Table A-1 summarizes the differences between IRIX FailSafe and SGI Cluster Manager for Linux for those readers who may be familiar with FailSafe.
 | Note: SGI Cluster Manager for Linux members and FailSafe nodes do not work together and cannot form a high-availability cluster. |
Table A-1. Differences Between FailSafe and SGI Cluster Manager
Topic | FailSafe | SGI Cluster Manager |
|---|---|---|
Operating system | IRIX | SLES9 with SGI ProPack 4 for Linux |
Terminology | ||
Size of cluster | 8 nodes | 2 members |
Node/member name | Hostname or private network address | Hostname only |
NFS lock failover | Supported | Not supported |
A node that is participating the cluster membership. FailSafe tries to include the tiebreaker node in the membership in case of split-brain scenarios. | The IP address of machine or a router that does not participate in the cluster membership. Usually it is the IP address of a network router that connects the SGI Cluster Manager members to the external world (clients). In a split-brain scenario, only those members that can contact the tiebreaker IP address can form a cluster. There is also a disk tiebreaker. | |
Supported | Not supported | |
Configuration information storage | Information is stored in the cluster database. The cluster database is replicated on all nodes automatically and kept in synchronization. | Information is stored in the /etc/cluster.xml configuration file and in the shared partitions. For initial configuration, you must copy this file to all members, such as by using scp . After making configuration changes, you must verify that configuration files are in synchronization. See “Step 15: Verify that Configuration Changes are Synchronized” in Chapter 4. |
Making changes while the service is enabled | Device parameter, IP address parameters, and check interval can be changed. | Device parameter, IP address parameters, and check interval cannot be changed. |
Heartbeat interval and timeout | You can specify cluster membership heartbeat interval and timeout (in milliseconds). | In the command line, you can specify the heartbeat interval (in microseconds) and the number of heartbeats that can be consecutively missed (tko_count). You can also specify the aggregate failover speed in the GUI. |
Heartbeat networks | Allows multiple networks to be designated as heartbeat networks. You can choose a list of networks. | Allows heartbeat on all networks or as a multicast on the hostname network. However, you cannot choose a list of networks. |
Separate scripts named start, stop, monitor, restart, exclusive. | A bash script that contains start, stop, and status parameters (see Chapter 6, “Creating a New Highly Available Application”). The equivalent for restart in SGI Cluster Manager is to perform a stop and then a start; there is no equivalent in SGI Cluster Manager for exclusive. | |
Resource timeouts | Timeouts can be specified for each action (start, stop, monitor , restart, exclusive) and for each resource type independently. | Timeout can be specified for each service irrespective of the action or the number of resources it contains. |
Resource dependencies | Resource and resource type dependencies are supported and can be modified by the user. | Applications have fixed dependencies. The start and stop order of applications cannot be modified. |
Failover policies | The ordered and round-robin failover policies are predefined. User-defined failover policies are supported. | Only the predefined ordered policy is supported. No user-defined failover policies are supported. |