Chapter 3 Troubleshooting

Installation Troubleshooting

Console or Server Cannot Communicate With the Agent

Verify that the agent has been seeded. Check for the existence of the file /var/opt/PrimeAlert/cfg/agent-engine-d.dat on the Linux Agent host. You must wait 5 - 10 minutes after starting a newly installed or re-seeded agent before it and the server authenticate one another.

Remember that you must add the Linux Agent to the topology first as an SNMP Ping type object, then change that object to a Sun Management Center Host type object.

To seed the Linux Agent host, see Reseeding After Changing Host Names or Port Numbers.

No other processes need to be restarted on any other hosts. Wait 5 – 10 minutes before using console actions. If you have changed the IP address or SNMP Port of the agent process, you might need to delete the existing topology host object and re-create it.

The console or server cannot communicate with the agent.

Runtime Troubleshooting

This list describes some of the problems you might encounter.

Objects Have Long Cryptic Strings

Problem: When you view the Linux Agent in the console, you might see objects with long cryptic strings, such as ca.halycon. primealert.modules.LinuxSystemAlert.LinuxSystemAlert:moduleName.

The server component of the Linux Agent is not installed on the Sun Management Center server host.

Workaround: Install the server component of the Linux Agent. See Chapter 1, Installing the Linux Agent.

Error Message: Unable to load module definition

After you select PrimeAlert SystemMonitor for Linux in the Load Module window, the Module Loader dialog box appears empty, and this message appears: Unable to load module definition.

Problem: The server component of PrimeAlert SystemMonitor for Linux is not installed on the server host that the console is connected to.

Workaround: Install the server component of PrimeAlert SystemMonitor for Linux. See Chapter 1, Installing the Linux Agent.

Linux Agent Cannot Be Added to the Domain

Problem: When you try to add a Linux Agent to a Sun Management Center domain, the Linux Agent cannot be added to the Domain.

This error message appears: Error occurred during node create.

Workaround: Follow the procedure for creating a Linux object inTo Create a Linux Object As an SNMP Ping Object.

Cannot Create a Sun Management Center Host Type Object

Problem: You cannot create a Sun Management Center host type object for the Linux Agent in the console topology.

Workaround: This issue has multiple possible causes and solutions:

• Ensure that you are creating an SNMP Ping type object first and then changing that object to a Sun Management Center Host type topology object.

• Ensure that you are logged into the Sun Management Center server layer as a user with sufficient privileges to create a topology object.

• Ensure that you have waited long enough to give the Sun Management Center server and the Linux Agent a chance to authenticate with one another. This period might be longer in a larger environment. Start with 5 minutes, then increase to 15 minutes if all else fails.

• Verify that you can ping the Linux host from the Sun Management Center server layer using both the IP address and the host name.

• Verify that you can ping the Sun Management Center server host from the Linux host using both the IP address and the host name.

• Verify that the host name resolves to the correct IP address on the Sun Management Center server.

• Verify that the IP address that you provided to the Sun Management Center server layer is on the same subnet as the Ethernet card containing the IP address being used by the Linux Agent.

• If the Linux host has multiple NIC cards, ensure that the host name on the Linux host resolves to the IP address for the proper NIC. If that resolution fails, modify the agent subsection of the file /var/opt/PrimeAlert/cfg/domain-config.x on the Linux Agent host:

  agentServer = IP-address

  where IP-address is the IP address of the appropriate NIC.

Runtime Issues and Bugs

Issues include information that you should know about, such as prerequisites, tips, troubleshooting hints, and bugs. Bugs are a subset of issues. Bugs have tracking numbers that are shown in parentheses. For updates on bugs and to obtain patches, see the SunSolve^SM web site at http://sunsolve.sun.com.

Runtime Issues

Attempting to Stop the Agent With the pa stop Command Results in the Message agent process (PID) error waiting for process to exit: child process lost (is SIGCHLD ignored or trapped?)

Workaround: You can safely ignore this message. To confirm that the agent has been stopped, run the command ps -elf | grep paAgent | grep -v grep.

Selecting PrimeAlert System Monitor > User Statistics > Console User value Always Reports None

The user whose TTY device is console, tty1, or :0 is reported as the console user. Either no user is logged into the console, or the TTY is not console, tty1, or :0.

Workaround: None.

The pa stop Command Indicates That the Agent Has Been Stopped, But the paAgent Process Still Exists

This rare condition is linked to multiple simultaneous attempts to start the agent.

Workaround: Kill the PID of the paAgent process. If that fails, use the kill -9 command.

Runtime Bugs

When PrimeAlert System Monitor Reports on Idle Processes, the Number of Sleeping Processes Does Not Match the Number Reported as State S in ps -elf (5009340)

The number of idle processes as reported in the PrimeAlert System Monitor module is the number of sleeping processes that are swapped in ps -elf. The arguments of such processes are encased in brackets ([] ). The number of sleeping processes reported in the System Monitor module is the number of sleeping processes minus the number of sleeping processes swapped out.

The Info Tab for the Linux Agent Host Is Missing Information (5010084)

The Info tab for the Linux Agent host in the Sun Management Center console is missing information. The netmask information is missing. Timezone is shown as unknown.

Workaround: None.

No es-setup Command on the Linux Agent to Change the Agent Setup After Installation (5007871)

Workaround: You can change settings for the Linux Agent by running the install script from the installation source directory or editing the domain-config.x file. For more information, see Reconfiguring the Linux Agent.

Linux Agent Does Not Support an Optional Install Directory (5010061)

Workaround: The agent is installed in /opt/PrimeAlert with configuration files in /var/opt/PrimeAlert. If another directory is more suitable, use a symbolic link.

If a DataView Table Is Made for a Linux Agent Property (such as Filesystem Usage/System Load), the Property Column Does Not list the Property Name But Shows N/A (5009314)

Workaround: None.