A large number of "Receive packets dropped" are occurring on an ESXi host. The issue is environment-isolated to a single host and causes an active alert within VMware Aria Operations for Networks indicating critical packet loss on the physical boundary.
Analysis of the host network statistics confirms that the dropped packets are strictly incrementing on a specific host. The esxcli network nic stats get command shows high drop counts:
[root@host:~] esxcli network nic stats get -n vmnic2NIC statistics for vmnic2 Packets received: 88924040750 Packets sent: 69870283494 Bytes received: 107018966907299 Bytes sent: 74345629163764 Receive packets dropped: 681455735 <<<<<<------------ Transmit packets dropped: 0 Multicast packets received: 654484574 Broadcast packets received: 494334834 Multicast packets sent: 62009 Broadcast packets sent: 32014
Reviewing deeper driver statistics highlights a non-zero value for the threshold counter
vsish -e get /net/portsets/ENK10_CLIENT_VDS01/ports/67108897/vmxnet3/rxSummary
::rx_bits: 204035916044800rx_buffer_passed_threshold: 417 <<<<<<-------------rx_pcs_symbol_err: 0::
VMware NSX-T Data Center
VMware NSX
The physical network interface card (NIC) is receiving traffic faster than the ESXi kernel driver can clear the rx descriptors, leading to hardware FIFO buffer exhaustion indicated by an incrementing rx_buffer_passed_threshold.
Identify the virtual machines residing on the impacted ESXi host generating the highest packets-per-second (PPS) or network throughput.
Execute a live vMotion migration to move lower-priority virtual machines off the affected ESXi host to alternate nodes within the cluster. This decreases the aggregate ingress packet arrival rate on the host.
Monitor the network interface statistics to confirm if the drop rate stabilizes or stops entirely:
esxcli network nic stats get -n <vmnic_name>
Also, check the Hardware Compatibility List (HCL) to ensure the network card's driver and firmware versions are fully aligned and supported.