Cisco Cisco Packet Data Gateway (PDG)

 

▄  VPC-VSM System Administration Guide, StarOS Release 19 

This section provides an overview of how this feature is implemented and the recovery process. 

The Session Recovery feature provides seamless failover and reconstruction of subscriber session information in the 
event of a hardware or software fault within StarOS that prevents a fully connected user session from being 
disconnected. 

Session recovery is performed by mirroring key software processes within StarOS. These mirrored processes remain in 
an idle state (standby-mode) wherein they perform no processing, until they may be needed in the event of a software 
failure (for example, a session manager task aborts).  

StarOS spawns new instances of “standby mode” sessions for each active control processor (CP) being used. 

There are some situations wherein session recovery may not operate properly. These include: 



 

Additional software or hardware failures occur during the session recovery operation 



 

A lack of hardware resources (memory and CPU cores) leaves StarOS unable to support session recovery. 

  After a session recovery operation, some statistics, such as those collected and maintained on a per 

task manager basis are in general not recovered. 

Session Recovery is available for a SecGW (Security Gateway) running as a VPC-VSM instance (StarOS VM) on an 
ASR 9000 VSM CPU. 

  Always refer to the SecGW Administration Guide and StarOS Release Notes for other 

possible session recovery and Interchassis Session Recovery (ICSR) support limitations. 

When session recovery occurs, StarOS reconstructs the following subscriber information: 



 

Data and control state information required to maintain correct call behavior. 



 

A minimal set of subscriber data statistics; required to ensure that accounting information is maintained. 



 

A best-effort attempt to recover various timer values such as call duration, absolute time, and others. 



 

The idle time timer is reset to zero and the re-registration timer is reset to its maximum value for HA sessions to 

provide a more conservative approach to session recovery. 

  Any partially connected calls are not recovered when a failure occurs. 

 

  Failure of critical tasks will result in restarting StarOS. Kernel failures, hypervisor failures or 

hardware failures will result in the VM restarting or going offline. The use of ICSR between two VPC-VSM instances is 
the recommended solution for these types of failure.