Hi Bernard,
I already test on another server collecting data from only one cluster
source and the problem is the same.
I keep a configuration for each cluster, where I use different ports for
interactive and xml data. And the gmond for each cluster is configured on a
different multicast address and this is working this way for about 4 years!
In this way, I only use a one server and monitor 11 gmetad "grids". Each
gmetad daemon use a different gmetad.conf and each grid has a separated
"webhome", each one with a different configuration on conf.php to use
different ports.
I didin't know I was the only one using a single server for multiple gmetads
because the software allows it easily!
Apparently, I solved the problem running on my new server the same version
of gmetad I'm running on my old server.
With gmetad version 2.5.6 the problem doesn't happens! Now, I'm using the
3.0.7 frontend, gmond is mixed, ranging from 3.0.1 to 3.0.7 and it depends
the cluster monitored.
What have changed in gmetad from version 2.5.x? I was thinking theres some
network issue here, because there's a lot of switches and some clusters are
in another datacenter.
Anyway, thank you for you attention,
Regards,
----
Leandro Tavares Carneiro
Analista de Suporte Linux/Unix
On Wed, Jun 18, 2008 at 5:54 PM, Bernard Li <[EMAIL PROTECTED]> wrote:
> Hi Leandro:
>
> On Tue, Jun 17, 2008 at 11:02 AM, Leandro <[EMAIL PROTECTED]> wrote:
>
> > I use ganglia to monitor the clusters I work with for some years and I
> love
> > it! I thinks there's no software like this in the entire planet, not to
> do
> > what it does. :-)
> >
> > I have a very complicated configuration, where I run multiples gmetad on
> one
> > server, monitoring some clusters. I have done this because I need to view
> > every cluster individually and it gives me the flexibility to change
> servers
> > or move the configuration without great impact.
> >
> > It worked pretty well until I upgraded to the new version... I created
> > another server, to replace the old one, and installed the latest ganglia
> > version, 3.0.7.\
> >
> > In this new server, some clusters/grids keeps alternating the node status
> > from all down to all up. Seeing the error_log from apache, when it turns
> all
> > nodes to down the following message appears:
> >
> > PHP Notice: Undefined index: HOSTS_UP in
> /var/www/html/cluster_view.php
> > on line 25,
> >
> > Note I have each gmetad webfrontend in a separated directory under
> > /var/www/html and the port numbers are different for each source and the
> > multcast channel for the gmond is different too.
> >
> > I have some guys looking at the network, but must of this nodes are on
> the
> > same vlan on a Cisco network infraestructure. This problem can be cause
> by
> > network?
> >
> > Maybe I have done something wrong, but I have turned all my configuration
> > upside down and I doesn't found what is wrong!
>
> I have to say that this is the first time I have heard of someone
> running multiple gmetads/frontends on the same server although I
> cannot see why it will cause problems, since you are already using
> separate XML ports (8651 & 8652) and different locations for frontend
> docroot.
>
> However, I guess to troubleshoot your issue, perhaps what you should
> do is test whether *one* installation works fine. If it does, then
> there must be something wrong with the new setup.
>
> Let us know how it goes.
>
> Cheers,
>
> Bernard
>
-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://sourceforge.net/services/buy/index.php
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general
