1    Introduction to TruCluster Server

The Cluster File System (CFS) supports a single clusterwide namespace and uniform coherent access to all file systems in a cluster. Context-dependent symbolic links (CDSLs) are used to maintain per-system configuration and data files within the shared CFS root (/), /usr, and /var file systems.

See Section 2.2 for more information on CFS. See Section 2.4 for more information on CDSLs.

The device request dispatcher facility provides highly available clusterwide access to both character and block disk devices, as well as tape devices. All cluster disk and tape I/O passes through the device request dispatcher.

See Section 2.3 for more information on the device request dispatcher.

The semantics of LSM have been extended to a cluster environment.

See Section 2.7 for more information on LSM in a cluster environment.

The connection manager ensures that all cluster members communicate with each other in order to control the formation and continued operation of a cluster. The connection manager calculates the votes required for quorum and decides when members are added to and removed from the cluster.

See Chapter 3 for more information on the connection manager.

The CAA facility provides resource monitoring and application restart capabilities. It provides the same type of application availability provided by user-defined services in the TruCluster Available Server Software and TruCluster Production Server Software products.

See Chapter 4 for a definition of the types of applications that can run in a cluster. See Chapter 5 for more information on CAA's role in making single-instance applications highly available.

In Version 5.0A, CAA introduces support for tape devices and media changers as monitored resources. CAA also provides additional features for existing resource types, such as application monitoring through the check entry point of a service's action script. CAA also supports additional profile attributes, including restart attempts, failover delay, failover threshold and interval, autostart, and active placement.

The cluster alias subsystem lets TCP and UDP applications address the cluster as though it were a single system. When the cluster is created, a default alias is defined that addresses all cluster members. A site can define additional aliases that address some or all cluster members.

See Chapter 6 for more information on cluster aliases.

For Version 5.0A, a virtual Media Access Control (vMAC) address can be assigned to an alias. When vMAC support is enabled, an alias vMAC address follows the alias's proxy ARP master from node to node as needed. Regardless of which cluster member is serving as the proxy ARP master for an alias, the alias's vMAC address does not change. For more information on vMAC, see Section 6.7.

Cluster alias has also been enhanced in the way in which it handles ICMP packets. Other changes include enhancements in loopback performance and a reduction in the use of distributed lock manager (DLM) locks for single-instance service selection.

As shipped, the cluster is a highly available NFS server. CFS ensures that file systems exported from a TruCluster Server cluster are highly available to clients. Clients use the default cluster alias as the name of the NFS server when mounting file systems exported by the cluster.

See Section 6.3 for more information.

The Memory Channel interconnect is a high-speed interconnect designed specifically for the needs of clusters. The Memory Channel interconnect provides both broadcast and point-to-point connections between cluster members.

TruCluster Server provides a Memory Channel application programming interface (API) library, which is the same as that provided in the TruCluster Production Server Software product.

See Chapter 7 for more information on the Memory Channel interconnect. See the TruCluster Server Highly Available Applications manual for a description of the Memory Channel API.

In Version 5.0A, the Internode Communication Subsystem (ICS) Memory Channel Transport (MCT) provides higher performance, better scalability, and quicker failure detection than the previous TruCluster Server implementation of ICS that used TCP/IP over Memory Channel. ICS MCT takes full advantage of the features offered by Memory Channel for optimized data transfer and copy avoidance. It also takes advantage of Memory Channel features for failure detection rather than relying on TCP/IP timeouts.

In addition, the ICSNET network driver provides a network interface for the Memory Channel. Because the driver uses ICS to communicate with other cluster members, it is interconnect-independent.

TruCluster Server supports the DLM and its API, which is the same as that provided in the TruCluster Production Server Software product.

See Chapter 8 for a description of the DLM. See the TruCluster Server Highly Available Applications manual for a description of the DLM API.

Because a cluster uses CFS, all systems' configuration files are available for management. The SysMan suite of graphical management utilities provides an integrated view of the cluster environment, letting you manage a single member or the entire cluster.

See Chapter 9 for an overview of cluster installation and administration.

Because a cluster uses CFS, there is a single copy of security administration files such as /etc/passwd and /etc/group. A user authenticated on one member has access to all members. A user with access to a file on one member has access to that file from any member. Access control lists (ACLs) are uniformly available to all members.