I'm not really familiar with wireless mesh networks, but I'm assuming
that it is a bit different than a wired mesh network (a mess!) and
allows for some flexibility with which nodes connect to which. Would
load-balancing/channel bonding be an option here? You might even be able
to bond all 3 links into "one," but this would keep a node from
simultaneously connecting to multiple neighbors (I think), and the wired
interface would be a lot less flexible than the wireless links...
Well, here's a walkthrough somebody did with wireless channel bonding:
http://www.linuxhelp.net/forums/index.php?showtopic=8478 (although it's
really no different than wired ethernet channel bonding)
Of course, follow his steps with caution. It's strange advice to set
miimon=0 (around 100 is normal, I think), and I'm not sure of the
relevance of the "my" interface/bridge to your situation.
Of course, it might be a better idea to read through some documentation
on kernel.org:
http://www.kernel.org/pub/linux/kernel/people/marcelo/linux-2.4/Documentation/networking/bonding.txt
Or in /usr/share/doc/iputils-[some date?]/README.bonding
Hope this gets you started, at least!
-adam
On Thu, 2007-08-02 at 11:44 -0400, Andy Brody wrote:
> Hi,
>
> I'm running Ganglia on a wireless mesh network where each Gmond node has
> three network interfaces, two of which are wireless and one wired. At
> the moment each node is manually set to send to the head node over
> unicast on whichever interface has seemed most reliable for that
> particular node. Most of the nodes won't actually be connected via the
> wired ethernet, and the wireless connections aren't the most reliable.
> Is there any easy way to configure the nodes to fall back on other
> interfaces if one doesn't work?
>
> Using multiple udp_send_channel entries (one for each interface) created
> what appeared to be three separate machines with identical metrics,
> because gmond on the head node doesn't know that 192.168.1.52 is the
> same machine as 192.168.3.52. Setting the /etc/hosts file to map
> 192.168.1.52, 2.52, and 3.52 to the same hostname appeared to work, and
> the nodes were only shown once in the charts. Really, gmond was still
> keeping track of three machines, so the listed numbers of hosts up and
> down showed 3x as many nodes as I actually have.
>
> Is there some way for gmond to identify the nodes so it realizes that
> what appear to be three nodes are actually just one? Any suggestions
> would be appreciated.
>
> Thanks,
>
> Andy Brody
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Splunk Inc.
> Still grepping through log files to find problems? Stop.
> Now Search log events and configuration files using AJAX and a browser.
> Download your FREE copy of Splunk now >> http://get.splunk.com/
> _______________________________________________
> Ganglia-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/ganglia-general
-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general
