[Linux-cluster] two nodes can't see eachother.

[rajveer singh]

On Thu, Feb 26, 2009 at 12:53 PM, Antonio Dias
<accdias+cluster at gmail.com<accdias%2Bcluster at gmail.com>
> wrote:

> Vincent,
>
> On Thu, Feb 26, 2009 at 17:39, Vincent VIAL <vincent at vial.tv> wrote:
> > I checked what you suggested but according to cman man page, broadcast is
> > used by default and multicast is only used if a multicast parameter is
> given
> > in the cluster.conf file.
>
> I'm not sure but I believe that now the default is multicast. And if
> so, the problem can be in the switch not letting igmp version packages
> pass through.
>
> I had this problem when deploying a RHEL5 cluster in a environment
> using Cisco switches. Since I didn't have access to the switches
> configuration, the workaround was to force my Linux boxes speaks igmp
> version 2. You can try this on your boxes. Edit /etc/sysctl.conf e put
> a line like this:
>
> net.ipv4.conf.all.force_igmp_version = 2
>
> And issue a "sysctl -p" or reboot your boxes.
>
> --
> Antonio Dias
>
> --
>  Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster
>

Hi,

I'm also trying to understand this problem. Is both cluster nodes are not in
same network?, if both nodes are pinging with each other through your
gateway router (may be your managed switch ). as cluster uses multicast so
you have to increase the TTL value of your outgoing packets.

for e.g. you can try out this iptables command

#iptables -t mangle -I OUTPUT -j TTL --ttl-set 60

I know this is not best rule, you can customize to suite your own setup and
requirement.

If doesn't help this then let us know your setup architecture.

Thanks,
Rajveer Singh
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20090227/8f194187/attachment.htm>