Automatically restarting server processes

There are several server processes that the operating system can monitor and automatically restart when the server processes stop abnormally.

Before you begin

To set up this function on a Linux® or supported UNIX operating system, you must have root authority to edit the inittab file.

To set up this function on a Microsoft Windows operating system, you must belong to the Administrator group and have the following advanced user rights:

• Act as part of the operating system
• Log on as a service

The Installation wizard grants you the user rights if your user ID is part of the administrator group.

If you are running on a Microsoft Windows Operating System, the Installation wizard displays a message that states that although the advanced user rights are now effective, they do not display as effective until the next time you log on to the Windows machine.

You can also add the advanced user rights manually if you are performing a silent installation on a Windows operating system. For example, to grant the user rights to your administrator group user ID on a Windows operating system, perform the following procedure:

1. Click Administrative Tools in the Control Panel.
2. Click Local Security Policy.
3. Click Local Policies.
4. Click User Rights Assignments.
5. Right-click Act as part of the operating system.
6. Click Security.
7. Click Add.
8. Click your user ID.
9. Click Add.
10. Click OK.
11. Click OK.
12. Right-click Log on as a service.
13. Click Security.
14. Click Add.
15. Click OK.
16. Click OK.
17. Reboot your machine to make the settings effective.

Consult your Windows help system for more information.

About this task

There are several environments where you might use this function of automatically restarting servers. You can restart the server1 managed node process, for example. Here is a list of processes you might consider restarting:

• The server1 managed node process
• The server1 process on a stand-alone Application Server
• The dmgr process on a deployment manager node
• The nodeagent server process on any managed node
• The IBM® HTTP Server process
• The IBM HTTP Administration process

On a Windows operating system, you can create Windows services during installation, using the installation wizard. Each Windows service controls a single process, such as a stand-alone product instance. Multiple stand-alone Application Server processes require multiple Windows services, which you can define. The wizard lets you create services for these servers:

• The server1 managed node process, defined as a manually started (versus automatic) service
• The server1 stand-alone Application Server process, defined as a manually started service
• The IBM HTTP Server process and the IBM HTTP Administration process, defined as automatically started services when you choose to install the IBM HTTP Server feature
• The dmgr process on a deployment manager node, defined as a manually started service

The installation wizard does not provide a way to create a service for a node agent because the deployment manager instantiates each node agent after installation when you add an Application Server node to the deployment manager cell. For this reason, you must manually create a function that automatically starts a failed node agent server process.

On a Linux or supported UNIX operating system, you must manually create a shell script that automatically starts any of the processes previously mentioned. Each UNIX shell script controls a single process, such as a stand-alone product instance. Multiple stand-alone Application Server processes require multiple UNIX scripts, which you can define.

In a WebSphere® Application Server Network Deployment environment, the addNode or startNode command starts a single unmonitored node agent only, the nodeagent process, and does not start all of the processes that you might define on the node. While running, the node agent monitors and restarts Application Server processes on that node, on either a Windows or a Linux and UNIX-based platform. Each Application Server process has MonitoringPolicy configuration settings that the node agent uses when monitoring and restarting the process.

It is recommended that you manually set up a monitored process for the deployment manager dmgr server and for any node agent defined for your system. To set up a monitored process:

• Use a Windows service. You can install the WebSphere Application Server Network Deployment product as a Windows service during installation, or at a later time.
• Use the WASService command. For more information, see the documentation for the WASService command. As an alternative, you can use the rc.was example shell script.
  Supported configurations: On the Red Hat Enterprise Linux Version 6 operating system, do not use the rc.was example shell script. Instead, follow the specific steps in this topic that are related to this operating system.
• Use the rc.was example shell script that is provided with the appropriate version of the product.

1. Use the Profile Management tool to set up a Windows service to automatically monitor and restart processes related to the product.
   • Perform the following procedure from the Profile Management tool to select services that the installation wizard can set up:

     1. Click Run WebSphere Application Server Network Deployment as a service.
        If you select this option, the installation wizard creates the following service during installation:

        IBMWAS6Service - node_name

        IBMWAS6Service - node_name service controls the node_name process.

        After you complete and verify the installation, use the Windows Services panel to change the IBMWAS6Service - node_name service to an automatic startup type.

        1. Right-click IBMWAS6Service - node_name and click Properties.
        2. Click Automatic from the Startup type list box and click OK.
     2. Click Run IBM HTTP Server as a service.

        Select this option on the machine where you are installing the IBM HTTP Server.

        If you select this option, the installation wizard creates the following services during the installation:

        • IBM HTTP Server 2.0.x
        • IBM HTTP Administration 2.0.x

        The installation wizard defines the startup type of these services as automatic. It is not necessary for you to change the type from manual to automatic.
     3. Enter your user ID and password and click Next.

     In a coexistence environment, you can change the default service names to make them unique. In a same version coexistence scenario for IBM HTTP Server 2.0.x on a Windows platform, you cannot use the default service names created by the installer because they are common.

