KBI 310554 Issue Addressed: Argent AT Causes Cancelled Tasks Due To Deadlock Thresholds
Version
Argent AT — all versions
Date
Tue, 25 Jun 2013
Summary
Argent AT has a built-in deadlock detecting mechanism using a registry-controlled deadlock threshold.
The default threshold is 60 seconds.
This value may be too small for large installations (> 300 nodes)
This causes unnecessary service recycling, which results in cancelled tasks.
Technical Background
Argent AT detects a potential deadlock if the engine attempts to acquire a shared resource, and cannot do so within the threshold controlled by the registry ‘DEADLOCK_THRESHOLD_IN_SECOND’.
When this occurs, the service is recycled, and all running or pending tasks are to cancelled.
As a result, customers may see a large number of tasks in the ‘Scheduled Tasks’ screen labeled as ‘Cancelled’.
For large installations, Argent AT does require more than 60 seconds to acquire a shared resource, such as a loaded database connection.
Resolution
For 3.1A-1304-L and earlier, edit the registry DEADLOCK_THRESHOLD_IN_SECOND to 300
(5 minutes seem to be a reasonable number for typical large installations)
Note: NO service restarts are required
Alternatively, customers can upgrade to Argent AT 3.1A-1307 or later, where the default value is now 300