Cluster: Best Practices

Cluster failures that cause a failover to occur include:

• Computer (hardware) failure.
• Manual move. (Move this service or application to another node on the CIMPLICITY project resource Popup menu).
• Process health subsystem issue.

In order to take full advantage of cluster benefits, it is recommended that:

• Run a single project only run in a cluster.
• Include an IP address as a dependency so the all of the components in the cluster will move as a group (e.g. The IP address will move with the project).

Points to Consider if running multiple projects on a cluster node

• The single side of the cluster has to have the capability to support all the projects.
• In a cluster configuration the router is a single point of failure; if the router fails all projects in the cluster fail.
• In a single drive set up as a shared drive where you host the projects, the projects are shut down in order on one node before the drive is transferred to the other node in reverse order.

If you have multiple drives set up to host each individual project, then you have dependencies on each individual drive set up for that specific project. The router dying will effectively take them down in order.

• Including an IP address as a dependency is required when there are multiple projects.
• It is essential that you allow enough time for all CIMPLICITY processes to shut down and restart (<CIMPLICITY Project Resource Name> Properties dialog box>Policies tab).

Although it is highly recommended that you only include one CIMPLICITY project in the cluster, if your system requires more than one project, the time needs to be increased accordingly. The projects shut down consecutively, not simultaneously. ??If enough time is not allowed some processes could still be in the process of shutting down when a restart is initiated.