Posted on December 27, 2021 3:09 am
 |  Asked by JAIME SANTOS AMANDI
 |  119 views
Tags:
1
0
Print Friendly, PDF & Email

Hello community,

I am struggling to make multicast static RP configuration work. I might be bringing some baggage from other vendor’s multicast conventions. Also I am using the container version of EOS c-EOS 4.26.4M

A few questions:

  • In EOS, we don’t need to (cannot) configure pim in the loopback interfaces. So we should be able to configure RP in a certain loopback interfaces even without pim running on it. Correct?
  • In my instances, static RP doesn’t work, no matter if I configure it in a loopback address (no pim on the interfaces) or in a vlan SVI (indeed pim on the interface). This is a sample of my configuration:

ip multicast-routing
router multicast
ipv4
routing
! with loopback not working
router pim sparse-mode
ipv4
rp address 10.0.0.3
sh ip int b | i 10.0.0.3
Loopback0 10.0.0.3/32 up up 65535
!
sh ip pim rp  ! this gives no output
!
! with svi not working
sh ip pim interface | i Vlan
10.1.8.3 Vlan100 sparse 1 30 1 10.1.8.4 0 0
router pim sparse-mode
ipv4
rp address 10.1.8.3

  • Then BSR RP does work but only with SVI (vlan) interfaces but not with loopback.

For testing purposes (lab), not having static RP is a bit of an inconvenience but not a big deal however not being able to have RP in a loopback is a bit odd. I am quite positive it works (from real deployments and from Arista docs) but not sure why it doesn’t work for me.

Any hint of this would be awesome.

Many thanks, J

1
Answered on December 27, 2021 7:18 am

Hi Jaime,

Thanks for reaching out.

Yes, In EOS, we should be able to configure RP on any interface even without PIM running on it i.e for the switch to consider itself the RP, there is no PIM required. However for RP to participate in multicast, we need to enable PIM accordingly on interfaces towards LHR and FHR. RP holds no significance without having a PIM neighbor.

Ideally, Static RP should work and be active on cEOS.Could you please provide the tech-support log to check configs and to investigate other outputs from the cEOS device?

show tech-support | no-more

 

Thanks,

Bhavana.

0
Posted by John Gill
Answered on December 27, 2021 10:48 pm

It may also be good to look at the output of:

show ip pim config-sanity

We would also like to know what it means when you say the multicast does not work?

0
Answered on December 29, 2021 11:32 am

Hello Bhavana,

I have pim enabled in all interfaces and neighborships established. However RP static is stubbornly reluctant to appear in the "show ip pim rp" command.

Regardless of the RP participating in multicast, this static setup should appear anyway, right? Nonetheless, as mentioned, I have peer PIM neighborships in up state.

Here I attach the tech support, rp sanity as well as the pim neighborship and lldp so you get an idea of the immediate adjacent topology.

When I multicast not working (i know,very generic :-)) I mean that I generate traffic with iperf on one side of the topology and I don't see any multicast state created (show ip mroute). Also tcpdump in the device itself show traffic not flowing.

Let me know if you need anything else and thanks for the help.

Jaime

0
Posted by John Gill
Answered on December 29, 2021 10:08 pm

It looks like since multicast has dependencies on the kernel, cEOS doesn't have support for multicast.  You probably see that Pimsm is crashing constantly, if you look at the logs.

There is an alternative to using the kernel for multicast state creation, and it *might* work for you by making this config change and restarting:

router multicast
ipv4
software-forwarding sfe

A bit more about the above config, which implements multicast forwarding and programming via BESS: https://eos.arista.com/eos-4-22-0f/multicast-forwarding-using-bess-mfa/

0
Answered on December 31, 2021 11:01 am

'software-forwarding sfe' made some improvements in terms of seeing the multicast state and RP is  now present in the topology. However I'm not sure this fully works. No (*,G) PIM join messages run up the topology from the receivers up towards the RP (using 'ip igmp static-group'). I'm tempted to go back to vEOS just for these multicast labs.

Thanks for the help.

0
Posted by John Gill
Answered on January 3, 2022 2:40 pm

Okay, just keep in mind that if you expect to send *,G PIM joins to the RP, you need to be the PIM DR on that segment.  Let us know if the same issues occur with vEOS-lab.

I know one thing that is tricky with iperf is that it doesn't seem to periodically send, or respond to, membership reports.  I have usually had to stop/start the client when testing multicast with iperf.

Post your Answer

You must be logged in to post an answer.