Defect ID

Resolved Issue

Existing installs of 32-bit Agent Manager instances on 64-bit operating systems are improperly updated to 64-bit installs, causing them to fail when they are restarted. New 32-bit installations of FglAM 5.8.1 or later are not affected.
Workaround: To upgrade existing and 32-bit FglAM installation on an 64-bit operating system:
1. Shut down the existing Agent Manager instance.
2. Install a new copy of Agent Manager, version 5.8.1 or later, on the same system.
3. In the new install, open the state/default/config/fglam.config.xml file for editing, and replace the <config:id> ... </config:id> element with the one obtained from the existing Agent Manager installation.
4. In the same file, replace the <config:host-display-name> ... </config:host-display-name> element with the one obtained from the existing Agent Manager installation.
5. Start the newly installed Agent Manager and deploy all agents that were previously deployed to the existing Agent Manager installation.

Newly installed Agent Manager on Windows 2008 R2 would not start. WindowsTopologyCollector failed to detect the Manufacturer name during a wmic call.

Support is added for cron- and non-cron-aware data collectors.

Warning messages now appear to inform end users that agents whose package names start with "com.quest.glue" are not deployed.

WMI (Native or Java) method executions that use object identity values containing special characters caused the internal query to fail because the special characters were not escaped.

Agent Manager upstream connections were required to use SSL connections for reverse polling or upstream connections.

Fix: A new property, require.secure.server.connection, is introduced to the state/default/config/client.config file. When set to true (by default), the Agent Manager uses an HTTPS connection. Setting this property to false allows the Agent Manager to use HTTP when connecting to the Management Server or an Agent Manager concentrator.

When undeploying a GAR file from the Adapter's MBean that resulted in a material change in negotiated resources, the Agent Manager client did not restart automatically. To reclaim these resources, a manual restart was required.

If the Agent Manager is stopped before completing its startup routine, a NullPointerException was thrown.

The init.d script now includes the status option that reports on the Agent Manager daemon status, such as running, dead, not running, unknown.

Threads were blocked when a large amount of common-layer messages in which the ID was being set, were processed.

The upstream disk cache size setting did not accept -1.

The support-bundle script could not generate a support bundle from a Linux platform when the Agent Manager stopped running.

Agent Manager self-monitoring metrics now use the common topology classes whenever possible. This ensures that agents of different types submit the same values for common objects, such as host names, which may be vulnerable to topology churn.

During an agent upgrade, a property change timing failure may result in a benign ERROR log message:

IllegalArgumentException: An agent with the ID <guid> cannot be found

VM properties set on the command line on startup were not passed to out-of-process instances.

Importing certificates from the command line resolved relative paths to the state directory when launched, instead of the current working directory.

The udp2icmp helper application now generates its own log file. This allows end-users to diagnose udp2icmp-related problems that do not make it into the main Agent Manager log file.

The following message was logged too often:

A domain name has not been provided, assuming <domain> will be acceptable based on the provided hostname <host>

The external Agent Manager process runner can be configured to write its own log file. This is now enabled by default.

An InternalError during upstream connection attempt prevented the Foglight Agent Manager from initiating polling connections.

Fast mode collectors defined with a CRON schedule never terminated.

Attempts to retrieve a support bundle using the browser interface or the command line on a Solaris 11 SPARC system failed.

The Agent Manager uncaught exception handler logged ThreadDeath exceptions.

HPUX: The system info file in the support bundle was empty.

During an Agent Manager upgrade, the config/vm.config file did not migrate to the latest version.

Locked agents could cause the support bundle to fail.

Batch sizes were not based on a weighted moving average of the connection time, instead on an instantaneous measure.

A failure was reported after the successful deployment of an agent package.

The Agent Manager JVM is updated to JRE 7.

AIX platforms are now supported only on 64-bit hardware.

Solaris: When monitoring the local system, the Agent Manager required extra swap space to be made available so that it could execute commands.

The FglAM Performance Overview dashboard did not display the Servlet write data.

udp2icmp error messages were simple numeric values and did not always indicate the real nature of the problem.

When agent deployment failed due to an agent type that was defined in an existing package and a new package, a NullPointerException occurred.

Incorrectly configured queues and agents that overflow data queues were not properly logged.

Excessive logging of credential management is reduced in this release.

On Windows hosts, the Agent Manager crashed when a memory leak occurred in the native Hyper-V host detection code.

