GTP-C path failure detection should be enabled for enabling this feature.

If a path failure is detected at the Demux, then the path failure notification is sent to all session managers at the P-GW. Next, the session manager cleans up the ongoing transactions. Once all the transactions are deleted, the sessmgr-egtpc deletes the tunnels by adding them into the pacing queue.

The P-GW will not delete the session immediately after detecting path failure with the S-GW restoration in place. The P-GW discards downlink packets received for a maintained PDN connection and stop charging for maintained PDN connections after an S-GW failure that has not been restored.

The MME/S4-SGSN controls the pace of the S-GW relocations to avoid core network node overload. The MME/S4-SGSN prioritizes the S-GW relocation for UEs engaged in a Service Request for RAU/TAU procedures over UEs which are not engaged in any mobility product procedure and do not have a signaling connection to MME/S4-SGSN.

If a session is marked for S-GW restoration and if a new request results in context replacement, then the existing session is aborted followed by a new request indication event.

When the Demux informs the eGTP-C about the path failure, abort all active procedures. The abort procedure indicates to the P-GW if S-GW restoration is enabled to for the session. Add all the session in the queue and start the session hold timer for S-GW. The MME restores these sessions by relocating sessions to the new or the same S-GW. When the MBReq is received at the P-GW and if the session is marked with S-GW restoration, then the S-GW flag is reset. At this point, the session hold timer expires and S-GW restoration is removed.

When the MME and SGSN detects path failure towards the S-GW and the MME supports S-GW restoration, then MME relocates sessions from the failed S-GW to another S-GW.

The EGTPIN manager detects a path failure and informs all session managers about the failure. Then, the session manager gets the path failure notification and S-GW restoration is enabled so it will stop all ongoing transactions. The Sessmgr-egtpc informs the P-GW-drv about the path failure. The P-GW deletes the session.

Next, the eGTP-C starts the session hold timer. If the MME triggers the S-GW service restoration before the session hold up timer, then sessions from the old S-GW will move to the new S-GW. This ensures that no sessions are handed over and none are deleted.

Once the session hold timer expires, all the sessions with that peer are cleaned up. At this point, new sessions are not moved to the new S-GW.

Important

If the S-GW goes down and comes back up in a very short interval, then the P-GW will not detect the path failure. In this case, S-GW restoration does not occur. If the old S-GW comes up again before the hold timer expires, then there is a chance that only some partial sessions move to the new S-GW. In this case, the eGTP-C does not need to delete the remaining sessions with that peer. The eGTP-C will stop the timer.

If the session manager detects the path failure, then it informs the Demux manager. The above, scenario still occurs in this case.

After detecting the path failure, the Demux will not send ECHO messages towards the S-GW. If the new session addition occurs or a new session is restored from the node, then the Demux sends the ECHO messages towards the peer.

The S-GW Restoration functionality complies with the following standards:

Use the following example to enable S-GW Restoration Support.

configure 
   context context_name 
      pgw-service service_name 
         egtpc sgw-restoration session-hold timeout seconds 
         { default | no } egtpc sgw-restoration session-hold 
         end 

Notes:

show pgw-service all

To verify the S-GW Restoration configuration, use the following command:

The following fields have been added to display configuration information for S-GW restoration.

This section provides information regarding show commands and/or their outputs in support of the S-GW Restoration.