Chapter 2 Runtime Issues

This chapter describes runtime issues that are known to be problems. You can find the Sun Management Center 3.5 documentation on the Sun Management Center 3.5 Software 2 of 2 CD. The information included in these Release Notes supersedes the information in the Sun Management Center Release Notes on the Sun Management Center 3.5 Software 2 of 2 CD. The Sun Management Center 3.5 Release Notes can be found in the Sun Management Center 3.5 Software Collection on http://docs.sun.com.

The following runtime bug descriptions have been added to this chapter since this document was published on the Sun Management Center 3.5 Software 2 of 2 CD.

• Netscape 4.7x Displays Pages Incorrectly in Simplified Chinese Locales (zh_CN.GBK and zh_CN.GB18030) (4864462)

• Netscape 4.6x Displays Help Pages Incorrectly in Traditional Chinese Locale (zh_TW.UTF-8) (4880488)

Sun Management Center 3.5 Software Bug

Sun Management Center 3.5 Java Processes Silently Terminate (4770335)

If you press Ctrl+C or Ctrl+D in the terminal window from which the Sun Management Center 3.5 processes were started, the Java^TM processes terminate without displaying any message. If you attempt to launch the Java console again, a warning message indicating that a remote exception has occurred is displayed in the console window.

Workaround: Run Sun Management Center 3.5 processes from the C shell rather than the Bourne shell.

Localization Bugs

Netscape 4.7x Displays Pages Incorrectly in Simplified Chinese Locales (zh_CN.GBK and zh_CN.GB18030) (4864462)

The page text in the Simplifed Chinese locales, zh_CH.GBK and zh_CN.BG18030, is garbled when viewed in the Netscape^TM version 4.7x browser.

Workaround: Specify a different browser as your default browser. The following browsers are known to work correctly in these locales:

• Netscape 6.2.2

• Netscape 7.x

• Mozilla^TM

• Microsoft Internet Explorer

Netscape 4.6x Displays Help Pages Incorrectly in Traditional Chinese Locale (zh_TW.UTF-8) (4880488)

Pages displayed by clicking the Help button in the Traditional Chinese locale (zh_TW.UTF-8) are garbled when using the Netscape version 4.6x browser.

Workaround: To avoid this problem, specify a different default browser, such as Netscape 6.x or later.

Web Browser Console Is in English When Browser's Language Is Simplified Chinese (zh) (4863879)

When you set your web browser's language to Simplified Chinese (zh), the Sun Management Center 3.5 web console displays English content. This behavior occurs even when the Simplified Chinese locale is installed.

Workaround: Set your web browser's language to zh-CN for the Netscape Navigator^TM browser or to zh-cn for Microsoft Internet Explorer.

Sun Fire Link Traditional Chinese Help Buttons Point to Incorrect Locale (4856610)

For Traditional Chinese users, some of the Sun Fire^TM Link Help buttons are displayed in Simplified Chinese. Help pages that are displayed in Simplified Chinese instead of Traditional Chinese include the Manage All Partitions and Manage Registered Nodes dialog boxes buttons.

Workaround: Use the Help button on the Fabric Details console to view the correct help pages. The Fabric Details console is the main graphical user interface (GUI) for Sun Fire Link.

Hardware Diagnostic Suite Buttons Might Be Truncated on French-Locale Machines (4862340)

When viewing the Hardware Diagnostic Suite console on a French-locale machine, two of the buttons might be truncated and difficult to read. These buttons should say Arreter les test and Controle Rapide.

Workaround: None.

Hardware Diagnostic Suite Software Bugs

Incorrect Error Message Displayed if JAVA_HOME Path Is Set to Value Longer Than 80 Characters (4854768)

If the JAVA_HOME path is set to a value longer than 80 characters, the following incorrect error message is displayed:

Could not start Hardware service

Workaround: This error message can be ignored. The Hardware Diagnostic Suite starts correctly.

Slow Memory Leak Might Cause Sun Management Center 3.5 To Crash (4768232)

When running multiple long-term processes on a server through the Hardware Diagnostic Suite console, a slow memory leak might occur. If additional free memory is not available on this server, Sun Management Center 3.5 might crash and possibly dump core.

