The User Plane Node Bring-Down Procedure is a Maintenance Operating Procedure (MoP) that outlines the steps to bring down a specific User Plane (UP) node for maintenance operations. The purpose of this procedure is to disable specific UP node on Control Plane (CP) node while node selection takes place for new incoming sessions.

This feature provides the following functionality:

• Configuration to mark a specific UP unavailable for any new session.

• Option to delete idle subscribers.

Following are the preconditions to bring down the UP node for maintenance operation:

• The UP and CP nodes must be associated with each other and that the calls land on a specific UP.

• You want to disable the specific UP node for maintenance, from being selected for new incoming calls, and clear existing idle users on that UP node.

• To avoid any call loss, there’s a provision of another UP node in the same CP group. If there’s only one UP in the group and we disable it for maintenance, the CP rejects the new incoming sessions with disconnect reason as “user-plane-info-not-available”.

Following are the known limitations of this feature:

• The multi PDN calls for same UE can fall to different UPs.

• The “busy-out” configuration is done on all Active and Standby CPs.

• Currently, the new IP Pool isn’t added for the specified UP or the belonging UP group so that some of the IP chunks doesn’t get assigned to this UP, leading to loss in capacity.

• All CPs must have the same configuration for “busy-out”. Else, UP uses the latest configuration value triggered from any one of the CPs. Similarly, when you do the “no busy-out” on any of the CPs, UP comes out of “busy-out”.

• To block the UP completely, “busy-out” both its IPv4 and IPv6 addresses using two separate CLIs.

• Currently, there’s a spike in CPU usage when idle timeout triggers for a huge number of calls. The CPU usage eventually goes down when all the calls get cleared out.

The MoP is applicable only when CPs and UPs are on the same software version where the support of the following “busy-out” CLI is available.

configure 
   busy-out user-plane { ipv4-address ipv4_address | ipv6-address ipv6_address } [ inactive-timeout inactive_time ]  
   end 

You must enable the configuration on CP to make a UP unavailable for new sessions and clear the idle sessions which exceeds the time configured in “inactive-timeout”. The ipv4_address /ipv6_address is the IPv4 or IPv6 address of the UP. “Inactive-Time” is configured in seconds.

If you don’t configure the inactive timeout, then the behavior for the idle session remains unchanged.

NOTES:

• Existing clear subscribers saegw-only uplane-address ip_address no-select-up CLI command is not extended or reused for “busy-out” logic as it is exec-level CLI which is not fit in the scenario of session-recovery. There is also no other way to roll back the up-selection logic without reassociating the UP using this CLI.

  Note	When the clear subscribers command is executed on UP, CP will not be informed and will consider the sessions as running.

• After “busy-out” CLI command is executed, the UP is removed from UP selection. Existing calls continue to function as usual. No extra operation is performed to the assigned IP pool chunk.

• If you want to bring the same UP after upgrade to handle new calls, you must undo the configuration by executing the no busy-out user-plane { ipv4-address ipv4_address | ipv6-address ipv6_address } CLI command.

• After “inactive-timeout” is configured for UP in “busy-out” CLI, calls get cleared and some of the assigned pool chunks to this UP gets reclaimed if the other UPs sharing the same IP Pool reach threshold of approximately 70% of the chunks allocated.

Following are the CLI commands available in support of this feature.