Comware

Hello Everybody,

I am new to Comware Switches and I have a little Problem with Microsoft Multicast NLB (Win 2003)

on the 7503 Switches.

I already have searched throug the Forum but i cant find the right answer to my Problem.

My Setup looks like this:

- I have two 7503 Switches running in an IRF.

- Both Terminal Servers are Connected on the Gig Interfaces (lets Say - gig 1/4/0/1 and gig 2/4/0/1)

- FW Running is the: 7500-cmw520-r6708p03.app

- In my routed VLAN 4 i have all of my Serves and all of my Client PC's

My  Config so far looks Like this:

  undo arp check enable

  multicast routing-enable

  interface Vlan-interface4
   description "SVR-CLT-VLAN"
   ip address 192.168.186.1 255.255.254.0
   ip address 192.168.188.1 255.255.255.0 sub
   pim dm

Also I have a test NLB but only with one Interface - I see in the ARP Table that my Switch learned the Multicast MAC Address and I am able to connect.

Also i have Checked with the MS Mulitcast Tool that the Routing of the Multicast Traffic is Working.

My  Questions are now:

1. Must I Set an command like that:

mac-address multicast <MAC-Adresse> interface <Interface 1> vlan <VLAN-Id>

for my Config to get the NLB running?

2. Did i forget something important (the NLB will be going Live on Sunday...)

3. When I will use Multicast + IGMP  i will just have to enter:

igmp-snooping (globally)

igmp enable (in my VLAN 4)

It would be nice if a real Comware Professional could explain me this.

Thank You

I personally have not tested multicast NLB before, but hoping this info will help..

Previous post on this topic

http://h30499.www3.hp.com/t5/Comware-Based/Microsoft-NLB-Cluster-in-Multicast-Mode-on-vmware-hosts/td-p/5969351#.VRrvwUaXJac

>1. Must I Set an command like that:

>mac-address multicast <MAC-Adresse> interface <Interface 1> vlan <VLAN-Id>

>for my Config to get the NLB running?

Yes if your NLB is multicast mode, good explanation here

 http://h30499.www3.hp.com/t5/Comware-Based/Microsoft-NLB-Support/td-p/5334135#.VRroIkaXJac

Microsoft NLB can work in two different ways: unicast and multicast. If the NLB is working in multicast mode, so the only thing you have to do is to configure a static multicast mac-address entry in each port that the server is connected to, as a static ARP resolution to the virtual IP and MAC address of the NLB.

At the other hand, there was a limitation - in most of players, not only 3Com/H3C/HP, just to let you know - that this feature works fine only at L2, when there is routing involved, it works unstable.

If the NLB is working in unicast mode, you can enter the command undo arp check enable, that actually disable the MAC Address learning in the ports, forcing the switch to act as a hub - what is a poor workaround.

An alternative is if the server could be configured in a aggregated mode, so the solution can work smoothly aggregating the ports manually in static mode, using Bridge-aggregation feature.

As you mentioned that the server was unstable, it seems that the NLB feature is working in multicast mode, what it means that the routing among VLANs - if needed - must be done by another switch, instead the one that the server with NLB is directly attached.

I don't know if I was clear, but I already faced the same problem before, and the solution was to attach the NLB server to a switch working at L2 and setup the inter-VLAN routing in another switch.

>2. Did i forget something important (the NLB will be going Live on Sunday...)

Refer to above explanation for good info

>3. When I will use Multicast + IGMP  i will just have to enter:

>igmp-snooping (globally)

>igmp enable (in my VLAN 4)

I'm pretty sure you have to enable PIM at your default gateway router to respond to IGMP joins etc

This info also seems useful

http://blog.patricknielsen.net/2012/01/hp-h3c-3com-7500-5800-5820-series.html

Daryl

First of all thank you Darylwan for having a look =)

I looked the Information in the Links which you Postet so i think the NLB should be working after setting the Multicast MAC to the Physical Interfaces which are on my IRF (Core) Switches.

Another Thing which Interestes me:

When the Servers are Teamed (HP) and i have one link of each Server at my IRF Core and the redundand Link on Seperate Switches - would it be enough to make on this Edge Switches the Same Entry - or how would that look like ?

To the Multicast IGMP Thing - Pim DM is already Running in my VLAN 4 and also the DGW is my Core Switch.

Thank You

HI.

I faced the same question some years ago and seems that doesn't changed nothing until now...

Daryl is correct when says that this scenario does not work properly when running in L3.

If your VLAN 4 has the gateway in another switch than the one that provides the servers connections, it's okay, but if the DGW is in your IRF you will face unstable multicast traffic forwarding. You can work this multicast traffic only in L2 in a stable way.

The solution in my case - similar as yours, working with a XRN stack of 5500G-EI - was move the gateway to another switch (at that time, a distribution switch in the server farm) and configure the same settings you mentioned.

Regarding your question about the server's connection it seems to me that you need the same settings applied in the redundant paths - I've never seen this kind of scenario before so I cannot ensure you that it will work fine in a convergence event.

Hope it helps.

Regards.