Clustering Support

Topics | Support


Overview

Software Operations in a Cluster

Adding or Removing Cluster Nodes

Network TCP Ports

Multi Instancing

Windows Clusters

NetWare Clusters

UNIX Clusters

License Requirements


Overview

A collection of multiple server computers into a single unified cluster provides an added measures of performance, reliability, and fault tolerance. The CommServe, MediaAgent, and certain Agents take advantage of the failover protection afforded to virtual servers within the clustering environment. If an active node fails, the software will still be able to function from the node that has not failed, and will continue to be able to perform data protection and recovery operations.

Terminology

When referring to the physical servers that comprise a cluster, individual computers are referred to as nodes. The physical server may be called one of the following depending upon the context in which it is being discussed: "physical computer," "physical node," "Client computer". Also, the "physical server" may be an "active node" or a "standby (passive) node" depending on its role within the cluster environment at a given point in time.

"Virtual Server" or "Virtual Node" refers to the hardware and software components (Cluster Resource Group) within the cluster that are managed by Cluster Service as a single, logical unit. The virtual server is actually not a physical computer but a logical construct within the system. However, like the physical servers, the virtual server has an Internet Protocol (IP) address and a network name. As such, the virtual server secures its own license on the CommServe, and it has its own registry information. This registry information is included on all physical nodes in the cluster. Once the proper installations are made to the physical nodes, the virtual server can be configured on each node to function within the clustering environment.

"Cluster Service" refers to the collection of components on each node that perform cluster-specific activity, managing the Cluster Resource Group, or virtual server.

The "Shared Disk" resource is connected to each physical computer. This disk resource is designed to include shared data for CommCell operations (e.g., data protection and recovery operations). Information that is relevant only to each physical server should be included only on the physical server's local disk and not on the shared disk.

Supported Cluster Configurations

Both Active/Passive, and Active/Active cluster configurations are supported.


Software Operations in a Cluster

The CommServe, MediaAgent and certain Agents can take advantage of the failover protection afforded to the virtual server within the cluster.

You can create schedules for data protection operation on virtual servers in the same way as you create data protection operation schedules for data on a physical node.

All schedules are saved on the CommServe. When a job is scheduled for the virtual server, the CommServe executes that job on the active server. If the active server fails while jobs are in progress, a failover will occur, and running jobs are temporarily placed in a pending state. Once the failover completes, the new active server will restart the jobs. In addition, remaining scheduled jobs are implemented on this new active node. This is shown in the illustration; the virtual server is associated with Node A and Node B. Since Node A is the active node (and therefore has access to the shared disk), scheduled jobs for the virtual server are executed on Node A. When Node A fails, after the failover, a new data protection operation is started on Node B which is now the active node. In addition, the remaining scheduled jobs are executed on Node B.

Regardless of the hosting node, the software can perform operations for a virtual server, including data protection and recovery operations, as long as the software has access to the network and to the virtual server's network name and Internet Protocol (IP) address.

Failovers

CommServe on a Microsoft Cluster (MSCS)

Failure of the CommServe software on the active server will cause a failover to a standby server.

MediaAgent or Agent on a Microsoft Cluster (MSCS), or Agent on a Non-Microsoft Windows Cluster

By default, a failure of the MediaAgent or any Agent software on the virtual server will not cause a failover to be initiated. To configure a cluster virtual server to fail over when the MediaAgent or Agent software fails, after you have finished installing the appropriate software on all the nodes in the cluster, perform the following on each of the computers configured to host the virtual server:

  1. Configure the sServiceMonitorList registry key with a list of services that GxClusPlugin will monitor for failure. For information about the services used by an Agent or MediaAgent, which can be listed in this registry key, see Services.
  2. Configure the sServiceMonitorFrequency registry key for the frequency, in seconds, at which those services should be polled for failure. This value should be less than the service restart interval being monitored, if any.
  3. Configure the properties of GxClusPlugin so that a failover will be triggered by a failure of any of the services listed in the sServiceMonitorList registry key. In Cluster Administrator, right-click GxClusPlugin and select Properties; in the Advanced tab of the GxClusPlugin Properties, select "Restart" and "Affect the Group".

