Symptom

The FM of Nexus may stop forwarding a specific flows after upgrading or enabling strict priority queue for user configurable Qos class. In detail, under stateless reboot (upgrade) or Qos configuration, the strict priority queue configuration may not be programmed properly for the FM's internal port, which may end up a DWRR queue with %0 bandwidth, hence no packet can be drained. issue "show platform internal hal qos uc_mc_queue_occupancy_counter" from the FM would tell if there is any ASIC had packet queued up.

Conditions

Strict priority is enabled global Qos, Spines was stateless rebooted because of the upgrade. Or String priority is enabled for a fabric which does not contain the fix.

Workaround

1- Identify the FM which are impacted, then reload the FM during a change window. or 2- Disable and Strict priority for the user class, then re-enabled it in a change window.

Further Problem Description