When scheduling collections using a CRON expression, the system attempted to derive a collection interval when submitting topology data.

Logging of the encryption algorithm when connecting to an HTTPS upstream is now implemented.

Hosting a high number of agents on a single Agent Manager instance results high memory use.

Fix: A process execution service is provided in this release to prevent direct process execution.

Agent Manager installations on Windows used the old service name, FoglightAgentManager, instead of using the new name, AgentManager.

Defect ID

Resolved Issue

Transferring an agent instance that previously existed on the target Agent Manager resulted in a failure.

Using the Management Server JMX console to change the global log level of a remote Agent Manager added an invalid message to the queue.

Unhelpful error messages were logged when the Foglight license is not installed.

FTR Agent data was discarded by a Canonical Data Transform failure, and no data was displayed in the FTR Agent interface.

Agents running on an Agent Manager host can now be manually migrated to another host.

Timestamp topology properties were not handled correctly by the CDT.

A race condition occurring when the Management Server restarts could cause an Allocate exception for the FglAMServlet if the Agent Manager HTTP request arrived before the Management Server adapter was fully activated.

The adapter was unable to create agents after un-deploying and re-deploying the agent package.

Agents were set to NOT_COLLECTING_DATA when the Agent Manager restarted.

If an Agent Manager was stopped and restarted several times in quick succession, some agents could end up in a wrong state.

An agent that was deleted from the Management Server while the hosting Agent Manager was not connected might be restored when the Agent Manager reconnected.

The FMS Adapter's ThreadPool now allows its configuration values to be persisted into its configuration service.

Defect ID

Resolved Issue

Queue utilization metrics were collected but did not appear on the FglAM Performance Overview dashboard.

Defect ID

Resolved Issue

The JRE is updated to version 7 on all supported platforms.

The installer did not clearly report if it could not configure a Foglight Agent Manager instance.

If the fglam-init-script-installer.sh script was generated with a version of the Agent Manager earlier than 5.6.2.4, it did not work correctly on Linux systems that use upstart to manage init scripts.

The installer performed an extensive cleanup, even in the event of failure, unless an environment variable (QUEST_DEBUG_SKIPCLEANUP) was set. That could make it difficult to determine the root cause of a failure.

Fix: Log files and any contents of the tmp directory are now retained after a failure. A message is generated, pointing the user at the logs in the tmp directory.

The --help documentation for the installer did not correctly indicate that TMPDIR could be used to specify the directory the installer unpacks in and runs from.

The error message that occurred when the silent installer could not create a required directory was unclear.

The JREs shipped with the Agent Manager are updated as a result of the Oracle security patch release.

The Agent Manger installation parameter --noservice was ignored during a silent installation when an Agent Manager service was already installed on the host.

The unbranded Agent Manager Windows service name was too generic.

Defect ID

Resolved Issue

Windows local string command execution failed due to quoting of arguments passed to the ProcessRunner.

WinRM connections failed in Foglight for Exchange.

ExecutionTaskThread handled interrupts incorrectly for PoolableWMINativeConnectionImpl causing task threads to be prematurely shut down and to close connections.

File transfer failed during a base64Conversion operation (on a WinRMCommandShell connection) due to a large file.

Running commands with "{}" in them when using LocalWindowsComamndShell could cause unknown execution exceptions.

Executing repeatedly a WMI method that does not take any parameters using jinterop throws a JIException with the error code 0x800401FB.

WindowsCommandShellFactory threw an incorrect UnsupportedCredentialTypeException when a LocalUserCredential was provided for a remote connection.

A DCOMWindowsCommandShell command execution could hang and cause timeouts for PowerShell commands.

A WinRMCommandShell command execution could hang and cause timeouts for PowerShell commands.

A DCOMWindowsCommandShell command execution threw an IllegalArgumentException when the specified command contained the greater than '>' character.

ConnectionPool did not log a message after failing to establish a connection to a host because of an agent pending state change.

Logging of established connections is improved in this release so that connections to the same host but with different properties can be differentiated.

A transient network failure during remote data collection could cause transient NullPointerExeceptions for some common topology object instances.

A WindowsShell DCOM implementation failed with a shell-interpreted if/else statement.

A connection service using an unsynchonized map that is servicing many threads concurrently could throw IllegalStateExceptions, or cause connections to not be released, or possibly in rare cases could trigger an infinite loop in the map lookup.

