[Linux-cluster] TTL on multicast packets set to 1 by default?

[Nikolas Lam]

Some weeks ago I posted to this list reporting problems with my 2-node
RHEL5 cluster where I had each node on a different VLAN, connected by a
layer 3 router.

The problem was that the OpenAIS (cluster communications) packets were
not reaching the other node, so all sorts of annoying things were
happening.

Well, I've just had someone with extensive multicast experience help me
with some troubleshooting. We've discovered that the problem appears to
be that the packets are going out with a TTL of 1, which means the first
router to receive it will drop it without forwarding.

Here's a piece of tcpdump -v

04:47:23.167506 IP (tos 0x0, ttl 1, id 0, offset 0, flags [DF], proto
UDP (17), length 102) 172.16.99.50.5149 > 239.224.72.11.5405: UDP,
length 74
04:47:23.336194 IP (tos 0x0, ttl 1, id 0, offset 0, flags [DF], proto
UDP (17), length 102) 172.16.99.50.5149 > 239.224.72.11.5405: UDP,
length 74
04:47:23.538871 IP (tos 0x0, ttl 1, id 0, offset 0, flags [DF], proto
UDP (17), length 146) 172.16.99.50.5149 > 239.224.72.11.5405: UDP,
length 118
04:47:23.658161 IP (tos 0x0, ttl 1, id 0, offset 0, flags [DF], proto
UDP (17), length 102) 172.16.99.50.5149 > 239.224.72.11.5405: UDP,
length 74
04:47:23.826268 IP (tos 0x0, ttl 1, id 0, offset 0, flags [DF], proto
UDP (17), length 102) 172.16.99.50.5149 > 239.224.72.11.5405: UDP,
length 74
04:47:24.026863 IP (tos 0x0, ttl 1, id 0, offset 0, flags [DF], proto
UDP (17), length 146) 172.16.99.50.5149 > 239.224.72.11.5405: UDP,
length 118


So, I guess the question is, what is the correct method to set the TTL
to be a bit more reasonable, say 128?

Regards,

Nik Lam