- Cluster Design Guidelines
- Creating a Failover Matrix
- Application-Specific Design Guidelines
- Clustering Without Shared Storage
This chapter is from the book
This chapter is from the book
This chapter is from the book
Clustering Without Shared Storage
Most clustered applications work with files on the file system and therefore require access to shared storage. Examples of such applications can be found in Chapter 8—GroupWise, iFolder, and Samba. But there are also applications that do not require access to files on the file system. So there really is no need for them to run on a cluster with shared storage. In this section we will discuss Novell applications that can run in a cluster without the need for setting up shared storage.
This means that you can use an existing cluster where these applications can be installed without configuring extra storage pools. But it also means that you can set up a basic twonode cluster without any shared storage at all to provide high availability for these types of applications. And this is free because Open Enterprise Server comes with a license for a twocluster node.
An example of a service that can be set up without shared storage is Lightweight Directory Access Protocol (LDAP). For LDAP the server reads its configuration from eDirectory, which, if set up correctly, is replicated to all servers. The only thing needed to set up LDAP for high availability is to configure the two servers in your cluster for LDAP and to set up a secondary IP address for use with this service.
We have chosen LDAP as an example, but there are more applications that just like LDAP do not rely on shared storage for their operation. DNS, DHCP, and SLP are examples of such applications. They work with data that is stored in eDirectory or on the local server hard disk. Chapter 5, "Creating Clustered Resources," describes how DHCP can be configured on both NetWare and Linux.
Prepare the Cluster
Setting up a cluster without shared storage is done in the same way as when shared storage is involved. We will provide the basic steps here. First, plan which two servers you want to use for this cluster and what the cluster's IP address will be. This must be an address in the same subnet that your servers are already in. Also, decide where the cluster container object must be placed in your tree.
The actual installation for a NetWare cluster is performed with the Deployment Manager (NWDEPLOY.EXE), located at the root of your NetWare 6.5 operating system CD. This utility contains the option Install/Upgrade Cluster in the Post Install tasks on the left, for installing a cluster. For Linux you start YaST to set up the cluster nodes, as described in Chapter 4.
During installation for NetWare, you will be prompted with a question, asking whether your server has shared storage. The answer is set to No by default if no shared storage is found. In a cluster scenario where you do have shared storage, this is the point when you would cancel the installation to find out what is wrong. However, for this cluster select Yes to continue. The cluster will be installed, and if you have selected to do so, the software will also be started automatically. For Linux you just leave the field where to configure the shared storage empty.
You now have a cluster you can use to create high availability for networking services. We will discuss how to put LDAP into this cluster next.
The first step is to set up LDAP the way it needs to work and to store eDirectory replicas on both servers. Create a setup just as you would do for any normal server, where you would be running LDAP for your application. For this typical scenario, you may want to set up one LDAP group and make the two LDAP servers a member of that group to keep the part that is configured through a group consistent throughout the LDAP environment. However, keep in mind that there are also LDAP server–specific settings you must set identically for both servers.
When you are finished with the LDAP configuration, create a new cluster resource. Do that from either iManager or ConsoleOne. Figure 3.5 shows an example of creating the resource in iManager. Do not use any of the available templates. Just create a new blank resource.
)
Figure 3.5 For an application without shared storage, create a simple cluster resource.
In the cluster resource for your LDAP server, configure the load script to add the secondary IP address to the server when loading the resource. Also, configure the unload script to delete the address when bringing the resource offline or migrating it to the other node in the cluster. Here's an example in which we used the address 192.168.1.120:
|
LDAP Resource Load Script |
LDAP Resource Unload Script |
|
add secondary ip address 192.168.1.120 |
del secondary ip address 192.168.1.120 |
After setting up the LDAP server in the cluster, it is ready to be used. There is no need to reset the LDAP server or to load or unload the NLDAP.NLM. The LDAP server automatically binds to the new secondary IP address as soon as it becomes available on the server.