Due to a race condition, SSH connection session count tracking could result in lowering the MaxSessions cap for a server below its configured value.

A Sun_SSH 1.0.1 bug on other (non-Sun) sshd implementations was reported.

An empty response code from a WinRM server could result in an extraneous log message, before failing over to WMI data collection:

[STDERR] [Fatal Error] :-1:-1: Premature end of file" before failing over to WMI data collection

Cartridge-defined credential types can now be queried.

WinRM data collection for a Vista 64-bit system prior to the Service Pack 1 upgrade failed with a NullPointerException when attempting to retrieve the host name.

An Agent Manager could experience intermittent failures with an EXCEPTION_ACCESS_VIOLATION error in the Jacob native COM implementation.

An Agent Manager installed on a Linux platform could not establish a WMI connection to Windows Server 2008, Windows 7, or Windows Vista with a local user of the Administrator group.

An Agent Manager installed on a Linux platform could not establish a WMI connection to a monitored Windows host when the Windows Firewall is enabled.

Defect ID

Resolved Issue

The support bundle file provider did not check for warnings about files being too large.

Possible access violation in a native ProcessRunner messaging code on Windows Server 2008 was observed.

Defect ID

Known Issue

When a large legacy Foglight 4-based agent package is used for the first time in an Agent Manager installation, if you create multiple agent instances simultaneously, an error occurs: UndeclaredThrowableException "Failed to acquire the agent lock".

Workaround: The first time you create an agent instance after installing the package, create only a single instance. After that instance is complete, you can create multiple instances simultaneously.

When sudo is not configured to allow root launch without a password for an agent that requests it, the sudo process (labeled fog4_launcher) persists for five minutes until it times out.

When a DB2 agent is installed in two different Agent Manager instances on the same host, auto-discovery logs the message:
WARN An agent named DB2_DB2 could not be found.

The system is unable to start data collection for a JFogbank agent with a space in the instance name.

Defect ID

Known Issue

Script agents that call Windows PowerShell scripts can hang if they are not provided with stdin on the command-line.

Workaround: Provide the script file as stdin when using a batch script agent to invoke a PowerShell script. For example:

@echo off
type yourpowershellscript.ps1 | powershell -ExecutionPolicy RemoteSigned -NoProfile -NonInteractive -command -

A NullPointerException occurred when you start or stop data collection for the script agent.

Script agents did not run properly if they were deployed with incorrect line endings for the target host.

Defect ID

Known Issue

The command line option to common launcher should be expanded to generate heap histograms as well as heap dumps on Sun VMs.

Changing the default temp directory on Windows 7 results in the following error:

QcnUtil\src\qcn_sysinfo - Error output from external program: D:\temp\quest-wmi-query-HKbaaa.vbs

Using multiple --add-debug-level options lead to a FATAL log message, even though the help text says they are allowed.

A script used to pull selected entries from the Windows event log for the Agent Manager support bundle did not exist on Windows 7 or Windows Server 2008. 

The temp directory detection code should determine if a directory exists before attempting to create a temporary file in it.

The command-line option -r for the Agent Harness was stripped out on Windows.

The command-line thread dump failed on Windows Server 2008 R2 with the message: Could not create outgoing pipe ... Access is denied.

Defect ID

Known Issue

Symantec Endpoint Protection 12 reports false positive virus warnings for Agent Manager installers.

When downloading a Windows Agent Manager installer (32- or 64-bit) from the Management Server, Symantec Endpoint Protection version 12 may report a Download Insight failure with the signature WS.Reputation.1. This is not an actual virus signature; rather, it is simply a reputation system which rejects downloads of new executables that have not been seen by many Symantec Endpoint Protection users. This warning may be safely ignored.
Workaround: To proceed with the download, click More under the Activity box, and then click Allow This File.

Running the 32-bit Windows installer may cause Symantec Endpoint Protection to also report a false positive detection with the signature Bloodhound.Sonar.9. Again, this is not an actual virus signature, but rather a heuristic which is triggering incorrectly. With default settings, Symantec Endpoint Protection will not allow the Agent Manager installation to proceed; the installer will abort with the message Configuration Failed.
Workaround: Open Symantec Endpoint Protection, and click Change Settings on the left. Under Proactive Threat Protection, click Configure Settings, and under Sonar, enable the checkbox Prompt Before Terminating a Process.