Sometimes, through the normal operation of a hypervisor host, network interfaces (NICs) can become “overwhelmed” and be unable to respond to traffic fast enough as being served by an upstream network switch. This condition results in frames, which are subsequently being dropped from the receive buffer of the host, to go unprocessed or “missed”.
Though it is generally not a good situation when a host (and/or a downstream element from the host, like a VM) “misses” network traffic, many applications are robust enough to handle a few misses from time to time. However, if this condition is frequent and/or perpetual, it can cause production issues and alarms would be exhibited from Prism accordingly.
To combat a frequent/perpetual condition, there are several available options. For ESX and Hyper-V hosts, the receive buffer size of NICs can simply be increased. For AHV hosts, instead of increasing the receive buffer size, it is recommended to employ load-balancing across the available uplinks. More information regarding load balancing across AHV uplinks can be found from the Load Balancing Within Bond Interfaces sub-section of the AHV Best Practices Guide.