Workaround: None. The server must be restarted.

Performance Reporting Manager Software Bugs

Problem With Java Service Causes Performance Reporting Manager to Fail to Collect Agent Data (4855306)

If 15 or more Performance Reporting Manager report requests are scheduled to run at the same time, you might encounter one or more of the following problems after the software has been running for a period of time:

• Data is not collected from agent machines.

• Some report requests remain in a run state indefinitely.

• Report requests cannot be scheduled as desired.

• Grouping task requests cannot be scheduled as desired.

• The Report Manager GUI cannot be launched.

• The Manage Jobs GUI cannot be launched.

For some of the previously mentioned problems, the console might display the following error message when you launch the Manage Jobs GUI:

Task Service is not available

Similarly, for some of these problems, the console might display the following error message when you launch the Report Manager GUI:

Report Service is not available

Workaround: When you encounter this problem, you must stop and restart your server components. Use the following commands:

# es-stop -A

# es-start -A

To avoid this problem, do not schedule more than 15 report requests to run at the same time. It is recommended that you space the run time of scheduled requests out over a period of time.

For example, if you have 20 requests required to run every hour, split the requests up into two separate batches, each containing 10 requests. Schedule the batches to run at 15 minute intervals. Do not schedule the batches to run at the same time.

Performance Reporting Manager Setup Rejects Directory Names That Have Mount Point Longer Than 18 Characters (4856682)

During setup of the Performance Reporting Manager server layer, you cannot specify a directory name that has a long mount point name.

For example, if you use the following mount point:

/Sun_MC_3.5_PRM_Data

and then specify this mount point as the target directory when prompted for three directories during the server layer setup, the directory name is rejected.

The following error message is displayed:

The directory must be on a local filesystem.

This error is displayed only when the directory name that is specified for data files has a mount point name that is longer than 18 characters.

If the mount point name is short but the directory name is long, the problem does not occur.

For example: if the mount point name is /PRM, the setup will work.

Workaround: Ensure that the directory's mount point is less than or equal to 18 characters.

Alarms Reports Not Filtered Correctly When All Hosts Option Is Selected (4788475)

Alarms logged against agent machines are not filtered correctly in Alarms Reports when the All Hosts option is selected and a filter is specified.

When the All Hosts option is selected along with a filter, the specified filter criteria is applied to all the alarms that are logged for all the hosts. The filtered alarms are returned in the report output. Additionally, the report includes any agent not responding alarms and host not responding alarms from all monitored hosts. The specified filter criteria is not applied to such alarms.

Workaround: This problem can be avoided by explicitly specifying host names for Alarms Reports. Choose from the following workarounds:

• When generating an Alarms Report for a single host, either select the host name by providing the hostname:port, or select the host name from the Sun Management Center topology view.

• When generating an Alarms Report for multiple hosts, either provide the different host names as hostname:port in a comma–separated list, or select the multiple host names from the Sun Management Center topology view.

Service Availability Manager Software Bugs

Unable To Launch Modify Service Object Window or Service Manager Window if EService Object Is Created With Fully Qualified Host Name (4852117)

If you provide the fully qualified host name during the creation of an EService Object, the object will be created. However, you will not be able to launch the Modify Service Object window or the Service Manager window from the Sun Management Center 3.5 GUI after the EService Object is created.

For example:

host-name.domain

Workaround: During the creation of an EService Object, only provide a host name. Do not specify a domain.

Modify Service Object Dialog Box Might Not Accept Characters on Microsoft Windows 2000 Systems (4738717)

The Modify Service Object dialog box on a Microsoft Windows 2000 system might not accept characters in certain text fields when you are running the Java runtime environment, version 1.3.1 to version 1.4.0. When you attempt to change the service type value from the default of HTTP, characters provided for the hostname and instance text fields are not accepted.

This problem does not occur if you are running the Java runtime environment version 1.4.1.

Workaround: Before using the Modify Service Object dialog box, install the Java runtime environment version 1.4.1 on Microsoft Windows 2000 systems.

System Reliability Manager Software Bug

OS Crash Dump Module Is Not Supported on Solaris 9 Operating Environment (4841420)

The Solaris 9 operating environment does not support the OS Crash Dump module of the System Reliability Manager add-on product.

