Stopping application server takes too much time

Issue

• Shutting down the application server waits indefinitely so we have to use kill command.

Environment

• Liferay DXP 7.0+

Resolution

• It can happen that a latch countdown to stop an OSGi service takes a lot of time, in any circumstance. Thus, the stuck OSGi service prevents the application server from stopping.
• However, Felix Declarative Services implementation (the OSGi framework used by DXP) has, via configuration, a timeout procedure to solve services gain and/or stop blocks during framework shutdown.
• The timeout configuration is accomplished by the following framework properties:
  

  org.apache.felix.scr.impl.config.ScrConfiguration.DEFAULT_LOCK_TIMEOUT_MILLISECONDS = 5000;
  
  org.apache.felix.scr.impl.config.ScrConfiguration.DEFAULT_STOP_TIMEOUT_MILLISECONDS = 60000;

• Nevertheless, these properties are redefined in DXP in portal.properties file with the following default values:
  

  #
  # Pass arbitrary extra properties to the framework using the prefix:
  # "module.framework.properties." to avoid avoid colliding with other
  # frameworks that may be running within the same JVM.
  #
  ...    
      module.framework.properties.ds.lock.timeout.milliseconds=1800000
      module.framework.properties.ds.stop.timeout.milliseconds=1800000
  ...    

• So, the application server shutdown process would be affected by the OSGi framework's wait before the latter considers a timeout has happened. 1800000 milliseconds = 30 minutes.
• As a suggested solution, defining both properties in portal-ext.propeties file with suitable values for your system would avoid unnecessary waits.

Additional Information

• Both properties were included in LPS-69090.