Network getting hanged for IGMP Packets from Centos 6.6

General support questions
Post Reply
smcllt
Posts: 2
Joined: 2015/07/14 12:10:14

Network getting hanged for IGMP Packets from Centos 6.6

Post by smcllt » 2015/07/14 12:19:53

In my server i have two NIC Cards em1 and em2 with Centos 6.6 64 Bit Minimal Installed.
em1 : 192.168.x.z with default gateway 192.168.x.x
em2 : 192.168.y.a
Scenario 1
I am sending Multicast Packets for Multicast Group :- 224.0.0.x (whole series) from both the interfaces.in this case while sending from em1 my network gets choked up and if sending from em2, then the network is working perfectly fine

Scenario 2
i tried on CentOS 6.4 and 7.0 MInimal
everything fine in both the cases

for Scenario 1 i used 3 different type of Hardwares with 1G and 10G Nic Cards and the problem remains the same.
Totally screwed up because of this problem.
Thanks in Advance, your views are highly appreciated.

aks
Posts: 3073
Joined: 2014/09/20 11:22:14

Re: Network getting hanged for IGMP Packets from Centos 6.6

Post by aks » 2015/07/14 18:11:16

Not a solution, but perhaps a hint:
So generally with IGMP the "interested" parties "register" with their local multicast capable router. The question is, is that happening. Now this depends on the kit you're using (well usually the price tag of the kit you're using) - do you have something like igmp snoop on your router? If so use that to see if "interested" nodes are "registering". If you don't have that, you might have to capture the packets router and client side to see what's happening with IGMP.
Also make sure they're all using the right version of IGMP.
in this case while sending from em1 my network gets choked up
I'm not sure what chocked up means....

smcllt
Posts: 2
Joined: 2015/07/14 12:10:14

Re: Network getting hanged for IGMP Packets from Centos 6.6

Post by smcllt » 2015/07/20 06:42:48

We are trying to isolate the multicast traffic to the specific host by setting the TTL value to 0. When the TTL value on the multicast is set to 0, the relevant interface on the host machine is not supposed to forward the packet.

This is working as expected in centos 7.0 but on centos 6.6 it is forwarding the packets to a L2 switch which is inturn causing a multicast flood.

Maybe there is some inconsistency in the way centos 6.6 and centos 7.0 handles default setting(e.g: TTL threshold for an interface). Can someone please shed some light on this?

Post Reply