Workaround: None.

Sun Enterprise 6500/5500/4500/3500 Server Bugs

Increased Memory Consumption by Config Reader (4808821)

In some Sun Enterprise^TM systems, the Config Reader increases memory consumption over time.

Workaround: Choose from one of the following workarounds:

• Restart the agent every one or two months.

• Edit the agent file to slow down the refresh interval.

  You can change the system nodes refresh interval in the /opt/SUNWsymon/modules/cfg/Config-Reader4u-d.x file from 300 seconds to 900 seconds. This change slows downs the data refresh interval to every 15 minutes, which enables the agent to run for up to six months before needing to be restarted.

Tape Drive Errors Might Be Displayed When Running Config Reader (4846230)

If you run the Sun Management Center Config Reader on a system with an IBM LT0 tape library attached, the following tape error might be displayed in intervals:

Periodic head cleaning required.

Workaround: Ignore the error messages. The tape library and Sun Management Center software will continue to run.

Sun Fire 6800/4810/4800/3800 Systems Bugs

Minor Memory Leak in Domain Config Reader (4862493)

The Domain Config Reader has a minor memory leak. This memory leak should not impact the use of the software.

Workaround: Lengthen the refresh interval.

Failed Power Supply Causes Black Splat to Display in Place of Power Supply Table (4844612)

If one of the power supplies in Sun Fire^TM 6800/4810/4800/3800 systems fails, the loss of power causes the Platform Administration Module to cease displaying any of the power supplies. A black splat is displayed in place of the Power Supply table. The Hardware Physical and Logical Views display only the system controller and fan trays. The remaining components, such as system boards and power supplies, cease to be displayed.

Workaround: To correct the problem, replace the failed power supply.

Sun Fire 15K/12K Systems Bugs

Black Splat Displayed in System Table When Failover Is in Activating Status (4852731)

A black splat that indicates an indeterminate status is displayed in the System Table when a failover is in the Activating status.

Workaround: Ignore the black splat. The functionality of the system is not affected. After the failover completes, the black splat is no longer displayed.

Memory Leak in System Controller Platform Agent (4860281)

The platform agent running on the Sun Fire 15K/12K system controller (SC) has a memory leak. The amount of leakage is minimal and will not noticeably affect the system.

Workaround: If you discover that the memory leak has increased in size to more than 10 megabytes (MB), stop and restart the platform agent on the SC.

Platform Icon Might Not Be Visible on Right Panel of Sun Management Center Console (4864183)

After creating a Sun Fire 15K/12K platform object, the platform icon view, which is located on the right panel of the Sun Management console, might no longer be visible.

Workaround: Choose from the following two workarounds:

• If you have no need to use the platform icon, use the heierarchy that is located on the left panel of the Sun Management Center console for navigation.

• If you prefer that the platform icon is visible, use the following command to copy the icon file into the appropriate directory:

  # cp /opt/SUNWsymon/classes/base/console/cfg/topoimages \ sf15000-platform-xlarge.gif /var/opt/SUNWsymon/cfg/deviceimages \ Starcat-sf15000-platform-xlarge.gif

  ---

  Note –

  The new file name is different than the original file name.

  ---

Platform Agent on System Controller Stops When System Management Services Is Stopped and Cache Refresh Occurs (4822944)

If System Management Services (SMS) is stopped on the System Controller (SC), the Sun Management Center platform agent will also stop when the agent's next automatic cache refresh occurs, typically once every hour. This behavior occurs only when you start and stop the platform agent repeatedly for very short periods of time.

For example, if you repeatedly stop the platform agent for one minute, and start the platform agent for two minutes, this behavior would occur.

Workaround: Delay the frequency of starting and stopping the platform agent. For example, stop the agent for three minutes and start the agent for five minutes. If the Platform agent does stop, restart SMS, and then restart the Sun Management Center platform agent.

Sun Fire Link Bugs

Node Failure During Link Bring-Up Might Cause Sun Management Center Console and Fabric Details Windows to Freeze (4857941)

If a node fails after a cluster has been configured, but before all the links have been brought up, the Sun Management Center console and Fabric Details windows might freeze.

