Multicast Does Not work in the same Vlan in the same Switch?

Daniel Mansilla · ‎05-26-2011

Hello.

My question is the next:

I am configuring one cluster red-hat (two machines). My problem is that in my "clúster" only show me one machine and appears this message in the log "conexion refused" . In the Switch is enabled igmp snooping and filter. I would like know if this information can help me for solution my problem. I Guess that something is cutting the multicast traffic between both machines.

http://www.cisco.com/image/gif/paw/6813/cat_multicast_prob.pdf

This document show the solution for problems between two switches. I only have one Switch for this machines.

Thanks so much.

Best Regards

Daniel Mansilla

Eugene Lau · ‎05-26-2011

Hi Daniel,

Firstly, I'd like to check what platform/version you are running. Example, IOS 6500 switches will constrain unknown multicast by default.

For multicast to be "known", there would be the need for a "multicast router" or similar function such as the IGMP Querier. Could you confirm whether you have a mrouter or similar function?

Some questions:

1. Does the cluster work when IGMP is disabled?

2. With IGMP Enabled, what happens when you configure a static multicast MAC on the port? Does this work as well?

3. Confirm that you ping is successful between the two servers (and for a sanity check, that they are in the same VLAN )

Can start from there.

Eugene.

Daniel Mansilla · ‎05-27-2011

Thanks for you answer.

Answering your questions:

First.- The platform is WS-C4948-10GE, IOS running cat4500-ipbase-mz.122-46.SG.bin

Second- Yes, my platform have features mrouter and query but aparent is not configured. attached the command output:

Some questions:

1. Does the cluster work when IGMP is disabled?

R: I don't know. This switch currently is in production and is very hard starting with tests.

2. With IGMP Enabled, what happens when you configure a static multicast MAC on the port? Does this work as well?

R:I have not configured anything yet. currently the switch is in production and can not make changes.

3. Confirm that you ping is successful between the two servers (and for a sanity check, that they are in the same VLAN )

R: The servers have normal comunication the ping is succesfull and I have not conectivity problems. Actually is in the same vlan.

My question is the next: If i configuring one mac statics or other solution of the guide. It the solution my problem?

Thanks so much

Daniel Mansilla · ‎05-30-2011

Hi Eugene.

have any news regarding my query?

Thanks so much

Daniel M

Eugene Lau · ‎05-31-2011

Hey Daniel,

It's difficult to say whether static multicast mac is the solution without knowing what's is the problem but we can continue to break it down. Sometimes, simple steps can help eliminate certain conditions but it seems that you are restrictred.

In your case, where there does not appear to be a L3 Multicast router in VLAN 11 (usually seen as an IP address in mrouter) the 4948 should allow any multicast and flood to all ports.

a. What is the multicast MAC & IP's being used by the redhat server for clustering?

b. Are you able to run a sniffer trace with a SPAN session if required? If it's simple for you, maybe get a sniffer capture of the heartbeat traffic from each server, like a layer2 check to ensure the servers are indeed doing what we should be expecting.

c. Are you able to do testing in a lab or pre-production type environment?

If you're running on a time line, it might be better for you to get a TAC SR opened and work with an support engineer. I'm flying around at the moment so I'll check in every few days

Eugene.

khwajanusrat · ‎07-20-2012

Hi,

I have a stack of 3750. Two AIX servers are defined in a cluster connect to the same switch-stack in the same vlan. Still the servers cant talk multicast together.

mping results are shown below:

beta:/#mping -v -r -c 5 -a 228.18.1.74

mping version 1.1

Localhost is beta, 172.18.1.84

Listening on 228.18.1.74/4098:

Replying to mping from 172.18.1.74 bytes=32 seqno=1 ttl=32

Discarding receiver packet

Replying to mping from 172.18.1.74 bytes=32 seqno=1 ttl=32

Discarding receiver packet

alpha:/#mping -v -r -c 5 -a 228.18.1.74

mping version 1.1

Localhost is alpha, 172.18.1.74

Listening on 228.18.1.74/4098:

Replying to mping from 172.18.1.74 bytes=32 seqno=1 ttl=32

Discarding receiver packet

Replying to mping from 172.18.1.74 bytes=32 seqno=1 ttl=32

Discarding receiver packet

alpha:/#

I disabled igmp snooping on this vlan but still no luck. Could anyone help please ?

khwajanusrat · ‎07-20-2012

Hi Daniel,

How did you solve this problem ? I am having the same problem with 3750 switch stack.

Eugene Lau · ‎07-22-2012

Question Nusrat,

1. Is there any particular reason why a 228.x.x.x is being used for local mcast?

2. are there any packets being dropped everytime you test mcast pings?

#sh controller ether-controller [int]

(tip:- clear counters for controller before testing but always get a snapshot prior)

3. Sanity check - are alpha and Beta in the same IPv4 network?

Eugene

khwajanusrat · ‎07-24-2012

Hi Eugene,

Answers follow your questions below:

1. Is there any particular reason why a 228.x.x.x is being used for local mcast?

no reason. Is 228.x.x.x reserved for something else ?

2. are there any packets being dropped everytime you test mcast pings?

#sh controller ether-controller [int]

(tip:- clear counters for controller before testing but always get a snapshot prior)

3. Sanity check - are alpha and Beta in the same IPv4 network? yes, same VLAN, ip subnet on the same switch

moreover,

donot know why:

cidbalphapte:/#mping -r -v -t -a 228.18.1.74

mping version 1.1

setsockopt() failed: Can't assign requested address