On Thu, Sep 13, 2007 at 08:48:29AM -0400, Fernanda Foertter wrote:
> Hi Everyone,
> 
>       I had a quick question.  We use Ganglia as part of our Rocks Cluster  
> setup.  But we have port 80 closed to the outside.  How can I  
> securely, somehow, monitor the cluster and allow others to monitor  
> them without opening the cluster to the world?

I am not familiar with the rocks setup, but most likely all you need in your
cluster is to have gmond/gmetric installed in all nodes and running so that
all the metrics are collected.

in one of the servers (most likely not as part of the cluster as the profile
for this machine is most likely different than your cluster nodes as it
requires much less CPU but a very fast disk or lots of memory for a ram drive),
you had to have gmetad which is collecting all those metrics and writing them
into RRD files, which then can be accessed by the web frontend to show them
for monitoring purposes.

>       Whats the best way to setup a secure web-monitoring?  Perhaps a  
> remote web-server that accesses the remote cluster DB?

the web server that runs the PHP scripts for showing the cluster monitor data
will be most likely the same one that contains the monitoring data (RRD files
generated by gmetad), there is no need for the whole cluster to be accessed in
port 80, but gmetad has to be able to pull the data from the gmond and for
that depending on your setup, you might need to allow access from this web
server to your cluster node in port 8649.

>       Another question: Is there a way that Ganglia can email admins when  
> servers go offline?  Perhaps its trivial and I missed...

no, but you could trigger it by a crontab that monitors the XML definition of
your cluster from ganglia and looks for HOST entries that are missing
compared with the last iteration.

and interesting distinctions is that, at least in my opinion, the description
of ganglia as a monitoring tool is deceiving, because it is not meant to serve
as a health monitor for specific machines like nagios would do, but to monitor
the health of the whole cluster and be probably better be labeled as a
"capacity trend" tool.

Carlo

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Ganglia-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ganglia-general

Reply via email to