This chapter describes runtime issues and bugs in the Solaris Container Manager 1.1 (Container Manager) software.
These runtime issues might occur in the Solaris Container Manager 1.1 software.
If you use the command line interface (CLI) version of the Solaris Container Manager 1.1 installation, it is only available in English. This version is not translated. The English language aspect does not affect the installation process or functionality.
Workaround: To perform the installation in your locale, use the GUI installation (es-guiinst) and setup (es-guisetup).
If you reference the Help in the New Container wizard for the Name field, the maximum character count is 64. This count is incorrect. The maximum number of characters for the Name field is 32, as noted in the field help and the error message.
If you choose a locale other than English, the title and headings of the Utilization Graph image appear in English. These values are not localized.
The Extended Accounting feature is disabled in Container Manager on systems that run the Solaris 8 OS. As a result, the Extended Accounting Data (Exacct) column of the CSV file for a container on a Solaris 8 system will be null.
If a project in an agent host's /etc/project database has the same project ID as a project in the Sun Management Center database but has a different project name, then the project will not be discovered by Container Manager. A unique project ID across all hosts in the same server context is a Container Manager requirement.
You should not use the projadd command to create projects. Use only Container Manager to manage projects.
Workaround: Delete the nondefault project that was created by the projadd command from the agent host's /etc/project database and create the container by using the New Container wizard in Container Manager.
These runtime bugs might occur in Solaris Container Manager 1.1 software.
When you move processes by using the Move Processes feature, unselected processes might also be moved. Hence, do not use the Move button on the Processes tab.
Workaround: To ensure that processes are not inadvertently moved, perform the following steps:
Become superuser.
Move processes to the correct project by using the following command:
% newtask -c <pid>
If an alarm condition exists, performance might be degraded. To remedy this situation, perform one of the following:
Manage alarms with Sun Management Center 3.5 Update 1b. For additional information, see Chapter 16, Managing Alarms Using the Web Console, in Sun Management Center 3.5 User’s Guide.
Modify the alarm criteria to eliminate the alarm. For additional information, see Chapter 7, Managing Alarms, in Installing and Administering Solaris Container Manager 1.1.
If an object on an agent which is now in a different server context is accessed, the following error message is displayed:
There was an error while retrieving the Container data.
Choose one of the following actions:
Return the agent to its original server context
Delete the agent from the old server's topology
Workaround: None.
An agent has crashed leaving scm-container* files in the /var/opt/SUNWsymon/cfg/ directory. When the agent restarts, it might not perform in an optimal manner. This condition might identify itself by the following error message:
Insufficient security privilege to complete the operation.
You can determine the state of the processes on the agent by typing the following command:
% ps -eaf | grep esd
Workaround: To improve performance, type the following commands:
Become superuser.
% su -
Stop the agent.
# /opt/SUNWsymon/sbin es-stop -a
Delete the state files.
# rm /var/opt/SUNWsymon/cfg/scm-container*
Restart the agent.
#/opt/SUNWsymon/sbin/es-start -a
Following a server reboot, the Java Web Console does not start automatically.
To start the Java Web Console, type the following commands:
Become superuser.
% su -
Type root password.
Restart the Java Web Console.
# /usr/sbin/smcwebserver restart
Ensure that Java Web Console restarts in the future.
# /usr/sbin/smcwebserver enable
Workaround: None.
If you select the Cumulative Utilization by Object option, graphs do not display for the following objects and time intervals:
Project object with Last Week time interval chosen
All objects with Last Month time interval chosen
Workaround: None.
A Resource Change Job has been previously scheduled, executed, and completed normally. If you update the Resource Change Job, the job fails.
Workaround: Instead of updating the Resource Change Job, delete it and create a new job.
If an invalid locale is selected from the Locale drop-down menu in the New Zone wizard, zone creation fails. The Locale drop-down menu might contain invalid locale values.
To determine valid locales, go to Supported Locales in International Language Environments Guide.
Workaround: Choose a valid locale from the Locale drop-down menu in the New Zone wizard. If you are unsure of a locale value, select C for the English locale.
If you use a breadcrumb to navigate to any pane and then start a wizard, an Application Error occurs.
Workaround: Do not start a wizard immediately after clicking a breadcrumb.
You have selected a zone and clicked one of the action buttons at the top of the table. Hereafter, the action buttons might not be correctly activated relative to the zone state.
To ensure that the action buttons are properly activated, navigate to another pane and return to the Zones table.
Workaround: None.
If you delete a zone, the row in the Zones table is not immediately deleted. The Zone entry in the Zones table is not removed until Zone deletion has been completed. This process might take a few minutes.
To view an updated Zones table, perform one of the following steps:
Navigate to another pane and return to the Zones table.
Click one of the action buttons at the top of the table.
Workaround: None.
If you attempt to delete a resource pool that contains zones, the resource pool is deleted and the zones are moved to the pool_default resource pool.
Workaround: None.
If the project status is changed, the Status field in the Projects table might not update with the correct status.
To display the correct status, navigate to another pane and return to the Projects table.
Do not use the browser Reload button to refresh the Projects table. This behavior is not supported.
Workaround: None.
A Resource Change Job has been previously scheduled, executed, and completed normally. If you update the Resource Change Job, the job fails.
Workaround: Instead of updating the Resource Change Job, delete it and create a new job.