MediaAgent or Agent on a UNIX Cluster

A failure of the MediaAgent or any Agent software on the active server in a UNIX cluster will not cause a failover to be initiated.

The cvclusternotify.sh script should be added as part of the normal cluster startup/shutdown procedure. The script is provided as a generic template, and it must be run whenever the current node shuts down during failover and as the new active node comes online. In both cases, data protection services must be up and running. The script works as follows:

  • During node shutdown, the script notifies File Archiver for Unix and the CommServe that the listed file systems will no longer be monitored on this node.
  • During node startup, the script notifies File Archiver for Unix and the CommServe that the listed file systems will now be monitored on this node.

Usage:

cvclusternotify - Program to notify Calypso of cluster failovers

cvclusternotify -inst <Instance> -cn <Client Name> -start|-shutdown \[-fs <dir1>] [-fs <dir2>] ...

This command notifies Calypso that the specified client from the specified instance is going up or down because of a cluster failover. The optional "-fs <dir>" parameters can be provided to notify File Archiver for Unix that certain file systems are disappearing or becoming available during the failover. For each of these file systems, the CDR/File Archiver for UNIX driver will be instructed to stop or begin file system monitoring.

Example:

Suppose you have a two-node cluster where the virtual machine client name is "virtual". (Physical node names are not needed.) Suppose that the application instance in this case is "Instance001" and that the archived data currently exists on two shared file system directories as follows:

/share1/filesystem1
/share2/filesystem2

For the node shutdown, the following command is issued:

cvclusternotify -inst Instance001 -cn "virtual" -shutdown -fs /share1/filesystem1 -fs /share2/filesystem2

This will stop File Archiver for Unix from monitoring the indicated file systems on the node that is shutting down.

For the node (active) startup, the following command is issued:

cvclusternotify -inst Instance001 -cn "virtual" -start -fs /share1/filesystem1 -fs /share2/filesystem2

This will enable File Archiver for Unix to start monitoring the indicated file systems on the node that is starting up.


Adding or Removing Cluster Nodes

After a computer has been added to a cluster as a new node, and has been verified to be functional, you can install CommServe, MediaAgent or Agent software on it for the virtual server. MediaAgent or Agent software can also be installed as non-cluster-aware (a "physical" installation.) Normal installation procedures can be used for this.

Before removing a node from a cluster, the CommServe, MediaAgent or Agent software can be uninstalled and/or Deconfigured, if it is not running any jobs, and if it is not the current active node. The normal uninstall and deconfiguration procedures can be used for this. For more information and procedures, see Uninstalling Components.


Network TCP Ports

Every physical node in a cluster which is configured to host a given cluster virtual server, must have the same TCP port numbers configured for that cluster virtual server. For more information, see Network TCP Port Requirements - Clusters.


Multi Instancing

Multi Instancing is supported for clusters, with some restrictions. For more information, see Multi Instancing.


Windows Clusters

This section contains the following topics:

Supported Windows Cluster Installations

The CommServe, MediaAgent, and certain Agents can be installed in a Windows cluster to take advantage of the failover protection afforded to virtual servers within the cluster environment. When installed on the virtual server, the software secures its own license on the CommServe, and it has its own configuration file.

Each node within a cluster, whether active or passive, must have the necessary software components installed. Installing these software components to be cluster-aware involves both the active and passive nodes, whether the software component is installed directly to a particular physical node or not. During the virtual server installation, software binaries are installed on all physical nodes in the cluster, and the configuration files on all physical nodes are edited to contain information about all installed virtual servers.

