standby node. Heartbeat typically controls the DRBD services, but commands are available to manually manage the services.

For details about DRBD and its terms and operations, see the documentation at http://www.drbd.org.