Problem with DHCP accross Link Aggregation

[jababr]

Good morning everyone.

I have a curious question. I've tried to get answers on various websites, but found nothing yet.

I have a small infrastructure of tests with a single ESXi 5.0 host and a managed switch. The host is connected to Switch through two Ethernet (1Gbps) in Link Aggregation. I made the appropriate settings: I activated the LAG (without LACP) on the switch and I'm using IP HASH as a method of Load Balance (in virtual switch)

Everything works normal, but I can not use dhcp, for example, in vlan10: I have to set fixed IP addresses for computers that vlan to work. Regardless, the other VMs that are not in any vlan (with Untagged packets), function normally. I even have a VM just to download torrent which is directly connected and running without vlan. I am setting mode VGT (Virtual Switch Tagging) and LAG receives both Tagged and Untagged packets (vlan10).

Doing more testing I found that if the dhcp server is virtual (ie in a VM on vlan10) works OK, but if I put the dhcp server physical connected to a Tagged port of the switcth, there's only works on physical Untagged ports off the switch, turning on the computers on the switch (port untagged in vlan 10).

So to summarize, Tagged with packages, dhcp does not work going from virtual to physical and vice versa, but it works only in the virtual, or just the physical.

Who have some light, I will be eternally grateful. lol .....

Hug.

[weinstein5]

Welcome to the COmmuniry - It sounds like the problem is with the configuration of the physical switch -have you set up trunk ports on the physical switch identifying vlan 10?

[jababr]

OK.

Thankx for answer.

My switch is an SG3109 of TpLink.

I've tryed Trunk and General, on LAN and on port of Physical dhcp server, and variations (Trunk on Lag with General on dhcpd port and vice-versa, etc) whitout results. For all the tryes, untagged packets works normaly.

Should be a problem with my physical switch?

[JimKnopf99]

Hi,

what about ip-helper address for your dhcp server? Maybe that could be the issue.

Set the ip-helper address to your dhcp server.

We use cisco switches. Therefor i am not sure how to configure that for hp devices.

Frank

[rickardnobel]

I agree that it is most likely an incorrect VLAN tagging configuration on the physical switch. Do you have any inhouse networking team that knows how to deal with the VLAN tagging setup on your kind of switch?

Also, how is the networking setup on the ESXi host? Could you post a screenshot of the configuration, and also check the vSwitch NIC teaming policy?

[jababr]

My dhcp server is a x86 machine with m0n0wall.

I´ve no idea about id-helper. How I found that information on m0n0wall?

Attached is my serial console of m0n0wall.

Thankx.

[jababr]

Look at the attached screen shots of my switch and vswitch configuration.

My esxi host is conected on Ports 1 and 2 of switch (Link Aggregation).

My m0n0wall dhcp server is connected on port 3 of switch.

Thankx.

[jababr]

Anybody have another idea?

Thanks

[akshunj]

I see you have promiscuous mode, mac changes, and forged xmits enabled on the portgroup "lan10." Do you have them enabled on the vSwitch as well? In order for DHCP to work over your LAG you will need them enabled at the vSwitch level. You actually only need one of them, but I forget which one. That's probably why it works as a virtual machine, but not as a physical machine.

Our telephony guys usually spot this problem first, as the soft client phones will grab a DHCP address from the call server, but the physical VOIP phones never do.

[jababr]

Oh Yeah.

The promiscuou related settings are on de vSwitch and was inherited from vSwitch to lan10.

The default settings are: only "Promiscuous mode" reject and others accept.

Thankx for answer.

[akshunj]

OK I wasn't sure if you were overriding the vSwitch settings at the port group level.