It is important to remember the following guidelines:

Note that the same component can be installed multiple times to the same instance. This means physical and virtual machines can be installed to the same instance. This can be useful, for example, when you want to install an Agent to protect the data on the physical computer, and the same Agent to protect the data on the virtual server, and you want both Agents to be in the same CommCell.  The advantage of such an installation is that you will have one set of binaries and Services, which in turn reduces the load on the system resources. However, note that such installations are not applicable for CommServe. The CommServe can only be hosted in one virtual machine; in other words CommServe instance cannot be shared by another physical or virtual machines.

The following sections describe the basic principles of installing the different software components supported on a Windows cluster.

Physical Installation

A physical installation of a software component is the default type of installation, where the software is installed on a physical server and acts independently of the cluster. Services installed in this way do not take advantage of clustering features like failover protection. For instance, the File System iDataAgent can be installed on each node in the cluster to protect the data on the physical servers. MediaAgents are also supported on the physical node of a cluster. (To install the software in this non-cluster-aware manner, select Install on the Physical node and not Install on a Virtual server on the "Cluster Install Type" screen during the software installation.) The CommServe cannot be installed in this manner; it can only be installed as described in CommServe Installation.

For software installation instructions, see Installation.

CommServe Installation

The CommServe must always be installed first in a CommCell. In a clustered environment, the CommServe software binaries are installed on the physical nodes (active, then all passive nodes) while the SQL database is installed on the virtual server. The install procedure is performed on the active node with the choice of an automated or separate install to all passive nodes, with no failover required at the end of the installation process. For any passive nodes unavailable at the time the install was performed from the active node to the virtual server, a separate install must be performed. Installed to each node, the CommServe software is tolerant to binary corruption; if the binaries are corrupted on one physical node, the cluster resources can fail over to another physical node, which has its own set of binaries, affording the same performance, reliability, and fault tolerance as a cluster provides for power, hardware, or application failure.

For software installation instructions for a clustered CommServe, see CommServe Deployment.

When the CommServe is to be installed on a virtual server together with a Windows File System iDataAgent, also refer to Install the CommServe and Windows File System iDataAgent on the same Virtual Server.

MediaAgent Installation

The MediaAgent software can be installed in one of the following ways:

For software installation instructions for a MediaAgent, see MediaAgent Deployment.

File System iDataAgent Installation on all Physical nodes

If you will be installing the File System iDataAgent on the virtual server, you must first install the File System iDataAgent on each physical node of the cluster.

For MSCS clusters, all configuration information is stored on a quorum disk that is selected during the installation of the Microsoft Cluster Server software. It is very important to Back Up Quorum Data, and this requires a File System iDataAgent on the physical server. The quorum databases are automatically backed up by the default subclient for the File System iDataAgent on each physical node. The Windows 2000 iDataAgent will include the quorum databases during a system state backup.

Agent Installation for Virtual Servers

All iDataAgent installations should be performed after the CommServe, MediaAgent and any File System components are installed. To protect the data on the virtual server, the Agent software must be installed to the virtual server for every node in the cluster configured to host that virtual server. When the virtual server install procedure is performed on the active node, you can choose to perform an automated install to all available passive nodes at the same time.

For example, a cluster may have a virtual server running Microsoft Exchange, with node1 as the active node. To take advantage of failover protection, an installation of the Exchange iDataAgent software should be performed to the virtual server (and all available passive nodes) from node1, and then separately on any unavailable passive node(s). If not already installed, the Windows File System iDataAgent will be automatically installed on the virtual server with the Exchange iDataAgent.

For software installation instructions, see Installation.

Data Protection for a Windows Cluster

In general, data protection operations in a cluster environment function as they do in a non-clustered environment. However, it is recommended that you include the shared data for data protection operations only on the shared disk, and conversely, that you include local data for each physical server only on the physical server's local disk, using the appropriate Agent software on the physical server to back up this disk.