Workaround: If a node fails while you are creating a partition and the Cancel button on the Add Node dialog box does not work, perform the following steps:

1. To stop the fabric, use the command stopfabric with the -f option.

   # /opt/SUNWwcfm/stopfabric -f fabric-name

2. Restart the stopped fabric:

   # /opt/SUNWwcfm/startfabric fabric-name

3. Resume managing the fabric from the Sun Management Center console.

Striping Level 1 for Sun Fire Link Switch Configuration Not Supported (4858050)

If you select a striping level of 1 for a Sun Fire Link Switch configuration, the striping level is automatically upgraded to a 2. However, that level is not reflected in the Fabric Details window of the Sun Management Center console.

Workaround: Do not select a striping level of 1 for a Sun Fire Link Switch. A striping level of 1 is not supported for a Sun Fire Link Switch configuration.

Sun Management Center Is Slow to Update Link Status When a Cluster Node Goes Down (4644785)

If a cluster node goes down while the link status is being displayed in the Fabric Details window, the Sun Management Center console might not update the status of the affected links for an hour or more. Moreover, if the cluster node is restored, and the links are re–established, the Fabric Details window might continue to display the links as down.

Workaround: Use the Refresh button to manually refresh the status information.

Sun Fire Link Interconnect Fabric Name Is Not Updated When a New Fabric Is Loaded (4771624)

If you unload and stop a fabric, and then load another fabric with a different name, the Sun Fire Link (SFL) object icon that is displayed in the Sun Management Center console window will retain the name of the removed fabric. However, the Fabric icon label will show the name of the new fabric.

Workaround: After unloading and stopping the fabric you want to replace, delete the corresponding composite object.

Configured Node Might Be Missing Under Sun Fire Link Object (4855154)

A cluster node that has been successfully configured into a fabric might not be displayed under the Sun Fire Link object for the same fabric.

Workaround: Before you add nodes to a fabric, you must discover all cluster nodes by using the Sun Management Center console. If you do not discover all cluster nodes before you add the nodes to the fabric, problems are encountered. This bug is one example of the types of problems you might encounter.

Manage Registered Nodes Menu Option Might Not Be Displayed (4855259)

If you access the fabric icon pop-up menu by pressing mouse button 3 from within Sun Management Center's Fabric Details window, the Manage Registered Nodes menu option might not be displayed.

Workaround: Use the Fabric menu to select the Manage Registered Nodes option instead.

Switch Physical View or Logical View Alarm Color Not Shown (4855551)

When an alarm is triggered in the Sun Fire Link switch module, the switch physical view or logical view (PV/LV) does not show the alarm color.

Workaround: You can view the correct alarm colors in the module's browser and the alarms table.

Scroll Bar Not Displayed on Right Side of Fabric Details Window (4856536)

The Fabric Details window does not include a scroll bar in the large icon view of the Chassis & Links perspective. The scroll bar should always be displayed when the image is larger than the Fabric Details window. This visibility allows viewing of all the nodes in a large partition.

Workaround: While in the Chassis & Links perspective, change from the large icon view to the small icon view. The scroll bar will be displayed. Then change back to the large icon view. The scroll bar will remain displayed.

Changing to the Nodes & Routes perspective from the large icon Chassis & Links perspective will also display the scroll bar. However, when you change back to the large icon Chassis & Links perspective, the scroll bar will no longer be available.

Empty Grey Box Might Be Displayed in Fabric Details Window for Chassis & Links Perspective (4856884)

In the Sun Management Center Fabric Details window, when you're in the Chassis & Links perspective, an empty grey box might be displayed in the following instances:

• Before you have added a switch to a newly created partition object.

• After you have deleted the last switch from a single partition when unconfiguring the fabric.

Workaround: Ignore the empty grey box. If you add a switch to that partition, the grey box will contain a switch graphic.

Fabric Manager Remote Shared Memory Proxy Has Memory Leak (4863270)

The Sun Fire Link Fabric Manager Remote Shared Memory (WRSM) proxy leaks memory whenever the proxy receives an event from the kernel system event notification daemon, syseventd(1M).

