I had a very challenging issue on various ESXi 6.5 hosts. The hosts were various types of Dell Poweredge servers. All equipped with Intel quadport 10 Gbps network interface cards. Servers are attached to Cisco Nexus switches.
There were two issues:
- I could not ping two systems in the same subnet as the vmkernel. I could ping many other servers in the same subnet successfully. Other systems could ping the servers which could not be pinged from ESXi. The issue starts when the uplink is changed from access port to trunk port. I used both standard switch and distributed switch. I tried one nic in the host attached to the switch, and two nics. In all cases I could not ping the two systems. One of them a Netapp storage array.
- The other issue was vague as well. Virtual machines not in the same subnet as the Netapp storage, which had to pass a firewall, had an extreme bad performance when files where copied using SMB from the Netapp to the virtual machine. However file copy from the Netapp to VM’s in same subnet on the same ESXi host was fast. And a filecopy from a QNAP storage over the firewall was fast as well.For issue 1: What I saw when logged into the ESXi host with SSH was that the MAC-address of the two systems unable to ping were shown as incomplete. I noticed the ESXi host did send out an ARP request. However the response was never received or processed by the ESXi server.
Pretty strange issue. As said, when the trunk was changed to access mode, I could successfully ping all systems.
I decided to update the VIB driver for the nic. The driver i40en included in the Dell ESXi 6.5 distribution was version 1.1.0-1. The latest available on the website of VMware was 1.4.3 released at September 8, 2017. It can be downloaded here.
After applying the update and rebooting the server, both issues were gone!