Rx_prio0_buf_discard Counter Keeps Increasing Despite Low Traffic

Infrastructure & Networking Software And Drivers Mellanox OFED
1

Hi,

On one of our servers equipped with a Mellanox/NVIDIA NIC, we noticed that the rx_prio0_buf_discard counter keeps increasing steadily over time, even though the actual traffic volume on the system is relatively low.

This seems to suggest that receive buffers are being dropped or discarded, possibly due to buffer starvation or unprocessed packets, but we are unsure why this would happen under such light traffic conditions.

We’ve verified that:

  • The system is not under memory pressure
  • There is no high CPU load or IRQ bottleneck
  • ethtool -S shows increasing discards only for prio 0

This behavior appears as if buffers are not being consumed or released properly in the RX data path.

Could this indicate a driver/firmware issue, QoS misconfiguration, or an issue in the RX path handling for prio 0 traffic?

We would appreciate any insights on what could be causing these rx_prio0_buf_discard increments and how to further debug or mitigate it.

My Enviroment

Ethernet controller: Mellanox Technologies MT2910 Family [ConnectX-7]

Tool Information
----------------
Firmware Version                   : 28.41.1000
amBER Version                      : 2.22
MFT Version                        : mft 4.26.1-3

rx_discards_phy: 0
rx_prio0_discards: 0
rx_prio0_buf_discard: 45892