Workaround: Stop and restart the WRSM proxies on each node. Use the following commands:

# /etc/init.d/wrsm_proxy stop

# /etc/init.d/wrsm_proxy start

Data Missing From Route Properties Window (4771419)

When displaying route properties data in a Fabric Details context, some data might be missing from the To and From fields of the table.

Workaround: None.

Bug That Applies to Sun Fire 15K/12K Systems and Sun Fire Link Systems

Attempts to Remove a Node From Sun Fire 15K/12K Chassis Might Fail When Chassis Has Multiple Nodes in Same Partition (4856525)

When a Sun Fire 15K/12K chassis has two or more nodes configured into the same partition, attempts to remove the nodes from the partition might fail if that chassis is the only chassis with multiple nodes in that partition. This problem is caused by the need to create memory mapping aliases when multiple nodes from a chassis are configured into the same partition. This aliasing is done independently by every chassis that contributes multiple nodes to a common partition. When you remove all but one of a chassis's nodes from a partition, the aliasing created for that chassis will become invalid. Consider the following example:

• Begin with a partition consisting of three nodes:

  xc1-a, xc1-b, xc2-a

  where xc1 and xc2 are two Sun Fire 15K/12K systems.

• If you remove either xc1-a or xc1-b from the partition, the memory map aliasing for the xc1 system will become invalid.

However, if the partition includes two or more chassis's that contribute multiple nodes, the memory map aliasing created for the second chassis will compensate for the loss of aliasing in the first chassis. This compensation, which allows continued use of the partition, is shown in the following example:

• Begin with a partition consisting of four nodes:

  xc1-a, xc1-b, xc2-a, xc2-b

• If you remove either xc1-a or xc1-b from the partition, the memory map aliasing for the xc1 system will become invalid. However, the aliasing for chassis xc2 will remain valid. The partition can continue operation.

Workaround: The following procedure explains how to reduce to one the number of nodes a Sun Fire 15K/12K chassis contributes to a chassis, when the chassis is the only multiple-node chassis in that partition.

1. On the chassis with multiple nodes in the partition, remove all the chassis's nodes from the partition.

2. Add back the one node you want to include in the partition.

   For example, if you have a partition consisting of the nodes: xc1-a, xc1-b, and xc2-a, where xc1 and xc2 are two Sun Fire 15K/12K systems, and you want the partition to include only nodes xc1-a and xc2-a, proceed as follows:

3. Remove xc1-a and xc1-b from the partition.

4. Add xc1-a back into the partition.

Volume Systems Products (VSP) High-End Entry Servers (Workgroup Servers) Issue

Sun Ultra 5 Workstation Might Identify Itself as Sun Ultra 10 Workstation After Upgrade With Agent Update Utility

If you use the Agent Update utility to update the Sun Management Center 3.5 agent on a Sun Ultra^TM 5 workstation, the workstation might identify itself as a Sun Ultra 10 workstation after the upgrade. As a result, the icon and physical view for the Sun Ultra 5 display in Sun Management Center windows as a Sun Ultra 10.

Workaround: To correct the problem, complete the following steps:

1. At the system prompt, use the es-stop command to stop the Sun Management Center agent.

   # es-stop -a

2. Change directories to the /var/opt/SUNWsymon/platform directory.

   # cd /var/opt/SUNWsymon/platform

3. Use a text editor to edit the platform prop file.

   1. Change the UltraModel=10 entry to UltraModel=5.

   2. Save the file.

4. At the system prompt, use the es-start command to start the Sun Management Center agent.

   # es-start -a

VSP High-End Entry Servers (Workgroup Servers) Bug

Some Platform Types Not Available in Discover Objects Menu (4822174)

When you perform a discovery request and choose to discover platforms by platform type, the available platforms appear in a drop-down menu. This drop-down menu is located under the Filters tab in the New Discovery Request dialog box.

The drop–down menu does not include entries for the following Sun Enterprise Ultra servers:

• Sun Enterprise Ultra 2

• Sun Enterprise Ultra 5

• Sun Enterprise Ultra 10

Workaround: If you need to perform a discovery request for any of the preceding Sun Enterprise Ultra servers, perform the request manually.