     To work around this problem:

     1. Install the first copy of IBM HTTP Server, either by itself or with the product, and select to install the services.
     2. Customize the service names for the first install by running the following commands from the first install location:

             apache -k install -n "IHS 2.0(1)"
             apache -k install -f conf\admin.conf -n "IHS 2.0 Administration (1)" 

     3. Edit the AdminAlias directive in the installLocation 1\conf\admin.conf file to point to the new service name, such as IHS 2.0(1).
     4. Remove the default service names installed by the first install by running the following commands:

             apache -k uninstall -n "IBM HTTP Server 2.0"
             apache -k uninstall -n "IBM HTTP Administration 2.0"

     5. Install the second copy of IBM HTTP Server, either by itself or with the product. The default service names correspond to the second install.
     Avoid trouble: Customized service names must be unique on your system.
2. After you install the product, set up a shell script to automatically monitor and restart any related server processes.
   Supported configurations: There are two sets of substeps to complete this task. The first set of substeps do not apply on the Red Hat Enterprise Linux Version 6 operating system. For more information on this operating system, see the second set of substeps.
   1. Locate the rc.was example shell script, which is in the app_server_root/bin directory.
   2. Create a new shell script for each process that the operating system is to monitor and restart.
   3. Edit each shell script according to comments in its header, which provide instructions for identifying a product process.
   4. Edit the inittab file of the operating system, to add an entry for each shell script you have created.

      Comments in the header of the rc.was file include a sample inittab entry line for adding this script to the inittab table. Each inittab entry causes the operating system to call the specified shell script whenever the system initializes. As each shell script runs, it monitors and starts the server process you specified.

      For example, if you create the following inittab entry for a process, the rc.was shell script is run whenever the system initializes, and if the process goes down while the system is initializing into a machine that is operating at a runlevel of 2, 3, or 5:

      was:235:respawn:/usr/WebSphere/AppServer/bin/rc.was >/dev/console 2>&1 

      If you create the following inittab entry, the rc.was shell script only runs once when you initialize into a machine that is operating at a runlevel of 2,3, or 5:

      was:235:once:/usr/WebSphere/AppServer/bin/rc.was >/dev/console 2>&1 

      Following is a list of the runlevels that can be specified. Runlevels usually default to either 3 or 5.

      • 0-halt
      • 1-Single user mode
      • 2-Multiuser, without NFS (The same as 3, if you don't have networking)
      • 3-Full multiuser mode
      • 4-unused
      • 5-X11
      • 6-Reboot

      If you don't know the runlevel into which your machine is booting, look at the following line in the inittab file:

      id:x:initdefault

      where x is the runlevel that the machine is booting into.

      For example, if your machine is booting into a runlevel of 5, then all of the processes that are declared to run with a runlevel of 5 are started.

      Avoid trouble: Everything that is ran from the inittab file runs under the root user. Therefore, if you need the server to automatically start the process under a non-root user ID when the machine starts, you must also add the following line to the inittab file:

      su user -c values

      where values is the file path and arguments that are used to call the rc scripts you created, and user is the non-root user that you have configured the product to run as.

   Each shell script monitors and restarts the following processes in a WebSphere Application Server Network Deployment environment:

   • A server process on a managed node
   • A node agent process on a managed node
   • A stand-alone Application Server process
   • A deployment manager process
   Note: The following set of substeps apply to the Red Hat Enterprise Linux Version 6 operating system.
   1. Create the as1.conf file to monitor and restart processes and place it in the /etc/init directory.
      The following code is example content for the as1.conf file:

      # upstart service for my application server profile
      
      description  "my application server"
      
      start on runlevel [235]
      stop on runlevel [!235]
      
      expect fork
      respawn
      
      exec /opt/IBM/WebSphere/AppServer/bin/rc.as1

   2. Create the rc.as1 file to restart the server and place it in the install_root/bin directory for WebSphere Application Server.
      The following code is example content for the rc.as1 file:

      #!/bin/sh
      cd /opt/IBM/WebSphere/AppServer/profiles/profile_name/bin
      ./startServer.sh server_name

      profile_name represents the name of your application server profile. server_name is the name of your server. Change these values to match your environment.

3. After installing the product, use the WASService.exe command to manually define the nodeagent server process as a Windows service.

   You can use the same command to manually define a Windows service for another installation instance or for another configuration instance of either the server1 process or the dmgr process.

4. In the administrative console, click Servers > Server Types > WebSphere application servers > server_name > Process.
   Start the administrative console. In the topology tree, expand Servers and click Application Servers. Click the name of the application server that you want to start automatically. Click Process Definition. Click Monitoring Policy. Change the Node Restart state to RUNNING. Click Apply. Save the configuration.
5. Select the application server that you want to automatically restart and then, under Server Infrastructure , click Java and process management > Monitoring Policy.
6. Select Automatic restart.
7. Click Apply and then click Save to save the change directly to the master configuration.

Results

On a Windows operating system, you can

• Use the net start and net stop commands to control the IBM HTTP Server services on a Windows system. For more information about these commands, see the Windows help file. Access these commands from the Start menu, clicking Start > Programs > IBM HTTP Server.
• Use the Start the Server and Stop the Server commands to control the product process. Access these commands from the Start menu, clicking Start > Programs > IBM WebSphere > Application Server.
• Use the Start the Manager and Stop the Manager commands to control the WebSphere Application Server Network Deployment dmgr process. Access these commands from the Start menu, clicking Start > Programs > IBM WebSphere > Application Server > Deployment Manager.

Processes started by a startServer command, a startNode command, or a startManager command are not running as monitored processes, regardless of how they are configured.

For example, you can configure a server1 process as a monitored process. However, if you start the server1 process using the startServer command, the operating system does not monitor or restart the server1 process because the operating system did not originally start the process as a monitored process.