The cluster node manager sits outside the cluster and provides services to all of the cluster nodes. It helps to support the mission-critical applications running inside the cluster. For example, the cluster node manager may:
Be the central account database for LDAP, NIS, or Active Directory accounts.[1]
Be a central print server using the LPRng software.
Run the Ganglia software to display colorful graphs of the cluster's activity.
Run the cron daemon to schedule batch jobs for submission to the cluster.
Run license management software (such as Flexlm).
Run the Mon monitoring package to monitor cluster nodes.
Act as a central fax server to send and receive faxes using Hylafax.
Act as a DNS or email server for the cluster.
By using the high-availability techniques we've discussed in Part II of this book, we can build the cluster node manager to make these cluster server services highly available. (See Figure 20-2.)
Figure 20-2: High-availability cluster node manager
Like the black box shown in Figure 20-1, the black box in Figure 20-2 represents a Stonith device. This device makes it possible for the backup cluster node manager to reset the power to the primary cluster node manager should it need to failover the resources. When this happens, the LDAP, sendmail, Hylafax, Named, Mon, and Ganglia (gmetad and gmond) daemons move to the backup server.
[1]An LDAP server can act like an NIS server, using software from PADL software (http://www.padl.com), as discussed in Chapter 19.