Hi out there

We have two DC's with 10G interconnection in  between - these connections are run as L2 links - put into a set of  nexus 5000 (the old nx5020) - acting access-switches - and uplinked to a  set of nexus 7009 which act as L3 switch for us.

We  have a cluster of vmware boxes in each site and are running MS windows  2008 machines with MS NLB for TerminalServices - in IGMP multicast mode -  in VLAN 21.

Now I looked in the log of the nexus 7000 and found that the PIM DR is "flapping" between the two sites from time to time:

2013  Nov 25 22:50:58 ve-coresw-01 %PIM-5-DR_CHANGE:  pim [26128]  DR change  from 172.21.159.253 to 172.21.144.3 on interface Vlan21

2013 Nov  25 22:51:54 ve-coresw-01 %PIM-5-DR_CHANGE:  pim [26128]  DR change from  172.21.144.3 to 172.21.159.253 on interface Vlan21

2013 Nov 25  23:26:07 ve-coresw-01 %PIM-5-DR_CHANGE:  pim [26128]  DR change from  172.21.159.253 to 172.21.144.3 on interface Vlan21

2013 Nov 25  23:26:10 ve-coresw-01 %PIM-5-DR_CHANGE:  pim [26128]  DR change from  172.21.144.3 to 172.21.159.253 on interface Vlan21

I am not that familiar with multicast but the basic concepts are there - in the vrf I have defined

ip pim ssm range 232.0.0.0/8

the vlan is defined as:

vlan configuration 21

  layer-2 multicast lookup mac

vlan 2001

under the SVI interface vlan 21 I have also defined - and there is a sample showning the nlb

interface Vlan21

  vrf member DMZ_21

  no ip redirects

  ip address 172.21.144.3/20

  ip pim sparse-mode

  ip arp 172.21.149.19 0100.5E7F.9513

these flapping should only occur if the keep-alives between the two sites are missed 3 times

The uplinks to the nexus 5000 are defined as mrouters

vlan 21

  ip igmp snooping mrouter interface port-channel5

  ip igmp snooping mrouter interface port-channel16

SW5020-01# sh ip igmp snooping vl 21

IGMP Snooping information for vlan 21

  IGMP snooping enabled

  IGMP querier present, address: 172.21.144.3, version: 2, interface port-channel5  -> the DR on the nx7k

  Switch-querier disabled

  IGMPv3 Explicit tracking enabled

  IGMPv2 Fast leave disabled

  IGMPv1/v2 Report suppression enabled

  IGMPv3 Report suppression disabled

  Link Local Groups suppression enabled

  Router port detection using PIM Hellos, IGMP Queries

  Number of router-ports: 3

  Number of groups: 3

  VLAN vPC function enabled

  Active ports:

    Po10        Po15    Eth1/3  Eth1/11

    Eth1/12     Eth1/13 Eth1/14 Eth1/15

    Eth1/16     Eth1/17 Eth1/18 Eth1/19

    Eth1/20     Eth1/25 Eth1/26 Eth1/27

    Eth1/28     Eth1/29 Eth1/30 Eth1/31

    Eth1/32     Po16    Po5

The  link between the two sites - and boxes - is running error-free. As far  as I can see there hasn't been any problems in that vlan since ??

If I look at f.ex spanning-tree the topology hast changed for long time in that vlan (2 weeks).

Could I harden the igmp multicast setup?

What is happening when a DR is changing? Will the multicast stop work or what happens?

As  far as I understood the DR is the service which forwards the multicast  traffic to the groups so if suddenly some re-negotiation occurs I would  expect that the active traffic will be interrupted.

here the actual MS NLB clusters adresses:

SW5020-01# sh ip igmp snooping groups vl 21
Type: S - Static, D - Dynamic, R - Router port

Vlan  Group Address      Ver  Type  Port list
21  */*                -    R     Po10 Po16 Po5
21  239.255.149.19     v1   D     Eth1/14 Eth1/19 Eth1/32
21  239.255.149.24     v1   D     Eth1/12 Eth1/15 Eth1/16
                                    Eth1/26 Eth1/31
21  239.255.255.250    v2   D     Po15 Eth1/11 Eth1/28
                                    Eth1/29
SW5020-01#

Any suggeestions?