STK Scalability 1.2

  • New Features
    • New options to control the lifetime of host processes have been added. Hosts are recyclable after any one of a variety of time, memory, or job-based conditions have been satisfied. The Agent Tray Application has a new Host Recycling panel to set values for these options. The values set in the Agent Tray Application UI will be overridden if a submitted job's TaskEnvironment's HostRecycleSettings have been set programmatically.
  • Improvements
    • Reduced the amount of logging the external capability checkers generated in the agent log.
    • Improved capture of task standard output.
  • Bug Fixes
    • Fixed a bg where running parallel coverage with tirem could cause a host to hang in rare conditions.

STK Scalability 1.2

  • The product formerly known as STK Server has been renamed to STK Scalability Extension.
  • New Features
  • Improvements
    • Added a new throttling capability to the coordinator to better manage its memory usage when running jobs with high memory requirements.
    • Improved memory management when executing tasks with large payloads/results.
    • The detection of unresponsive elements in the system and their associated heartbeat messages used to rely on the local machine time and could therefore be wrongfully triggered if the local machine time was modified (e.g. manually or due to daylight savings). This logic has been reworked to no longer depend on the local machine time and is thus now immune to local machine time modifications.
    • Any non-default configuration value is now logged.
    • Added caching for the byte arrays used during message serialization to improve performance.
  • Bug Fixes
    • Fixed a race condition that could result in the agent service crashing if a host repeatedly failed after finishing a task.
    • Fixed the documentation on the expected behavior when Job.RaiseEventsUsingSynchronizationContext is true (.NET API only).
    • In some cases, IJobSchedulerContext.reserveResource(java.lang.String, long) would return an exception without waiting for the resouces to be available. This has been fixed and IJobSchedulerContext.reserveResource(java.lang.String, long) will now block until the requested resources become available or the specified timeout expires.
    • Fixed spurious debug messages in the agent log relating to invalid format strings.
    • Fixed a bug where the agent service might not have retried a task when the host failed to start due to low memory conditions. In these cases the task will now be retried similarly to the other failure cases.

STK Server 1.0.2

The first release of STK Scalability API for Java!