Objectives:This chapter discusses setting up a 2 node NCS cluster in the lab. The objectives are:
Concepts:Verify NCS System RequirementsThe chapter begins with a list of requirements for creating an NCS cluster:
Create a Cluster by Installing NCSThe NCS installation portion of Deployment Manager is used to create a cluster, add servers to a cluster, or upgrade NCS software. The process will create a cluster object in eDirectory, and it will install NCS on your cluster servers. In Deployment Manager, choose Post Installation Tasks, then choose Install or Upgrade a NetWare Cluster. In the options screen that appears, you can choose to copy NCS files to your servers, or skip copying. If you are upgrading or have never copied the NCS files to your servers, you should choose to copy the files. These files, however, are copied by default to NetWare 6 servers when they are installed, so you can safely skip the copying if all you are doing is creating a cluster. When creating a cluster, you must give it a name, and choose which container to place its eDirectory object in. This is followed by selecting the servers to be members of the cluster. The cluster is given its own unique IP address. The cluster creation interface asks where to create the "special cluster partition". This is the SBD partition. Each server in the cluster is assigned a sector in this partition, and must record epoch numbers in it. An epoch number is like a version number for the cluster. The epoch numbers in each sector must match. A server whose epoch number does not match will be cast off from the cluster. Check Cluster Configuration SettingsThe chapter lists five types of eDirectory objects useful in checking the cluster configuration:
Some settings for the cluster object are confusing until they are explained. The properties are arranged in tabs (when viewed through ConsoleOne). Quorum tab - A quorum is the minimum number of cluster servers that you want to be available before services are loaded and made available. This number is stored as the Membership number. However, this setting can be overridden by the setting for Timeout, which is the number of seconds to wait before loading services, no matter how many servers are available to provide them. Assume the Membership number is set to 3, and the Timeout is set to 60 seconds. Services will be loaded and made available as soon as three servers are up and running, if they come up in less than 60 seconds. However, if less than three servers are running at the end of a minute, services will still be loaded on whatever servers are available for use at that time. Protocol tab - Heartbeat is set in seconds. All nodes, except the master, are required to contact the master this often to prove they are still running on the cluster. This appears to be in conflict with Tolerance, which is the time limit that the master actually imposes on Heartbeat transmissions. Take them as a range. Assume Heartbeat is one second, and Tolerance is eight seconds. We want nodes to contact the master once a second (Heartbeat) but will not cast them off unless they exceed the time allowed in Tolerance (default is 8 seconds). This is similar to Master Watchdog and Slave Watchdog. You want the master to contact each slave node in the time interval specified as Master Watchdog. The slave nodes will not cast off the master unless it exceeds the interval specified as Slave Watchdog. Resource Priority tab - This is used to set the load order for resources when a server comes up, or when these resources must fail over to another server. Notification tab - The cluster can notify up to eight email addresses of cluster events. Each address can be configured separately for the kinds of notifications to send to it. The Master IP Address Resource object can be used to inspect (not edit) script, policy, and node information for the cluster. The Cluster Server Node objects can be used to inspect and edit the node numbers and IP addresses for each node in the cluster. Test and Monitor the ClusterClusters can be managed in ConsoleOne or Remote Manager. In Remote Manager, be sure to make contact with the servers using https protocol, at port 8009 on their IP addresses. Remote manager shows icons for the node servers, color coded for their status:
|
-
actually a volume object, it allows you to grant rights to the cluster
-
like a container, this object contains the next three types of objects.
-
holds the IP address assigned to the cluster
-
an object to hold the settings for individual servers in the cluster
-
templates for creating standard services on the cluster, such as DHCP,
iFolder, and GroupWise services.