I'd like to add to that, that in the specific case of cpu_num I have
to replace the collect_once with a collect_every in the
collection_group definition in gmond.conf in order to have the correct
SummaryInfo
collection_group {
collect_every = 1000
time_threshold = 1200
metric {
name = "cpu_num"
title = "CPU Count"
}
Gregory
On Tue, Jan 20, 2009 at 12:53 PM, Gregory Gabriel <[email protected]> wrote:
> Hi,
>
> I'm pretty new to ganglia so I'm still trying to get my head around
> some aspects.
>
> What I noticed is that in the __SummaryInfo__ for a cluster the total
> for cpu_num is not the total number of cpus for that cluster at all
> but rather for (in case of a 2 cluster) one node. I got around this by
> by defining another udp unicast channel for the gmond on the other
> node in the cluster (on each node) so all nodes have the data of all
> nodes.
>
> gmond on eg. node1 is configured like this:
>
> -----
> cluster {
> name = "clusterXX"
> }
>
> udp_send_channel {
> host = node1
> port = 8649
> ttl = 1
> }
>
> udp_send_channel {
> host = node2
> port = 8649
> ttl = 1
> }
>
> udp_recv_channel {
> port = 8649
> bind = node1
> }
>
> tcp_accept_channel {
> bind = node1
> port = 8649
> }
> -----
>
> gmetad on the collector is set up like this:
>
> -----
> data_source node1 node1
> data_source node2 node2
> -----
>
> This makes me wonder: does gmetad generate the SummaryInfo for a given
> cluster based on data it pulls from 1 node only? Or how does it work?
>
> I noticed a post about the same here:
> http://www.mail-archive.com/[email protected]/msg04765.html
>
> Thanks in advance,
> Gregory
>
------------------------------------------------------------------------------
This SF.net email is sponsored by:
SourcForge Community
SourceForge wants to tell your story.
http://p.sf.net/sfu/sf-spreadtheword
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general
