Cluster administration is more complex than system administration because a cluster has multiple points of control and multiple points of failure. In a cluster, for example, there is more than one place where a print job can get stuck and more than one place where a user's logon can fail. The cluster administrator needs to know where all of these points of control and points of failure are in order to troubleshoot problems.
Because Linux Enterprise Cluster administration is a new discipline, best practices for Linux cluster administration have not yet evolved. In this chapter, I will therefore describe several case studies in Linux cluster administration:
Administering accounts without Active Directory
Building a fault-tolerant print spooler
Rebooting nodes for preventative maintenance
Sending and receiving email in a cluster environment
Remote administration of cluster nodes