DHCP (broadcast) packets not transmitted between hosts of a dvPortGroup/dvSwitch?

[cforger]

Hello,

I've been banging my head against the desk for most of the day and need some help.

From everything I read/researched, I'm told that within a dvPortGroup broadcasts are distributed, not blocked - Regardless of if the dvPortgroup contains multiple hosts or not - i.e. a broadcast from VM on Host1 in dvPortGrou-Network can be received by a VM on Host2 in the same dvPortGroup, assuming the IP address/subnet are matching.

However, I am unable to get DHCP requests from a VM client in Host1 to be seen by a VM DHCP Server in Host2.

If I migrate the VM Client from Host1 to Host 2, then the DHCP requests are successful, and the VM Client receives a IP from the DHCP Server.

This says to me that broadcasts are being blocked between Host1 and Host2, even though both are part of the same dvPortGroup (called dvPortGroup-Network).

I have a fairly simple setup;

ESXi 5.0 Update 1 (but I had this problem with non-Update 1 as well)

2 Dell M915 Blade Hosts, running above software

vSwitch0 - Contains 1 nic on each Host, set just for management traffic

dvSwitch-Network which contains dvPortGroup-Network

- Set to "None" for a VLAN (in fact I'm not using any VLANs anywhere at the moment)

- Originally didn't have a vmKernel, added one with the standard 255.255.255.0 subnet, the IP is part of my LAN

- Set to 1 physical nic per Host, to keep things very simple

- Load Balance is set to "Route based on physical NIC load" although I have tried Originating Virtual Port as well, and it may be a moot point because I only have the 1 nic per host in the dvPortGroup-Network.

I can enable  "IP Helper" from my L3 switch, but it doesn't seem to be receiving any of the packets (the stats counters stay 0), which is a further indication that my broadcasts never leave the Host that they originate from.

First of all, am I correct in thinking that Broadcast traffic should be supplied to the entire dvPortGroup, even across multiple hosts?

Second: What else should I be checking?

I've even tried toggling Promiscuous mode, and it didn't change anything (which I expected, as that's not the issue I'm seeing).

Thanks.

[VTsukanov]

> First of all, am I correct in thinking that Broadcast traffic should be supplied to the entire dvPortGroup, even across multiple hosts?

Yes, it really works

> Second: What else should I be checking?

I think that you need to check settings on physical equipment (cable, switch configurations, etc)

Additional suggestion :  "Route based on physical NIC load"  does not make much sense if the host have only one uplink for dvSwitch (for detail Load Based Teaming ...), try to change Load balancing to "Route based on originating virtual port"

[cforger]

Thank you for your reply,

So it can work - That's good, anything I'm doing outside of the vdSwitch is just a kludge then - I should focus on trying to make this work from the VMware side.

However, I don't know what to check on the physical switch if the dvSwitch isn't passing the data between the hosts. Each host in the dvGroup has an uplink to a physical nic, but that will simply pass everything as any NIC would. My switch is a Dell PowerConnect M6348, and is pretty much setup as default. Any suggestions to what could be blocking broadcast traffic at the NIC or physical switch level? I'll dig through my M6348 documentation and see if there is any additional "default enabled" features I should know about.

For Routing: Yes, I normally have 4 NIC's per host in the dvSwitch/dvPortGroup, but to simplify the problem for troubleshooting, I've dropped down to 1 NIC per host.  I have tried "Route based on originating virtual port" and it's the same problem as "Route based on physical NIC load" so I've left it alone for now - And it's all a moot point, as there isn't any load balance when I have the single NIC.

I'll report back if I find something in the M6348 docs, but I'm still open to suggestions.

Thanks.

[VTsukanov · Accepted solution]

You can also try:

* Assign a static ip for the machines are located on different hosts, and test connection (by icmp tcp does not matter).

* Install a sniffer and  and see what kind of traffic you see on both hosts.

* Сheck settings Native VLAN on switch ports, because then portgroup in dvSwitch has settings VLAN - None, it works in native vlan (from point of view physical switch)

[cforger]

I'm going to have to use a sniffer and see what I can make out of the traffic between the hosts.  I'll get back to the list once I have new information.

[cforger]

UPDATE:

I still experience the issue, but it's looking to be a tcehnical issue with the Dell blade switches I'm using. Dell is working with me on the issue, and can report one similar ticket in their system about this, so I hope to have it resolved soon.

[cforger]

Hello,

I wanted to follow up to report that I did get the problem resolved, and it was an issue with the Dell blade hardware. Once I updated my NICs and switches, the problem went away.

Additionally, I eventlayy had a 10 Gig M8024k go bad on me - Bad ports that wouldn't light. Dell replaced under warranty.

While there is not a correct answer for this, I will award do the only fellow who bothered to try and help. 🙂