On Tue, Nov 25, 2003 at 03:41:24PM -0800, steven wagner wrote:
> Steve Feehan wrote:
> >On Tue, Nov 25, 2003 at 03:20:21PM -0800, steven wagner wrote:
> >
> >
> >>So I would check out your multicast config, both in the monitoring core 
> >>and on that system in general (what happens when you ping 239.2.11.71 
> >>while the monitoring core's running?).
> >
> >
> >This very well may be the problem. I am not able to ping  239.2.11.71
> >on the es40 whether the gmond is running or not. I also have to admit
> >that I know nothing about multicast... it "just works" on linux so
> >I never had to muck with it. Oh, and I am able to ping  239.2.11.71
> >on the linux systems.
> >
> >So I guess I need to go read a bit about multicast on tru64.
> 
> [insert critique of my lack of regard for quoting conventions here.]
> 
> It's my understanding (someone please correct me if I'm wrong) that a 
> ping to a multicast IP should return ICMP responses from every host on 
> the multicast channel.
> 
> In other words [pseudoshell alert]:
> 
> % ping 239.2.11.71
> 
> 64 bytes from host1 (icmp_seq=1)...
> 64 bytes from host2 (icmp_seq=1)...
> 64 bytes from host3 (icmp_seq=1)...
> 64 bytes from host1 (icmp_seq=2)...
> 
> And so forth.
> 
> It's slightly quicker than using tcpdump, which'd be the next step.  :)

Yes, that is what I saw on one of the working linux clusters. 
Actually, it didn't give the above results until I added the 
following route:

  route add -net 224.0.0.0 netmask 240.0.0.0 dev eth0

On this linux cluster, eth0 is a private network shared by the 
cluster (also the mcast_if in gmond.conf).

The ES40 doesn't have a private network... it is only connected to 
the public network. So in gmond.conf I have:

mcast_if alt0

When I ping 239.2.11.71 I get nothing, but when I ping the multicast
address from another host I DO get a reply from ES40's IP address.
Strange. And I don't know how to set the multicast route on the ES40...
well, I try the following:

  route add -net 224.0.0.0/8 -dev alt0

But that fails...

# route add -net 224.0.0.0/8 -dev alt0
writing to routing socket: Invalid argument
add net 224.0.0.0: dev alt0: ioctl returns 22
Invalid argument


Although, I suppose it would be just as good to use lo0 as this is
a one node "cluster". So what is the recommendation for a one node
cluster? Multicast is, well, is there a point? And what sort of 
route do I need?

I'll keep playing with this tomorrow :)

Steve

> 
> 
> 
> 
> -------------------------------------------------------
> This SF.net email is sponsored by: SF.net Giveback Program.
> Does SourceForge.net help you be more productive?  Does it
> help you create better code?  SHARE THE LOVE, and help us help
> YOU!  Click Here: http://sourceforge.net/donate/
> _______________________________________________
> Ganglia-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/ganglia-general

-- 
Steve Feehan

Reply via email to