| The following list specifies the minimum hardware requirements for installing NCS: A minimum of two OES servers. At least 512MB of memory on all servers in the cluster. This provides sufficient memory to support failover of multiple applications to the same server node. At least one local disk device on which the root filesystem will be installed for each node. A shared disk system, either Storage Area Network (SAN) or iSCSI, is required for each cluster in order for all cluster data to be available to each node. This is how high availability of data is achieved. NOTE NCS will create a special cluster partition using one cylinder of one drive of the shared disk system. This will require roughly 20MB of free disk space on the shared disk system for creating the cluster partition.
WARNING Be aware that it is possible to access shared disk systems by servers both in and out of the cluster. With NCS, this can cause corruption and possibly loss of the entire volume. To avoid this problem, ensure that only servers within the cluster access the shared disk system. If a noncluster server must access the same disk system, ensure that access is only granted to noncluster volumes.
Make sure the disk system is installed and configured in accordance with the manufacturer's instructions. Make sure that the disks in the shared disk system are configured in some type of fault-tolerant configuration, such as mirroring or RAID 5. If this is not done, a single disk error can potentially cause a volume failure across the entire cluster.
Configuring the Shared Disk System Prior to installing NCS, access to the shared disk system should be configured. One of the easiest methods of accessing a shared disk is through iSCSI. iSCSI is a new standard that makes it possible to transmit SCSI communications over a network by encapsulating them in standard TCP/IP data packets. This enables you to create a low-cost Storage Area Network (SAN) using regular high-speed network hardware, and avoid the considerable costs previously associated with fiber-based SAN architectures. ISCSI BASICS Two main components are required for iSCSI access to a shared disk: iSCSI Initiator software is installed on each server that will use the shared iSCSI storage. The initiator software allows an OES Linux server or cluster to communicate with an iSCSI storage server or other iSCSI target over a normal TCP/IP network. iSCSI Target can either be a dedicated iSCSI router or a software-based iSCSI controller. NetWare servers offer a software-based iSCSI controller that allows external iSCSI initiators to access to local disks.
Configuring your Linux server as an iSCSI target is beyond the scope of this book. However, access to a shared disk is required for fully utilizing NCS. The remainder of this section will explain how to configure access to a NetWare-based iSCSI target using an iSCSI initiator on the OES Linux server. For information on configuring the NetWare iSCSI target, see the online OES NetWare documentation. INSTALLING AND CONFIGURING ISCSI INITIATORS To configure an OES Linux server as an iSCSI initiator, you must first load the initiator software on the server using YaST. The following steps describe this process: 1. | Access YaST from a terminal using yast, or from a graphical environment using yast2 or the YaST launcher from the application menu.
| 2. | Select the Software category in YaST. (This is typically the category selected by default.) From within the Software category, click on the Install and Remove Software module.
| 3. | Use the Filter drop-down box to select the Search category.
| 4. | In the Search window (left pane of the main window), enter linux-iscsi.
| 5. | Click Accept to begin the software installation. At this point you may be prompted to install a few additional packages to satisfy dependencies. If so, select to continue the installation to install all required packages.
| 6. | At the conclusion of the software installation and SuSEconfig process, select Finish to exit the Add and Remove Software module.
|
With the linux-iscsi package installed on your OES Linux server, you are now ready to configure access to the iSCSI target. Accessing an iSCSI target on NetWare can be accomplished either with or without LDAP Access Control. The following steps document the required configuration process: 1. | In order to use LDAP Access Control with an iSCSI target, create an iSCSI initiator object in eDirectory for each Linux server you want to access the iSCSI target. If LDAP Access control is not in use, skip to step 4.
| 2. | Assign the newly created iSCSI initiator objects as trustees of the iSCSI target objects in eDirectory.
| 3. | Using your favorite text editor, edit the /etc/initiatorname.iscsi file on the OES machine. Locate the InitiatorName entry, which should resemble the following:
InitiatorName=iqn.1987-05.com.cisco:01.23bcc5247683
Modify this entry to contain the LDAP distinguished name of your iSCSI initiator object as follows:
InitiatorName=iqn.1987-05.com.cisco:cn=InitiatorObject,o=Novell
Replace cn=InitiatorObject,o=Novell with the distinguished name of your initiator object in LDAP format.
| 4. | Using a text editor, edit the /etc/iscsi.conf file. Add the following information to this file:
DiscoveryAddress=iSCSI_Target_Server_IPAddress
After you enter the correct information, the iscsi.conf entry should appear as follows:
DiscoveryAddress=192.168.1.150
| 5. | In order to connect to the iSCSI target, start the initiator using the following command:
/etc/init.d/iscsi start
| 6. | After starting the initiator, use the iscsi-ls command on the OES machine to verify connectivity. If everything has been configured properly, the iSCSI target information will be displayed following the iSCSI driver version. If no targets are displayed, you may have a configuration error on the NetWare side.
|
With the initiator configured and started, you can access the iSCSI disk subsystem as if it were a directly attached resource on each OES Linux server configured as an initiator for that iSCSI device. Mapping drives, trustee rights, and directory and file attributes will all work identically to a directly connected storage device. Novell Cluster Services can be installed during the original OES Linux installation, or can be added to an existing server after the fact using YaST. To add NCS to an existing OES installation, complete the following steps: 1. | Access YaST from a terminal using yast, or from a graphical environment using yast2 or the YaST launcher from the application menu.
| 2. | Select the System category in YaST. From within this category, locate and select the Novell Cluster Services (NCS) module. This module will detect that the RPMs for NCS are missing and ask if you want to install them. Select Continue to install the necessary packages.
| 3. | At the conclusion of the software installation, SuSEconfig is executed to update the system configuration. When this completes, the configuration of the OES component will begin automatically.
| 4. | At the Novell Cluster Services (NCS) LDAP Server Configuration screen, enter the following information and click Next:
- Local or Remote Directory Server Select the radio button, which indicates whether eDirectory is running on the local server, or a remote server.
