Cisco Cisco Packet Data Gateway (PDG)
Interchassis Session Recovery
▀ ICSR Operation
▄ VPC-VSM System Administration Guide, StarOS Release 19
254
ICSR Initialization
When StarOS is simultaneously initialized on each VPC-VSM instance, the instances send Hello messages to their
configured peer. The peer sends a response, establishes communication between the instances, and messages are sent
that contain configuration information.
configured peer. The peer sends a response, establishes communication between the instances, and messages are sent
that contain configuration information.
During initialization, if both instances are misconfigured in the same mode - both active (primary) or both standby
(backup), the instance with the highest priority (lowest number set with the ICSR priority command) becomes active
and the other instance becomes the standby.
(backup), the instance with the highest priority (lowest number set with the ICSR priority command) becomes active
and the other instance becomes the standby.
If the instance priorities are the same, StarOS compares the two MAC addresses and the instance with the higher MAC
address becomes active. For example, if the instances have MAC addresses of 00-02-43-03-1C-2B and 00-02-43-03-01-
3B, the last three sets of octets (the first three sets are the vendor code) are compared. In this example, the 03-1C-2B and
03-01-3B are compared from left to right. The first pair of octets in both MAC addresses are the same, so the next pairs
are compared. Since the 01 is lower than the 1C, the instance with the MAC address of 00-02-43-03-1C-2B becomes
active and the other instance the standby.
address becomes active. For example, if the instances have MAC addresses of 00-02-43-03-1C-2B and 00-02-43-03-01-
3B, the last three sets of octets (the first three sets are the vendor code) are compared. In this example, the 03-1C-2B and
03-01-3B are compared from left to right. The first pair of octets in both MAC addresses are the same, so the next pairs
are compared. Since the 01 is lower than the 1C, the instance with the MAC address of 00-02-43-03-1C-2B becomes
active and the other instance the standby.
Operation
This section describes how the VPC-VSM instances communicate, maintain subscriber sessions, and perform a
switchover.
switchover.
Communication
If one VPC-VSM instance is in the active state and one in the standby state, they both send Hello messages at each hello
interval. Subscriber sessions that exceed the checkpoint session duration are included in checkpoint messages that are
sent to the standby instance. The checkpoint message contains subscriber session information so if the active chassis
goes out of service, the backup instance becomes active and is able to continue processing the subscriber sessions.
Additional checkpoint messages occur at various intervals whenever subscriber session information is updated on the
standby instance.
interval. Subscriber sessions that exceed the checkpoint session duration are included in checkpoint messages that are
sent to the standby instance. The checkpoint message contains subscriber session information so if the active chassis
goes out of service, the backup instance becomes active and is able to continue processing the subscriber sessions.
Additional checkpoint messages occur at various intervals whenever subscriber session information is updated on the
standby instance.
Switchover
If the active VPC-VSM instance goes out of service, the standby instance continues to send Hello messages. If the
standby instance does not receive a response to the Hello messages within the dead interval, the standby instance
initiates a switchover. During the switchover, the standby instance begins advertising its srp-activated loopback and
pool routes into the routing domain. Once the instance becomes active, it continues to process existing AAA services
and subscriber sessions that had checkpoint information, and is also able to establish new subscriber sessions.
standby instance does not receive a response to the Hello messages within the dead interval, the standby instance
initiates a switchover. During the switchover, the standby instance begins advertising its srp-activated loopback and
pool routes into the routing domain. Once the instance becomes active, it continues to process existing AAA services
and subscriber sessions that had checkpoint information, and is also able to establish new subscriber sessions.
When the primary instance is back in service, it sends Hello messages to the configured peer. The peer sends a response,
establishes communication between the instances, and sends Hello messages that contain configuration information. The
primary instance receives an Hello message that shows the backup instance state as active and then transitions to
standby. The Hello messages continue to be sent to each peer, and checkpoint information is now sent from the active
instance to the standby instance at regular intervals.
establishes communication between the instances, and sends Hello messages that contain configuration information. The
primary instance receives an Hello message that shows the backup instance state as active and then transitions to
standby. The Hello messages continue to be sent to each peer, and checkpoint information is now sent from the active
instance to the standby instance at regular intervals.
When switchover occurs, the session timers are recovered. The access gateway session recovery is recreated with the
full lifetime to avoid potential loss of the session and the possibility that a renewal update was lost in the transitional
checkpoint update process.
full lifetime to avoid potential loss of the session and the possibility that a renewal update was lost in the transitional
checkpoint update process.