There are several cluster-specific issues to consider as you plan to install the NetWare MTA and POA in your clustered GroupWise system:
The GroupWise agents listen on all IP addresses, both primary and secondary, that are bound to the server on their specified port numbers. This means that any time there is a possibility of two of the same type of agent loading on the same node, it is important that each agent use a cluster-unique port number, even though each agent is using a unique secondary IP address. The best way for you to avoid port conflicts is to plan your cluster so that each agent in the cluster runs on a cluster-unique port. Print out a copy of the IP Address Worksheet to help you plan secondary IP addresses and cluster-unique port numbers for all GroupWise agents.
The following filled-out version of the IP Address Worksheet illustrates one way this can be done:
|
Domain |
MTA IP Address |
MTA MTP Port |
MTA HTTP Port |
||
|---|---|---|---|---|---|
|
Provo1 |
172.16.5.81 |
7100 |
7180 |
||
|
Post Office |
POA IP Address |
POA C/S Port |
POA MTP Port |
POA HTTP Port |
|
|---|---|---|---|---|---|
|
Development |
(same as MTA) |
1677 |
7101 |
7181 |
|
|
Manufacturing |
172.16.5.82 |
1678 |
7102 |
7182 |
|
|
Internet Agent |
GWIA IP Address |
MTA MTP Port |
MTA Live Remote Port |
MTA HTTP Port |
GWIA HTTP Port |
|---|---|---|---|---|---|
|
GWIA Domain MTA |
172.16.5.83 |
7110 |
7677 |
7183 |
N/A |
|
Internet Agent (GWIA) |
(same as MTA) |
N/A |
N/A |
N/A |
9850 |
|
WebAccess Agent |
WebAccess IP Address |
MTA MTP Port |
MTA HTTP Port |
WebAccess Agent Port |
WebAccess HTTP Port |
|---|---|---|---|---|---|
|
WebAccess Domain MTA |
172.16.5.84 |
7120 |
7184 |
N/A |
N/A |
|
WebAccess Agent (GWINTER) |
(same as MTA) |
N/A |
N/A |
7205 |
7205 (same as agent) |
This example places the Development post office on the same node and on the same GroupWise volume with the Provo1 domain; therefore, the Provo1 MTA and the Development POA can use the same secondary IP address. The Manufacturing post office is placed on a different node on a different GroupWise volume, so that the Manufacturing post office has a different secondary IP address.
The example also illustrates that the MTA, the POA, and the Internet Agent use different port numbers for agent ports and HTTP ports. In contrast, the WebAccess Agent uses the same port number for the agent port and the HTTP port.
The example uses default port numbers where possible. For example, the default MTA message transfer port is 7100 and the default POA client/server port is 1677. Incrementing port numbers are used in the example when multiple components have the same type of ports. For example, port numbers 1677 and 1678 are both POA client/server ports and port numbers 7180 through 7184 are all HTTP ports. Incrementing from the default port numbers generates unique, though related, port numbers.
If you are going to set up a GroupWise name server to help GroupWise clients locate their post offices, make sure that the default POA port number of 1677 is used somewhere in the cluster. For more information, see Simplifying Client/Server Access with a GroupWise Name Server
in Post Office Agent
in the GroupWise 8 Administration Guide.
|
IP ADDRESS WORKSHEET |
|---|
|
Fill out the IP Address Worksheet to help you plan secondary IP addresses and cluster-unique port numbers for all GroupWise agents in the cluster. (MTA, POA, Internet Agent, WebAccess Agent). |
After you have filled out the IP Address Worksheet, transfer the secondary IP addresses and cluster-unique port numbers from the IP Address Worksheet to the System Clustering Worksheet and the Agent Clustering Worksheet so that they are available in the sequence in which you will need them as you set up GroupWise in a cluster.