For additional cluster-specific information related to backups, refer to Backup Filters for Windows Clusters and Backing Up Quorum Data for MSCS Clusters.

Data Recovery for a Windows Cluster

In general, data recovery operations in a cluster environment function as they do in a non-clustered environment. You can restore data that resides locally on a physical node, or that resides on the virtual server's shared disk, just as you would restore data for a non-clustered computer. The process of restoring client data to a virtual server is the same as it is on a non-clustered computer; the virtual server is viewed as another computer on the CommCell.

Note that the virtual server always requires that at least one physical server is up and running. Therefore, if all the physical servers fail, be sure to restore at least one of these servers before you restore the virtual server.

Restoring Cluster Databases on Windows 2000

Once a full system restore is completed (including the system state), the node should be functional with the cluster services online. Cluster databases are restored to a temporary location and should only be restored if they are lost or corrupted on the quorum drive. If it is necessary to retrieve the restored cluster databases, simply type authorutil -cluster from a command prompt. This will move the cluster databases from their temporary location to the MSCS folder on the quorum drive.

Recovering the Virtual Server

Recovering a node in the cluster requires a full system restore. At the least, you must establish a connection to the CommServe and restore the System State and Windows File System iDataAgent. Before you restore the virtual server, you should make sure that all of the physical nodes are functioning. If any of the physical nodes have failed, restore them first. When all the cluster physical nodes are functioning, restore the virtual server by restoring the data for all the iDataAgents installed on the virtual server.

Additional Considerations for Windows Clusters

Back to Top


NetWare Clusters

This section contains the following topics:

Supported NetWare Cluster Installations

The MediaAgent and certain Agents can be installed in a NetWare cluster to take advantage of the failover protection afforded to virtual servers within the cluster environment. When installed on the virtual server, the software secures its own license on the CommServe, and it has its own configuration file.

Each node within a cluster, whether active or passive, must have the necessary software components installed. Installing these software components to be cluster-aware involves both the active and passive nodes, whether the software component is installed directly to a particular physical node or not. During the virtual server installation, the configuration files on all physical nodes are edited to contain information about all installed virtual servers.

It is important to remember the following guidelines:

The following sections describe the basic principles of installing the different software components supported on a NetWare cluster.

Physical Installation

A physical installation of a software component is the default type of installation, where the software is installed on a physical server and acts independently of the cluster. Services installed in this way do not take advantage of clustering features like failover protection, unless you also install software on the cluster virtual server. For instance, the File System iDataAgent can be installed on each node in the cluster to protect the data on the physical servers. MediaAgents are also supported, but not required, on the physical node of a cluster.

For software installation instructions, see Installation.

MediaAgent Installation

The MediaAgent software can be installed in one of the following ways:

For software installation instructions for a MediaAgent, see MediaAgent Deployment.

File System iDataAgent Installation on all Physical Nodes

If you will be installing the File System iDataAgent on the virtual server, you must first install the File System iDataAgent on each physical node of the cluster.

For software installation instructions for a clustered NetWare File System iDataAgent, see NetWare Deployment.

Agent Installation for Virtual Servers

All iDataAgent installations should be performed after the CommServe, MediaAgent and any File System components are installed. To protect the data on the virtual server, install the Agent software from the Windows computer with the Novell Client software, to the virtual server. When the virtual server install procedure is performed, an automated configuration of the Galaxy.INI file is performed on all available nodes at the same time. If any node was not available during the installation, a separate procedure must be performed on that node to configure its Galaxy.INI file.

For software installation instructions, see Installation.

Data Protection on a NetWare Cluster

In general, data protection operations in a cluster environment function as they do in a non-clustered environment. However, it is recommended that you include the shared data for data protection operations only on the shared disk, and conversely, that you include local data for each physical server only on the physical server's local disk, using the appropriate Agent software on the physical server to back up this disk.

