I have recently started running some captures on my network using wireshark and I have noticed many TCP Out-of-Orders and TCP DUP Ack errors that come from my hosts on both of my ESX 3.0 servers.
Both the clients and servers are running on a 1GB connection.
Has anyone else seen this type of behavior with their ESX hosts?
How's your NIC teaming setup on ESX?
I don't have teaming setup. The two NICs on the ESX server are on separate subnets. One is on our internal LAN (192.168.x.x) and the other is on a subnet used only for iSCSI. The subnet in question here is the internal LAN (192.168.x.x)
i've never used wireshark before but I'm assuming like other sniffers - it might like to have promiscuous mode and forged transmits turned on.
try turning that on in your ESX and give it a go. also does your physical switch report anything out of the ordinary on that ESX port?
You're correct on the Wireshark config. I have my laptop set in promiscuous mode and I have my switch configured to mirror all traffic to that port. The switch hasn't reported anything of the ordinary. Also these errors do not show for any of my physical servers.
Also my laptop is on a 1GB connection as well so there shouldn't be any packet resizing.
just to clarify - when you say you have your laptop set in promiscuous mode; do you really mean your laptop or on the ESX server config? i.e.: ESX config and under networking properties.
My laptop. The server shouldn't need to be in promiscous mode.
I guess I should clarify my reply to the teaming question.
I only have the one physical NIC on this network but the ESX teaming configuration is set to "Route based on the originating virtual port ID" and vmnic0 is the only Active Adapter in the configuration.
Hello,
Wireshark is not the most reliable tool. It gave me those same results but when I used Observer the captures did not have the dup acks and ordering issues.