The Fabric Manager high availability cluster includes cluster groups; a cluster group is a collection of resources that are dependent on one another. In the Fabric Manager clustered configuration, all resources run on one node or the other.
However, the postgres database is central to all other operations. Subgroups along Fabric Manager application lines are defined separate from the database to isolate any failures in a particular application. For example, the cluster defines an authentication group that consists of samba, winbind, wine32, and the Unisys authentication server. If any of these components fail, the log-ins from the desktop partition are temporarily not allowed until the cluster stops and restarts the related services. Thus services affected during a resource failure recovery are isolated.
The Fabric Manager high availability mount point for data replication is not part of a Fabric Manager cluster group. This mount point includes data from the postgres and Apache Fabric Manager cluster groups defined as follows:
Data Replication Resource Group: This master group for the Fabric Manager high availability cluster consists of the data replication block device (DBRD) and runs on both nodes as a master slave apart from the application. One of the cluster nodes must be promoted to master before any other cluster services can start.
Database Resource Group: This group is the main group related to the postgres database that contains the postgres, jetty, ldap, nagios, and nagios_feeder services. The mount point is defined as /mnt/ffmha. This file system is mounted on the system where the DRBD resource is master. It is only mounted on one system at a time. If any of these resources fail, the cluster attempts to restart the group on the same system without affecting the other groups. If the cluster cannot start the failing resource, all of the Fabric Manager resources fail over to the standby cluster node.
FFM Web Resources Group: This group contains cluster resources related to Apache—including Apache2 and Tomcat. This group depends on the DRBD resource being promoted to master. Any failure in Apache or Tomcat is isolated to this resource group only. If the cluster cannot resolve the resource failure, all resources on this cluster node fail over to the standby cluster node.
Authentication Resources Group: This cluster resource group contains the services that support partition desktop and authentication—including winbind, wine32, and Unisys authentication. A resource failure in one of these services results in a possible restart of all of the related services. If the cluster cannot resolve the resource failure, a node failover occurs and all Fabric Manager resources move to the standby cluster node.
Access Point Resources: These resources include the ClearPath Forward Management LAN (FM LAN) floating IP address, and an optional customer corporate LAN (public LAN) floating IP address.
Note: The public floating IP address may be added or removed after creating a cluster.
Cluster Monitor Resources: This resource monitors changes in cluster resources, and produces an audit log that is visible from the FMP Manager user interface.
Note: Do not change the network settings of an active cluster. The cluster uses the FM LAN for heartbeat communication between the Fabric Management Platforms in the cluster, and changing the network settings while the cluster is up may result in loss of communication and STONITH reboot. (There will not be any data loss, since the replication immediately disconnects if a split brain situation occurs.)