If an alarm condition exists, the tool tip of the alarm badge might appear instead of the alarm badge itself.
If you navigate to another pane, then return to the alarmed object's icon, the badge on the icon is correctly displayed.
Workaround: None.
If you navigate to the Zone - Properties pane, the Additional Attributes values might not be displayed.
If you encounter this situation, leave this pane and navigate back to the Zone by using the links in the left pane and click the Properties tab.
Workaround: None.
If you attempt to create a resource pool with a name that starts with any character other than an alphabetic, creation fails.
If the host does not appear to be respond after a failure, you might need to restart the host:
Become superuser on the host.
% su -
Stop and restart the host.
# /opt/SUNWsymon/sbin/es-stop -a
# /opt/SUNWsymon/sbin/es-start -a
Valid characters are the following:
Alphabetic (A–Z, a–z)
Numeric (0–9)
Hyphen (-)
Underscore (_)
Period (.)
Workaround: Begin all resource pool names with an alphabetic character.
If the project status is changed, the Status field in the Projects table might not update with the correct status.
To display the correct status, navigate to another pane and return to the Projects table. Using the Reload button is not supported.
Workaround: None.
When you click the browser back button, Container Manager might throw an exception. The browser back button is not supported in this release.
To recover from this exception, go to https://server_name:6789/containers and navigate to the appropriate page.
Workaround: None.
On a Solaris 8 system with Container Manager installed, projects might exist in the /etc/project file with project names that do not conform with the naming convention (user.username). These project are discovered as an application-type container. This assignment can be confirmed in the properties sheet for the container.
If an lnode of username and a project name of user.username exist for this project, the container is identified as a user-based container. The container then functions normally in Container Manager.
If no lnode is associated with this project, you can deactivate the project. The entry in the /etc/project database is deleted. If you subsequently attempt to activate the container, the following error message is displayed:
There was an error while activating the Container. Container activation failed. |
Workaround: To re-create the container definition through Container Manager, perform the following steps:
Delete the container definition.
Re-create the container definition through Container Manager.
Occasionally, utilization graphs are not displayed correctly. All data points relative to the x-axis appear as a single line on the y-axis.
Workaround: To reload the frame in which the utilization graph appears, perform one of the following workarounds:
NetscapeTM 4.79 software:
In the frame in which the utilization graph is displayed, press mouse button 3.
Choose Reload Frame from the pop-up menu.
Netscape 6 software, at minimum:
In the frame in which the utilization graph is displayed, click mouse button 3.
Choose This Frame from the pop-up menu.
Choose Reload Frame from the submenu.
Internet Explorer 5.0 software, at minimum:
In the frame in which the utilization graph is displayed, click mouse button 3.
Choose Refresh from the pop-up menu.
Do not click the Reload or Refresh button in the browser's toolbar. You might return to the Hosts navigation window.
When the Container Manager module is loaded, the module refreshes the container table by reading the /etc/project database. On agent hosts that run the Solaris 8 OS, a memory leak occurs because of a bug in the APIs of the libproject library.
Workaround: To resolve the memory leak problem, perform one of the following workarounds:
For a permanent solution, install the following patches:
108528-29 108987-13 108993-33 109147-28 111023-03 111111-03 111317-05 112396-02 113648-03 115827-01 116602-01 |
To temporarily fix the problem, edit the /etc/nsswitch.conf file and remove nis from the project line. The revised project line should resemble the following example:
project: files |
On the Solaris 9 Operating System, some processes, for example, the JavaTM application run by user nobody, are not moved to a container because of a bug in the newtask command.
Workaround: Start the applications that are run by user nobody in the appropriate container by using the newtask command.
# /usr/bin/newtask -p projectname command |
projectname is the project in which the application is to be started and command is the command to start the application.
For more information, see the newtask(1) man page.
Clicking the Help button displays the Help Table of Contents. Context-sensitive help is not currently available.
Workaround: Use one of the following options to access Help for a specific panel:
Each panel contains inline help immediately below the page title. Click the link following the inline help to open the Help window for this panel.
Use the Table of Contents to access the Help for a specific panel. The naming convention of the links within the Table of Contents is based on the page title. For example, to view the Help window for the Host and Groups panel, click the About Host and Groups link in the Table of Contents.
If you use one of the following match expressions when creating a container definition and subsequently activate it, the agent host hangs:
n
ns
nsc
nscd
Workaround: Start the nscd process in the appropriate project by using the newtask command:
# /usr/bin/newtask -p projectname command |
projectname is the project in which the application is to be started and command is the command to start the application.
For more information, see the newtask(1) man page.
If you plan to reinstall the Solaris Container Manager 1.1 software on the server layer, be sure to save the data when uninstalling. Similarly, if you run setup again on the server layer, be sure to answer No when asked if you want to re-create the database. Otherwise, agent hosts previously discovered might be missing from the database if the data was not saved and you then reinstall or set up again on the server layer.
Workaround: If hosts are missing after you reinstall or rerun setup, restart the Sun Management Center agent on the hosts by doing the following procedure:
Log in to agent system.
Become superuser.
% su -
Stop the Sun Management Center agent by typing:
# /opt/SUNWsymon/sbin/es-stop -a |
Start the Sun Management Center agent by typing:
# /opt/SUNWsymon/sbin/es-start -a |