Thanks for the information Jesse.
So I have tried some of the suggestions below.
Here is a sample config from hadoop-metrics.properties:
# Configuration of the "jvm" context for ganglia
jvm.class=org.apache.hadoop.metrics.ganglia.GangliaContext
jvm.period=10
jvm.servers=10.1.1.25:8649
In my gmond.conf on the 10.1.1.25 machine, there are several lines
that are similar to this one,
udp_send_channel {
mcast_join = 239.2.11.71
port = 8649
ttl = 1
}
1) it has a mcast value, should this be removed since I am not using
the mcast value in the hadoop properties?
2) how do you disable the local ganglia monitors from running?
Because when I run "gmetad --debug=9" I get a ton of data, but I think
that it is all from local, because it is the only environment that
shows in the web ui.
-John
On Nov 30, 2009, at 11:24 PM, Jesse Becker wrote:
> On Mon, Nov 30, 2009 at 17:08, John Martyniak
> <[email protected]> wrote:
>> Hi,
>
> Hello.
>
>> I am very new to ganglia, I have ganglia installed and working for
>> the
>> local machine.
>
> Welcome, and congratulations.
>
>> But when I try to have ganglia collect information on the running
>> cluster it does not get any results.
>>
>> So I guess I have a couple of specific questions:
>> 1) does each member of the cluster need to run gmetad or gnome?
>
> No. Typically, only the machine that will handle the RRD files needs
> to run gmetad. This is usually also the webserver, but it does not
> strictly have to be.
>
>> 2) I have machine 10.1.1.25 which is running ganglia web, gmond, and
>> gmetad, should that be the machine that each cluster machines sends
>> it's data to? On port 8649 or should it use port 8651?
>
> I'm not completely sure how Hadoop handles sending metrics around, so
> bear with me...
>
> The gmond processes communicate using UDP port 8649, for both sending
> and receiving metrics. It also listens on TCP port 8649 for polling
> requests from gmetad (or anything else that wants to read the XML dump
> for that daemon).
>
> Gmetad listens on port 8651 for incoming requests, usually from the
> web frontend. It also supports rudimentary interactive sessions on
> port 8652. Gmetad is responsible for polling one or more gmond
> processes using TCP port 8649, as mentioned above.
>
> There's a pretty picture of this interaction in figure two on the page
> http://www.linux-mag.com/id/6776 about halfway down.
>
>
>> 3) if #2 is corect then in the hadoop-metrics.properties, it should
>> be
>> set to 10.1.1.25:8649 or 8651?
>
> Not sure about how hadoop works here... If this setting relates to
> *sending* metrics, you probably want use 8649. If this setting is to
> allow an external gmetad process to poll the hadoop box directly, then
> you want 8651.
>
>> 4) when creating the data source on the 10.1.1.25 machine, should the
>> associated machine be the 10.1.1.25:port or should it be all of the
>> individual machines in the cluster?
>
> Probably just the 10.1.1.25 box. If you have mutlple hadoop boxes,
> you will probably want them all to send metrics to 10.1.1.25.
>
>
>>
>> Also I am running hadoop 0.20.1 and ganglia 3.1.3 or 3.1.4 (don't
>> have
>> the exact version in front of me.
>>
>> Thanks in advance for the help.
>>
>> -John
>>
>> ------------------------------------------------------------------------------
>> Join us December 9, 2009 for the Red Hat Virtual Experience,
>> a free event focused on virtualization and cloud computing.
>> Attend in-depth sessions from your desk. Your couch. Anywhere.
>> http://p.sf.net/sfu/redhat-sfdev2dev
>> _______________________________________________
>> Ganglia-general mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/ganglia-general
>>
>
>
>
> --
> Jesse Becker
------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general
