EXOS VM (VMware ESXi): confuses MAC addresses

  • 0
  • 1
  • Problem
  • Updated 1 week ago
  • In Progress
  • (Edited)

ARP requests carry the MAC address of the Mgmt port instead of Port 1 which is member of the VLAN I am using to try to ping something outside.

This looks like a pretty basic fault that should not exist. So my question is whether I might be doing something wrong?

I have correctly activated only vNIC1 (Mgmt port) during installation of the VM. I have activated vNIC2 (Port 1) later. Using "show port" the port state correctly reflected all this.

I then did the following:
unconfigure vlan mgmt ipaddress
configure vlan m tag 12
configure vlan m delete port 1
configure vlan m add port 1 tag
ping vr vr-default

What I see at is what I described above: The ARPs are arriving via the correct vNIC but carrying the source MAC of the mgmt port, and not that of Port #1. Obviously, communication will not be established like that.

Vswitch port list:

[root@ESX:~] net-stats -l
PortNum          Type SubType SwitchName       MACAddress         ClientName
33554442            5       9 vSwitch0         00:50:56:82:7b:47  LAB_exos_MGMTVM
50331673            5       7 vSwitch1         00:50:56:82:2d:87  LAB_exos_DCA_CORE_F2
50331674            5       9 vSwitch1         00:50:56:82:95:af  LAB_exos_MGMTVM
50331684            5       9 vSwitch1         00:50:56:82:a6:41  LAB_exos_MGMTVM2
50331694            5       7 vSwitch1         00:50:56:82:bd:e0  LAB_exos_DCA_CORE_F1 *** vNIC2 / Port1
50331695            5       7 vSwitch1         00:50:56:82:64:23  LAB_exos_DCA_CORE_F1 *** vNIC1 / Mgmt / System MAC

Packet capture at vswitch illustrates it again:

[root@ESX:~] pktcap-uw --switchport 50331694 --dir input
The switch port id is 0x0300002e
The dir is Input
No server port specifed, select 36483 as the port
Output the packet info to console.
Local CID 2
Listen on port 36483
Accept...Vsock connection from port 1066 cid 2
16:00:22.47462[1] Captured at PortInput point, TSO not enabled, Checksum not offloaded and not verified, length 60.
        Segment[0] ---- 60 bytes:
        0x0000:  ffff ffff ffff ***0050 5682 6423*** 0806 0001
        0x0010:  0800 0604 0001 0050 5682 6423 c0a8 63d3
        0x0020:  0000 0000 0000 c0a8 6363 0000 0000 0000
        0x0030:  0000 0000 0000 0000 0000 0000
We see an ARP request with a source MAC address of 0050 5682 6423 whereas that should be 00:50:56:82:bd:e0 because I am using port switchport 50331694 with MAC address 00:50:56:82:bd:e0.


Photo of jeronimo


  • 1,384 Points 1k badge 2x thumb

Posted 1 week ago

  • 0
  • 1
Photo of Brandon Clay

Brandon Clay, Escalation Support Engineer

  • 13,486 Points 10k badge 2x thumb
This is expected. EXOS does not use per-interface MAC addresses. There is only one MAC address used across the entire switch, regardless of interface.
Photo of jeronimo


  • 1,384 Points 1k badge 2x thumb
Ok, then that is not the problem.

In any case, ARP replies sent to EXOS VM from outside: (this is a packet capture of the ESXi vswitch port egress)

11:40:58.27805[2] Captured at PortOutput point, TSO not enabled, Checksum not offloaded and not verified, VLAN tag 12, length 60.
        Segment[0] ---- 60 bytes:
        0x0000:  ffff ffff ffff 0050 5682 95af 0806 0001

seem to be correctly egressed from the vSwitch to the EXOS VM.

When the vSwitch is set to VLAN 12 and the EXOS port is set to untagged it's replying alright.

When the vSwitch portgroup is set to VLAN 4095 (VGT / all VLANS) and the EXOS port is set to tagged (which it would need to be in that case) then EXOS is not replying to the ARP request. The FDB is not being populated.

("configure vlan m add port 1 untagged" vs. "tagged")