Lost packets when using switch

[CiscoKid1981]

I'm totally stumped on this one. Can anyone think of a reason why I would be experiencing 25% packet loss when pinging between any two machines on my network when connected via a switch, and 0% packet loss when connected via a normal little hub?? This results in hosts losing connection frequently.

I have two ESX hosts, one VC, and one shared storage server networked via a Dell 2216 unmanaged switch. This is just a little test network to help me learn and experiment. Pick any machine in that setup and ping to any other and you experience dropped packets. Swap the switch with a hub and the problem is fixed. I have tried another Dell 2216 and even went and got a Cisco catalyst 3550 off the shelf with no configs, to rule out a faulty switch. Still get packet loss. As soon as i plug the hub back in, zero problems.

I have verified that the NIC's have automatic speed sensing and duplex, and even forced to a 100 full, 10 full, and 10 half, just to make sure. all experience packet loss when any of the switches are being used to network the devices.

There was a similar post yesterday that sounds the same to me:

Per someone's suggestion in that thread, I tried deleting the agent from one of my hosts and re-adding it to VC with no luck.

Someone also suggested a DNS issue which sounds promising. I don't have anything acting as a DNS server right now and neither of my hosts or VC have any DNS servers assigned though.

Thanks for any help anyone can provide!

[Scissor]

Maybe a MAC Address Conflict where two of your guests have the same MAC address?

[Scissor]

Perhaps a Speed/Duplex mismatch between the Ethernet card and the switch port? Are the Ethernet card and switch port both set to autodetect?

Bad Ethernet cable?

[mcowger]

Perhaps a spanning tree loop?

--Matt

VCP, vExpert, Unix Geek

[CiscoKid1981]

Scissor - thanks for your response.

A MAC address conflict it is. I originally ghosted my first ESX host and pushed the image to a second machine. So that explains it. I changed IP and host name and foolishly thought that would be good enough.

I ran an ifconfig and see that they both have the same vmwif0 address and different vmnic0 addresses.

How can I get the second machine assign a new MAC address?

[mcowger]

delete teh vswif interface and recreate it:

esxcfg-vswif

--Matt

VCP, vExpert, Unix Geek

[CiscoKid1981]

Hey mcowger, thanks for the response.

I was able to delete the interface. I am now trying to add it back with this command:

esxcfg-vswif -a --ip 20.1.1.20 -n 255.255.255.0 vswif0

It comes back with an "Error trying to perform operation Must specify portgroup"

so I add:

esxcfg-vswif -a --ip 20.1.1.20 -n 255.255.255.0 -p Service Console vswif0

and I get "Missing vswif name parameter or invalid syntax"

What is the correct syntax for the service console portgroup name?

When i give the esxcfg-vswitch -l command I have a "Service Console" portgroup but when I run esxcfg-vswitch -C Service Console I get a 0 value returned...

[mcowger]

Put 'Service Console" in quotes.

--Matt

VCP, vExpert, Unix Geek

[kjb007]

You need double quotes around "Service Console"

-KjB

VMware vExpert

[CiscoKid1981]

Thanks kjb!

I was able to add it back....but the MAC is still not unique. It added the same one. I went to my other ESX host and deleted and added back and it also assigned it the exact same MAC address. vmnic0 on both machines is still unique, but its assigning the same vswif0.

[CiscoKid1981]

http://communities.vmware.com/thread/18307

I went and created a new service console port on a new virtual switch on each of my esx hosts and they both generated the same MAC. That's not what is supposed to happen right?

[Lightbulb · Accepted solution]

What is the output of

cat /etc/sysconfig/network-scripts/ifcfg-vswif0

Try manuly adjusting the last octet and perform network restart. I am actually just kind of curious if this will work.

Or check out http://communities.vmware.com/thread/134858

[Scissor]

Not sure if using Ghost to install ESX to multiple hosts is supported? Can you reinstall one of your ESX hosts from scratch to rule out Ghost as the cause of the problem?

Edit: Found this thread that recommends against the procedure: http://communities.vmware.com/thread/179317

[CiscoKid1981]

Lightbulb!

Thanks so much for your response. I sucessfully changed the MAC address and i am now back on my switch with zero packet loss. Excellent work sir.

Let this be a warning to anyone who takes the easy way out and ghosts an ESX host instead of building one from scratch...

[CiscoKid1981]

Thanks Scissor

Well this would definitely be a reason ghosting is not supported. Perhaps whatever key the ESX host uses to populate a unique MAC is uniquely tied to each install (?) So simply ghosting would not create a new key. Which makes sense. I will keep that in mind. Thank you everyone for your input, you guys are extremely helpful as always. The tough question now is, who do I award the correct answer points to!