- Directory Server Address If a remote eDirectory server is in use, enter the IP address for this server.
- Admin Name with Context Enter the eDirectory administrator's credentials using fully qualified dot notation, for example, cn=admin.o=novell.
- Admin Password Enter the password for the administrator user.
- Port Details If necessary, select this button to change the configured ports for the eDirectory server that you specified earlier. The default LDAP port for unencrypted communications is 389, and port 636 is used for SSL-encrypted communications.
| 5. | At the Novell Cluster Services (NCS) Configuration screen, enter the required information (see Figure 9.2), and click Next.
- New or Existing Cluster Select whether this server will be added to a new or existing cluster. The Configure Later option is normally only used during the main OES installation, when the final configuration of NCS is being postponed until after the server is fully installed.
- Cluster FDN If the server is being added to a new cluster, enter the fully distinguished name of the new cluster. If the server is being added to an existing cluster, enter the fully distinguished name of the existing cluster.
- Cluster IP Address If you are creating a new cluster, enter a unique IP address for the new cluster. This address must be separate from the server's IP address, and is required to be a static address on the same subnet as all cluster servers.
- Optional Device for SBD Partition If you are creating a new cluster, enter the device where the SBD partition will be created. If you are not using a shared disk system, leave this field blank.
Figure 9.2. Novell Cluster Services (NCS) configuration options. 
WARNING As previously mentioned, you must have roughly 20MB of free (unpartitioned) space on one of the shared disk drives to create the cluster partition. If no free space is available, Novell Cluster Services can't use the shared disk drives. | 6. | If the server has multiple IP addresses configured, ensure that the IP address to be used for clustering with the current node is selected. In order for NCS to be active, select Start Clustering Services now. Click Next to finish the installation of NCS.
|
When NCS is installed and running, you will use the Clusters category within iManager for the remaining configuration tasks. To start and stop NCS manually, use the following init script with a start or stop command-line parameter: /etc/init.d/novell-ncs <parameter>
Converting NetWare Clusters to Linux If you have an existing NCS cluster, using NetWare 6.5 Cluster Services, you can convert NetWare nodes to OES Linux through a rolling conversion process. This process allows you to convert one server at a time, without bringing the cluster down. NOTE NetWare 5.1 and NetWare 6 clusters must first be upgraded to NetWare 6.5 clusters before they can be converted to OES Linux clusters.
In the rolling conversion process, nodes in the NetWare cluster are converted one at a time to OES Linux. During this process, normal failover of the cluster will ensure that user access is not lost in this process. To perform the rolling conversion, execute the following tasks: 1. | Shut down the NetWare server you are converting to OES Linux. Cluster resources running on this server should fail over to other nodes in the cluster.
NOTE If a cluster node holds the Master replica of any eDirectory partitions, convert other servers in the cluster first. The Master replica holder should be among the last servers converted to OES Linux. | 2. | In eDirectory, remove any references to the downed NetWare server. This includes the Cluster Node object, the Server object, and all other objects referencing the server.
| 3. | Run DSRepair on the server holding the Master replica of the partition in which the downed server existed. The eDirectory tree should be fully synchronized and free from error prior to installing the new OES Linux server. If problems are reported, run DSRepair multiple times to ensure that the tree is clean. If problems persist and cannot be resolved, contact Novell Technical Support before proceeding.
| 4. | Install OES Linux on the downed server, or replacement hardware. During the OES Linux installation, do not install Novell Clustering Services as this will be installed later. Optionally, the new server can be installed using the same IP address and server name as the NetWare server it is replacing. See Chapter 2, "Installing OES Linux," for information on installing OES Linux.
| 5. | Set up and verify SAN connectivity on the new Linux server. These steps are specific to the SAN being used. For more information, refer to documentation supplied by your SAN vendor.
| 6. | Install Novell Cluster Services to your OES Linux server, and select to add the server to your existing NetWare cluster. See "Installing Novell Cluster Services" earlier in this chapter for more information.
| 7. | Execute sbdutil f via a terminal on the OES Linux server to verify connectivity to the cluster (SBD) partition on the SAN.
| 8. | Start NCS by executing /etc/init.d/novell-ncs start.
|
After the conversion process completes, cluster resources will automatically fail back to the Linux server if the following conditions are met: If necessary, manually migrate cluster resources to the new Linux server to complete the conversion process. The same eight conversion steps can now be restarted on additional NetWare 6.5 nodes in the cluster, or the cluster can be left using both NetWare and Linux. If all nodes in the NetWare 6.5 cluster are converted to Linux, the cluster conversion process must be finalized by writing the cluster resource load and unload scripts into eDirectory. This must be performed on one OES Linux server after all servers in the cluster have been converted from NetWare. The following two steps are required for this process: 1. | From a terminal on one OES Linux server in the converted cluster, execute cluster convert preview <resource_name>. This will display the resource load and unload script changes that will be made when the conversion is committed.
| 2. | Execute cluster convert commit from a terminal on one OES Linux server to perform the script conversion and finalize the NetWare to Linux conversion.
|
After the cluster convert commit command has been executed, all load and unload scripts will be permanently converted to Linux scripts, and your conversion will be complete. |
