Issue I noticed when moving from HPE to DELL hardware for VMware environments and sharing solution from VMware support. When deploying Dell PowerEdge R640/R650 servers with the Broadcom NIC’s, we have seen issues with Jumbo frames being dropped. This has been caused by the bnxtnet drivers attempting to create more rx queues than the ESXi host will allow – as such traffic is intermittently dropped.
The rx queue limit is increased as of ESX 7.0, however while deploying 6.7 the following work around is required on each ESXi host:
- esxcli system settings kernel set -s netPagePoolLimitPerGB -v 15360
- esxcli system settings kernel set -s netPagePoolLimitCap -v 1375920
You can confirm the changes with the following:
- esxcli system settings kernel list -o=netPagePoolLimitPerGB
- esxcli system settings kernel list -o=netPagePoolLimitCap
The ESXi host will require a reboot for these to take effect.