Symptom

Combination on cbQoS cache and telemetry collection from qos PD component may block LC qos stats processing: #show processes blocked location 0/X/cpu0 Fri Apr 26 12:04:33.347 CEST Jid Pid Tid ProcessName State TimeInState Blocked-on 203 7465 7465 qos_ma_ea Reply 0509:42:03.0269 6502 qos_ma 207 6502 6502 qos_ma Reply 0509:42:03.0275 5813 ifmgr Restarting qos_ma on LC fixes the issue for some time, until next blocking state observed.

Conditions

[1] Observed on eXR/cXR running 6.5.2/6.5.3 [2] cbQoS cache enabled 'snmp-server mibs cbqosmib cache' [3] telemetry requests PD QoS stats = Cisco-IOS-XR-asr9k-qos-oper:platform-qos Suspected conditions to run into the issue: - 10x10 breakout interface with QoS policy applied - high amount (>10) of interface with QoS policy applied

Workaround

Remove QoS policy from 10x10 breakout interface and restarting qos_ma process on relevant LC in some scenarios stop issue from appearing again.

Further Problem Description