|
SYSTEM CLUSTERING WORKSHEET |
|---|
|
If you are setting up a new GroupWise system, under Item 6: Shared Volumes for GroupWise Administration, specify secondary IP addresses for your GroupWise administration volumes. Under Item 7: Shared Volume for Domain, use the domain MTA secondary IP address from the IP Address Worksheet as the domain volume IP address. If you are planning the post office on a different volume from the domain, under Item 8: Shared Volume for Post Office, use the post office POA secondary IP address from the IP Address Worksheet as the post office volume IP address. |
|
AGENT CLUSTERING WORKSHEET |
|---|
|
Under Item 4: MTA Network Information, transfer the secondary IP address and cluster-unique port numbers for the MTA from the IP Address Worksheet to the Agent Clustering Worksheet. Under Item 7: POA Network Information, transfer the secondary IP address and cluster-unique port numbers for the POA from the IP Address Worksheet to the Agent Clustering Worksheet. |
By default, a GroupWise volume is configured to have all nodes in the cluster in its failover path, organized in ascending alphanumeric order. Only one node at a time can have a particular GroupWise volume mounted and active. If a GroupWise volume’s preferred node fails, the volume fails over to the next node in the failover path. You will want to customize the failover path for each GroupWise volume based on the fan-out-failover principle.
When a node fails, its volumes should not all fail over together to the same secondary node. Instead, the volumes should be distributed across multiple nodes in the cluster. This prevents any one node from shouldering the entire processing load typically carried by another node. In addition, some volumes should never have the potential of being mounted on the same node during a failover situation. For example, a post office and POA that service a large number of very active GroupWise client users should never fail over to a node where another very large post office and heavily loaded POA reside. If they did, users on both post offices would notice a decrease in responsiveness of the GroupWise client.
|
AGENT CLUSTERING WORKSHEET |
|---|
|
Under Item 3: Domain Failover Path, list the nodes that you want to have in the domain volume failover path. The MTA might need to run on any node that the domain volume fails over to. If you are planning the post office on a different GroupWise volume from where the domain is located, under Item 6: Post Office Failover Path, list the nodes that you want to have in the post office volume failover path. The POA might need to run on any node that the post office volume fails over to. |
When you install the NetWare MTA and POA in a clustering environment, you can choose between two different installation locations:
Table 2-3 Agent Software Installation Locations
|
Location |
Description |
|---|---|
|
sys:\system on each node in the cluster |
This is the default location provided by the Agent Installation program. Because the agents must be installed on each node where they might need to run during a failover situation, you need to do one of the following if you select this alternative:
|
|
system directory on each GroupWise volume |
If you create a vol:\system directory on a GroupWise volume, the agent software and startup files fail over and back with the domains and post offices that the agents service. Unless you have a very small GroupWise system with all domains and post offices on a single GroupWise volume, you still need to install the agent software multiple times, once to each GroupWise volume. |
A simple way to look at the agent location alternatives would be that if you have fewer nodes on failover paths than you have GroupWise volumes for domains and post offices, then it would be most efficient to install the agent software to the nodes. Conversely, if you have fewer GroupWise volumes than you have nodes on failover paths, then it would be most efficient to install the agent software to the GroupWise volumes. However, there are issues to consider that extend beyond efficiency during installation.
The following sections can help you choose which installation location would be best for your clustered GroupWise system:
Using a vol:\system directory on each GroupWise volume has several advantages:
If you change information in the agent startup files, you only need to change it in one place, not on every node on any GroupWise volume failover path.
Having the agent startup files on the same GroupWise volume as the domain or post office makes them easy to find.
When you update the agent software, you only need to update it in one place for a particular domain or post office, not on every node on a GroupWise volume failover path. This prevents the potential problem of having a domain or post office fail over to a location where a different version of the agent software is installed.
If you ever need to add or replace a physical server in the cluster, you only need to install NetWare and Novell Cluster Services to the new server, then add that node to the appropriate failover paths. No extra GroupWise configuration is necessary because there are no sys:\system dependencies for the GroupWise agents.
If you want to back up the GroupWise software, you do not have to include the sys:\system directory in the backup.
Installing the agents on a GroupWise volume does have some disadvantages:
GroupWise administrators who are used to the GroupWise agents being installed in sys:\system might be confused by not finding them there in the clustered GroupWise system.
You must remember where you installed the GroupWise agents when you update the agent software. Accidently installing a GroupWise Support Pack to the default location of sys:\system would not have the desired results if the original agent software was installed to the vol:\system directory on a GroupWise volume.