Data Recovery on a NetWare Cluster

In general, data recovery operations in a cluster environment function as they do in a non-clustered environment. You can restore data that resides locally on a physical node, or that resides on the virtual server's shared disk, just as you would restore data for a non-clustered computer. The process of restoring client data to a virtual server is the same as it is on a non-clustered computer; the virtual server is viewed as another computer on the CommCell.

Additional Considerations for NetWare Clusters

Back to Top


UNIX Clusters

This section contains the following topics:

Supported UNIX Cluster Installations

The MediaAgent and certain Agents can be installed in a cluster to take advantage of the failover protection afforded to virtual servers within the cluster environment. When installed on the virtual server, the software secures its own license on the CommServe, and it has its own configuration file.

Each node within a cluster, whether active or passive, must have the necessary software components installed. Installing these software components to be cluster-aware involves both the active and passive nodes, whether the software component is installed directly to a particular physical node or not. During the virtual server installation, software binaries are installed on all physical nodes in the cluster, and the configuration files on all physical nodes are edited to contain information about all installed virtual servers.

It is important to remember the following guidelines:

The following sections describe the basic principles of installing the different software components supported on a UNIX cluster.

Physical Installation

A physical installation of a software component is the default type of installation, where the software is installed on a physical server and acts independently of the cluster. Services installed in this way do not take advantage of clustering features like failover protection. For instance, the File System iDataAgent can be installed on each node in the cluster to protect the data on the physical servers. MediaAgents are also supported, but not required, on the physical node of a cluster.

For software installation instructions, see Installation.

MediaAgent Installation

The MediaAgent software can be installed in one of the following ways:

For software installation instructions for a MediaAgent, see MediaAgent Deployment.

File System iDataAgent Installation on all Physical Nodes

If you will be installing the File System iDataAgent on the virtual server, you must first install the File System iDataAgent on each physical node of the cluster.

For software installation instructions for a clustered UNIX File System iDataAgent, see UNIX Deployment.

Agent Installation for Virtual Servers

To protect the data on the virtual server, the Agent software must be installed to the virtual server from every node in the cluster configured to host that virtual server. It must be installed to a different directory for each virtual server-physical server combination.

NOTES

For software installation instructions, see Installation.

Data Protection on a UNIX Cluster

In general, data protection operations in a cluster environment function as they do in a non-clustered environment. However, it is recommended that you include the shared data for data protection operations only on the shared disk, and conversely, that you include local data for each physical server only on the physical server's local disk, using the appropriate Agent software on the physical server to back up this disk.

If an Oracle database is included in the system, do not perform simultaneous data protection operations from the virtual server and the physical server(s) because Oracle locks its database files in such a case.

Data Recovery on a UNIX Cluster

In general, data recovery operations in a cluster environment function as they do in a non-clustered environment. You can restore data that resides locally on a physical node, or that resides on the virtual server's shared disk, just as you would restore data for a non-clustered computer. The process of restoring client data to a virtual server is the same as it is on a non-clustered computer; the virtual server is viewed as another computer on the CommCell.

Additional Considerations for Unix Clusters

Back to Top


License Requirements

A specific Product License must be available for each component you install, whether CommServe, MediaAgent, or Agent, and a license is consumed for each installation. When a File System Agent is installed on a virtual server, a special license is required, for Virtual File System.

Windows Cluster

As an example of a Windows cluster, consider one with three physical nodes and Microsoft SQL server on the virtual server. License requirements would be as follows:

NetWare Cluster

As an example of a NetWare cluster, consider one with three physical nodes and one virtual server, with the NetWare File System iDataAgent, File Archiver for NetWare, and a MediaAgent installed. License requirements would be as follows:

UNIX Cluster

As an example of a UNIX cluster, consider one with five physical nodes configured to host the same virtual server, with an Informix instance on the virtual server. License requirements would be as follows:

 

Review general license requirements included in License Administration. Also, View All Licenses provides step-by-step instructions on how to view the license information.

Back to Top