8

I am using ESXi 5.1 in the hosts, I have setup a distributed switch with corresponding uplink and port groups. When migrating to distributed switch, it seems that VMs from the same host can ping one another, but pinging VMs on different hosts fail.

I have used the default settings on creation of the uplink and port groups, it seems very strange and I cannot find anything useful on the internet.

Distributed switch topology figure shows all VMs on the same port group having green colors as well as the uplinks, it seems perfectly normal but VMs from different hosts cannot ping.

Any ideas? Anyone with similar problem?

enter image description here

maiky
  • 139
  • 2
  • 3
  • 12
  • Is the physical network good? All the physical ports share the same VLANs, etc? – Jim G. Dec 03 '14 at 23:34
  • Yes it seems ok. And there are no vlans. – maiky Dec 03 '14 at 23:37
  • Just to get this clear: Everything worked on non-distributed virtual switches and then you migrated everything to a distributed virtual switch and it stopped working? Same configuration on the port groups, same uplinks, no changes in the network? Can you migrate back to local switches (at the moment, it doesn't work anyway) and show us the configuration that works? – Mario Lenz Dec 06 '14 at 17:03
  • I could not ping VMs from different hosts before trying distributed virtual switches. So "everything worked" is not true, it was just normal the VMs could ping one another as parts of the same switch in the same host. The configuration was exactly the same, apart from the port group that was assigned to network interfaces. – maiky Dec 06 '14 at 21:02
  • On the second host itselft if you issue ping does it work ? – yagmoth555 Dec 07 '14 at 21:07
  • Ping works when I ping ESXi host to ESXi host and between VMs in the same ESXi host. But ping between VMs in different hosts does not work. – maiky Dec 07 '14 at 21:42

3 Answers3

3

I'm not sure that I understand your network setup. In your management port group you work with a VLAN trunk range of 0-4094. This implies you're working with tagged VLANs on your uplinks, see Edit the VLAN Policy on a Distributed Port Group in the vSphere Web Client.

On the other hand, the port group that gives you trouble works untagged. To the best of my knowledge, switches allow you to work with tagged or untagged VLANs, you can't mix them.

Did this work on your local virtual switches, i.e. were you able to mix the untagged port group and the VLAN trunking port group there without any problems?

edit: Try to use VLAN or VLAN Trunking on the port group that's not working at the moment.

edit2: If you use specific VLANs (not VLAN Trunking, I don't know if this works) on your port groups, you can make use of Network Health Check (KB article: Enabling vSphere Distributed Switch health check in the vSphere Web Client) to troubleshoot your problems.

Mario Lenz
  • 1,612
  • 9
  • 13
  • ok thanks, this is not the problem. I have made everything untagged and still no luck. I tried also creating a new vds with only one port group , either tagged or untagged didnot work. – maiky Dec 06 '14 at 21:07
  • I saw in a blog that for them they had to change the "teaming and failover load balancing" to "route based on IP hash" with the corresponding LACP changes , but still no luck.. its very strange.. – maiky Dec 06 '14 at 21:09
  • Your hosts only have one uplink so the load balancing policy shouldn't matter- all network traffic will go through this one uplink. Since you're talking about LACP changes: Do you use LACP? Can the VMs in vIMS02_IO ping each other across hosts? – Mario Lenz Dec 07 '14 at 09:25
  • No, this is what I want to solve. VMs cannot ping each other across hosts. When they are in the same host they can ping. – maiky Dec 07 '14 at 21:41
  • Sorry, I'm running out of ideas. I think it's time for you to open a support request with VMware. – Mario Lenz Dec 08 '14 at 19:47
0

Though this post is for Esxi 5.5, you may check whether this solves your issue.

You may also check here.

serverstackqns
  • 722
  • 2
  • 16
  • 39
  • Thanks but the first one has already been checked, the uplinks are OK and there is connectivity between different ESXi hosts, also no VLANs are used. The second one tried also no luck – maiky Dec 08 '14 at 16:55
  • Are all you hosts of the same version? – Yan Skursky Dec 12 '14 at 10:19
-2

i dont know if you have already solved that but thats a bug in some part of the cisco ucs firmware. should be fixed with the newest version if u want i can send u the link

thomas
  • 1
  • 1
    If you have the link (and one for an errata), why don't you just post it, together with a short summary of the problem? – Sven Apr 09 '15 at 13:41
  • I'm currently down voting this answer because of the lack of documentation / specifics. However if this ends up being the solution, I'll gladly up vote. – Reaces Apr 09 '15 at 14:12