Whichever method you choose, be consistent throughout the entire cluster. Either put all the GroupWise agents on the GroupWise volumes with the domains and post offices they service, or put them all in sys:\system directories. If you put them on GroupWise volumes, make sure there are no agent files in sys:\system directories to confuse the issue at a later time.
Even if you choose to install the agents to multiple sys:\system directories, you can still store the agent startup files on the GroupWise volumes. The significant advantage of this approach is that you only have one startup file to modify per agent.
|
AGENT CLUSTERING WORKSHEET |
|---|
|
Under Item 1: Agent Installation Location, mark whether you will install the agent software to a vol:\system directory on a GroupWise volume or to sys:\system on each node in the cluster. If necessary, specify where the agent startup files will be stored. Under Item 2: Domain Name, transfer the domain name and location from the System Clustering Worksheet to the Agent Clustering Worksheet. Under Item 5: Post Office Name, transfer the post office name and location from the System Clustering Worksheet to the Agent Clustering Worksheet. |
On a NetWare server, using protected memory allows you to create isolated memory spaces where NLM programs can run without affecting other NLM programs running on the same node. This contributes to the high availability of the cluster. Using protected memory has the following advantages:
When using protected memory, the node can restart a specific memory space if any NLM program within that memory space abends. This allows for recovery without failing the entire node, which enhances both up time and database integrity.
Using protected memory gives you the ability to unload a single instance of an agent, rather than all instances.
If you use protected memory, you can run the agents in active/active mode, rather than active/passive mode.
If you have any possibility of the same type of GroupWise agent loading multiple times on any node in the cluster, you must use protected memory so that you can unload agents individually. Check your failover paths (Agent Clustering Worksheet items 3 and 6) for failover combinations where multiple instances of the same type of agent might need to run on the same node.
Protected memory does result in higher memory utilization (about 5% to 10%) and a slight performance penalty. Make sure your nodes have sufficient memory to handle the number of memory spaces that might reside on them. Keep in mind that if you load the MTA and the POA in different memory spaces, the agent engine (gwenn5.nlm) will load twice on the node. Remember to provide memory for any GroupWise volumes that could fail over to a node, in addition to that node’s regular processing load.
IMPORTANT:For optimum stability, we strongly recommend that you run the agents in protected memory, with one agent per memory space.
|
AGENT CLUSTERING WORKSHEET |
|---|
|
Under Item 8: Load Agents in Protected Memory?, mark whether or not you need to run the GroupWise agents in protected memory. If you will use protected memory, provide one or two unique protected memory space names. If you will create the domain and post office on the same GroupWise volume, the MTA and POA can use the same memory space, although this is not recommended. If you will create the domain and post office on different GroupWise volumes, the MTA and POA must use different memory spaces. |
Aside from the cluster-specific issues discussed in the preceding sections, the considerations involved in planning to install the GroupWise NetWare agents are the same in a clustering environment as for any other environment. Review Planning the GroupWise Agents
, then print and fill out the GroupWise Agent Installation Summary Sheet
in Installing GroupWise Agents
in the GroupWise 8 Installation Guide for each location where you will install the NetWare MTA and/or POA.
Fill out the NetWare Agent Worksheet, taking into account the following cluster-specific issues:
|
GROUPWISE AGENT INSTALLATION WORKSHEET |
|---|
|
Under Agents and Locations, mark POA Local to Post Office and MTA Local to Domain. In a clustering environment, a domain or post office and its agent must reside on the same GroupWise volume in order to fail over together. Under Installation Path, take into account your decision based on Deciding Where to Install the Agent Software. Under Installation Options for your agent platform, mark Yes for Configure the GroupWise Agents for Clustering. This will cause the Agent Installation program to customize the agent startup files for clustering. Under Domain Information and Post Office Information, refer to the Domain and Post Office Worksheets you filled out in Section 2.3, Planning a New Clustered Domain and Section 2.4, Planning a New Clustered Post Office, and to the IP Address Worksheet you completed during Planning Secondary IP Addresses and Cluster-Unique Port Numbers for Agents in the Cluster. Under Launch GroupWise Agents Now, mark No. |
IMPORTANT:Do not install the NetWare agent software until you are instructed to do so in Section 3.0, Setting Up a Domain and Post Office in a NetWare Cluster.
Skip to Section 3.0, Setting Up a Domain and Post Office in a NetWare Cluster.