Task Server

Resolved Issue

During a migration from Task Server 6.0 to Task Server 7.0 SP2, remote task servers were being uninstalled and left in that state until user-intervention. To resolve this specific issue, the scope of the filter used for Task Server uninstall has been narrowed so that the correct computers are targeted.

Web Controls

Resolved Issue

Defect ID

GridControl had a memory leak issue only when UseInternalGridCallback property was set to false. This is because in this scenario, the GridControl uses a callback to bring down an entirely new CA grid on each refresh, without telling CA to dispose of its resources before triggering the refresh callback. The change was a 1 line change to tell the ca grid to dispose its resources before the callback takes place. Memory usage when UseInteralGridCallback == false is now more of a flat line rather than a constantly increasing amount.

Symantec Installation Manager

Resolved Issue

SIM was originally coded to require dependent products to explicitly call out the fact they are compatible with a newer version of a product by adding a dependency (with the same group ID) to the new product. We have changed the design now to assume that superseding packages are always backwardly compatible and satisfy the requirements met by the old dependency.

Cancel button on uninstall prompt for dependent items does NOT cancel the uninstallation

When the user chooses to uninstall s/he may be presented with a dialog explaining that with that particular selection additional components will also be uninstalled. However, if the user chooses the Cancel button on this dialog, the uninstall proceeds anyway.

Fixed problems in SIM code with uninstalling the Platform itself when the HF1 was applied. Services were still running in some cases when they shouldn't have been which was the root cause of the service shutdown and reboot notifications reported in this bug.

The repair of the platform is now disabled when updates to it are installed.

Suppressed reboots during repair. Changed repair initiated from the UI to only repair missing or older files. Repairs due uninstall of updates still perform a full repair that replaces the files from the previous revision level.

Occurs infrequently, but crashes install when it does. Removed conn.Close(); statements inside of a Using block inside SQLAuth.

PPA/PAL

Resolved Issue

Defect ID

Caching with a timeout was added to the connection profile manager. This means that when a connection profile is retrieved with its associated credentials, it is cached in the profile manager for a specific amount of time (defaulted to 300 seconds). When the connection profile is accessed, the timestamp of the retrieval of the CP (connection profile) is checked and, if it has expired, the CP is retrieved again from the database. If not, the cached CP is retrieved from memory.

From Symbol SDK documentation - In order to properly manage its network resources, the controller firmware will automatically terminate client connections that have been idle for some predetermined amount of time. This cleansing mechanism provides key resource preservation capabilities in the TCP environment. Without it, connections associated with clients that were abruptly terminated would simply accumulate undetected and use up precious controller network and memory resources.  PPA has been modified to avoid this situation and keep the connection open.

A WSMAN session can “hang” if targeted device does not respond to a TCP or HTTP request. In particular this can happen with the Network Discovery process if one or more devices are not responding.

No server response to requests due to environmental and system issues such as a slow network or incorrect connection parameters. Adjustments to how PPA handles the connection resolve this issue.

Event Console

Resolved Issue

Defect ID

The current purging architecture allows unresolved alerts to be purged. In the case of Monitor Solution, if an alert that is unresolved is purged, it leaves the agent in a triggered state because Monitor Solution relies on a Notification Server Message indicating a resolution so agents can be reset. When an unresolved alert is purged(deleted), Monitor Solution does not get the message so the agent remains triggered and will not send another alert. On the server, Event Console displays the health as normal because the alert was deleted. The fix adds restraints to the Purge logic to prevent the purging of any unresolved Monitor Solution alerts. This does not apply to SNMP traps. The caveat is that the user needs to respond to every alert, manually or automatically, before the Monitor Solution alert can be purged.

High volume alert flow was causing memory usage to rise and stay up until the alert volume droped (allowing garbage collection to catch up). We added an interface to the alert object to allow it to be disposed more efficiently. We also added a call to the garbage collector on a timer (when alerts are being received).

Alert Forward Rules and Alert Task Rules with Time of Day conditions were evaluating incorrectly due to a UTC timestamp comparison against local server time. The alert is now converted to local server time prior to Forward/Task rule evaluation.

On Notification Servers with several thousand or more managed resources, the ResourceKey table grows very large. An Event Console query was not constrained to reduce the result set to a manageable size. The stored procedure was altered to include a where clause limiting the result set to needed rows. Impacted Regions: The EventEngine.exe and EventReceiver.exe Event Console services both use this query to match alerts to resources as well as evaluate alert rules.

Network Discovery

Resolved Issue

Defect ID

When we added a new table field in HF1, we missed modifying one of the existing stored proceedures that references this table. This caused any attemped modification on our Device Classification page to fail.

Correlation fails when servers with an existing Altiris Agent is discovered

The correlation logic was changed so existing devices can be correlated.

Monitor Core

Resolved Issue

Defect ID

25% of the strings were not translated due to a string submission issue. This has been corrected.

Monitor alerts are individually sent to the Notification Server using NSE. This proved to be inefficient during scalability testing. Now the alerts are sent to the server in Bulk to improve performance.

There was a synchronization bug where OnDemand requests would not properly prioritize metric polls, causing metrics to not get polled and evaluated when expected. This was exposed during OnDemand scalability testing.

During migration from 6x to 7.0, the NT Event metric would be set with the wrong data type. This would cause any NT Event to be processed and cause a rule to trigger during rule evaluation, rather than only NT Events that satisfied the rule criteria. The correct data type is now being set.