VMware vSphere

 View Only

  • 1.  Multicast flooding vSwitch

    Posted Apr 07, 2010 02:15 AM

    Hi all. Been trying to solve this problem all day. We have an application running on a windows 2003 server vm which sends multicast data out at around 50mb/s. Whenver the app starts to run, I notice that all the other VMs that are connected to the same vSwitch begin to experience heavy network latency to the point where connectivity is totally lost to the vNICs on those VMs.

    I've ruled out the physical switch (Dell Powerconnect 6224) being the issue since VMs on other hosts using the same network do not experience any latency. The latency only occurs on the machine running the multicast app.

    There was a post which suggested running some console commands to enable IGMP snooping but it used a command that is no longer available in ESXi4. Has anyone else experienced similar issues? If so, how did you solve it?

    Thanks!



  • 2.  RE: Multicast flooding vSwitch

    Posted Apr 07, 2010 08:12 AM

    Flooding is a problem with multicasting. We had this problem with NLB clusters. They use multicasting too. We changed our NLB clusters to Server Load Balancing (SLB) clusters. That solved our flooding problem.

    Robert



  • 3.  RE: Multicast flooding vSwitch

    Posted Apr 07, 2010 01:19 PM

    Hi. Thanks for your response but I'm not sure if it's directly related to the problem I'm seeing which is the multicast overwhelming the vSwitch within one host. The multicast data does not seem to affect the other hosts with vSwitches on the same network though.

    Thanks.



  • 4.  RE: Multicast flooding vSwitch

    Posted Apr 07, 2010 05:01 PM

    The flooding is a result of the multicast protocol in that the multicast floods all packets to all ports in the vlan. This is probably blocked on a physical switch in your environment. Therefore you see it only on the virtual switch connected to the server that generates the multicast packets. The article Using IP Multicast with VMware® ESX 3.5 gives an introduction about multicast in a VMware environment. Hopefully it will help you further.

    Regards, Robert



  • 5.  RE: Multicast flooding vSwitch

    Posted Apr 07, 2010 06:28 PM

    I've read that guide but as far as I can tell, vmware should be blocking the multicast traffic correctly like a physical switch. For now I've just allocated a physcial switch to the vm using the passthrough functionality but I'd much rather get it working via the vswitch.

    Thanks.



  • 6.  RE: Multicast flooding vSwitch

    Posted Jun 17, 2010 02:25 PM

    Hi. Ive discoverd the same problem. I have a ESX4i HOST with over 5x RHEL 5.5 VM's running.

    One of the VM's has a Multicast Based Bootsystem for some STB's in my lab and when running tcpdump on any another vm in the same subnet i see a ton of innbound multicast packets:

    15:59:37.704414 IP 80-xxx-x-xxx.dd.xxx.com.personal-agent > 23x.xx.xx.xxx.personal-agent: UDP, length 1068

    15:59:37.704414 IP 80-xxx-x-xxx.dd.xxx.com.personal-agent > 23x.xx.xx.xxx.personal-agent: UDP, length 1068

    15:59:37.704414 IP 80-xxx-x-xxx.dd.xxx.com.personal-agent > 23x.xx.xx.xxx.personal-agent: UDP, length 1068

    Its clearly stated in the ESX docs that IGMP snooping isnt needed, but infact it seems so...

    Just now 1 vm is flooding over 55Mb/s to all the other vms from what i can see from netstat and tcpdump.

    BR

    TE