Failover in Case of a Routing Engine Failure

Graceful Routing Engine switchover (GRES) is supported in case of a Routing Engine failure. A failure of the Routing Engine stops the PGCP process. When the JUNOS software high-availability framework detects the Routing Engine failure, it switches control of the packet gateway to the PGCP process on the backup Routing Engine. The PGCP process stores completed H.248 states, and these states survive a restart or a switchover. After the restart or switchover is completed, the PGCP process reapplies the states.

Gate Synchronization Procedure

When the IPC connection between the PGCP process and the MultiServices PIC is being reestablished, a synchronization process restores the previous state of gates, so session and call continuity is achieved. That is, existing calls stay active and voice sessions are not disconnected.

The synchronization process can result in the mismatching of gates. Mismatched gates are handled as follows:

• If the PGCP process detects gates that exist in the Routing Engine, but are missing in the PIC, the PGCP process reinstalls the gates on the PIC. Existing calls receive service from the PIC when the synchronization is complete.
• If the PGCP process detects gates that exist on the PIC, but not on the Routing Engine, the PGCP process removes the gates from the PIC. This removal causes the associated sessions to be closed, and the session resources to be released.
• If the PGCP process detects gates that exist on the PIC and on the Routing Engine, but the versions of the gates do not match, the PGCP process forces the Routing Engine version. Doing so maintains call continuity by making sure that the databases are synchronized, while limiting the effect to existing calls and sessions.

You can control the number of mismatches by configuring the synchronization properties.

Configuring Synchronization Properties

You can configure the maximum number of mismatches allowed during the synchronization process between the PIC and the PGCP process.

Step-by-Step Procedure

To configure properties for synchronization between the PIC and the PGCP process:

1. Access the graceful restart configuration.

   [edit services pgcp]

   edit gateway vpg-1 graceful-restart

2. Configure the maximum number of mismatches allowed during the synchronization procedure. If the number of mismatches exceeds this number, the PGCP process clears the state of the PIC and the state of the PGCP process. All calls and sessions are terminated and existing resources are released.

   [edit services pgcp gateway vpg-1 graceful-restart]

   user@host#set maximum-synchronization-mismatches 20

3. Configure the number of milliseconds within which you want the synchronization process to complete. If the process is not complete when this time expires, the PGCP process clears the state of the PIC and the state of the PGCP process.

   [edit services pgcp gateway vpg-1 graceful-restart]

   user@host#set maximum-synchronization-time 300

Displaying the Status of the Routing Engine Synchronization

Purpose

To determine the status of the synchronization between the main Routing Engine and the backup Routing Engine, display the value of the replication socket field.

Action

. . . 
Packet gateway configuration:
  Name                                     :  vpg—1              
  IP address                               :  10.10.10.1           
  Port                                     :  2944              
  Status                                   :  In-Service (Registered) 
  Active gateway controller                :  dt3         
  Cleanup timeout [secs]                   :  3600              
  Gate inactivity delay [secs]             :  240               
  Gate inactivity duration (Q-MI ) [secs]  :  86400  
  Replication socket                       :  Disconnected           

. . .

Meaning

The replication socket can be in one of the following states:

• Connected (Ready)—The replication is ready, and it is safe to perform a switchover.
• Connected (Syncing)—The replication is synchronizing. It is not safe to perform a switchover.
• Connected (Error)—An error occurred in the previous switchover.
• Disconnected—The backup Routing Engine is down. It is not safe to perform a switchover because there is no backup Routing Engine.

Related Topics

• Part 3, Graceful Routing Engine Switchover in JUNOS High Availability Configuration Guide
• Chapter 27, Summary of Packet Gateway Configuration Statements in JUNOS Services Interfaces Configuration Guide
• Chapter 25, Packet Gateway Control Protocol Operational Mode Commands in JUNOS System Basics and Services Command Reference

Copyright © 2008, Juniper Networks, Inc. All rights reserved. Trademark Notice.