Symptom

SW: IOS XR Software, Version 7.6.1 HW: NCS-6008 RP/B0/CB0/CPU1:NCS-6008(config)#interface Bundle-Ether110.1035 RP/B0/CB0/CPU1:NCS-6008(config-subif)#service-policy output OUT:GWR:fiber:50G RP/B0/CB0/CPU1:NCS-6008(config-subif)#commit % Failed to commit one or more configuration items during a pseudo-atomic operation. All changes made have been reverted. Please issue 'show configuration failed [inheritance]' from this session to view the errors RP/B0/CB0/CPU1:NCS-6008(config-subif)#show configuration failed !! SEMANTIC ERRORS: This configuration was rejected by !! the system due to semantic errors. The individual !! errors with each failed configuration command can be !! found below. interface Bundle-Ether110.1035 service-policy output OUT:GWR:fiber:50G !!% 'SSE_QOSEA' detected the 'warning' condition 'Bandwidth over-subscription not supported.' When bundle members change state (up/down) for various reasons or the line card with bundle members reloads then the bandwidth of the Bundle-Ethernet interface is changed. There is a check that keeps track of the priority traffic reservations for all sub-interfaces of a parent bundle interface (sse_qosea_par_child_info_subif_bw_used_check). When a bundle member flaps that has both ingress and egress policies, the running total of the bandwidth used by all the sub-interfaces on the bundle is not updated correctly due to CSCwh13086. Issue is observed only when both the ingress and egress policies applied simultaneously on sub-interface. A flag is not set properly for first sub-interface of bundle main interface while updating cur_sum bandwidth. This is leading to incorrect bandwidth updates. The sse_qosea_par_child_info_subif_bw_used_check checks to make sure that the sum of all the eqress QOS bandwidths on sub-interfaces of a parent Bundle-Ether interface does not exceed the interface bandwidth on the line card. This sse_qosea_par_child_info_subif_bw_used_check keeps a running total of the cur_sum on the BE interface each time a new egress QOS policy is added or removed to a sub-interface under that BE. When the egress QOS policy commit attempts to add policy the amount of bandwidth needed by the policy (new_propagated_bw) is calculated and the existing cur_sum is read from a register. The new_propagated_bw is the amount of priority class bandwidth since only that traffic is guaranteed. If the sum of the cur_sum + new_propagated_bw is greater than the interface bandwidth on the LC then the check fails and the commit fails. The ?show qos ea trace errors loc all? command reports the error that is causing the check to fail.

Conditions

If a bundle member flaps that has both ingress and egress policies applied to the bundle or a sub-interface of the bundle, the running total of the bandwidth used by all the sub-interfaces on the bundle is not updated correctly The issue can be triggered by either individual bundle member interface state changes or during line card reloads. There are no CLI commands or traces available to display the ?cur_sum? of any Bundle-ethernet interface when the ?cur_sum? is accurate and the sse_qosea_par_child_info_subif_bw_used_check succeeds. A debug SMU (for CSCwh13086) was built that would include QOS traces and syslog to track calculations made for new_propagated_bw and the updates made to ?cur_sum? when QOS policies are added or removed from bundle sub-interfaces or other event that impact egress QOS policies take place including: bundle member flaps, LC reloads, BE interface state changes. This issue was reproduced in both Cisco and customer labs on both Single Chassis and Multi-Chassis systems.

Workaround

No workaround possible.

Further Problem Description