|Extended monitoring [message #16899]
||Tue, 26 April 2005 02:59
Originally posted by: ng.MYFIRSTNAMEkuhn.ch|
I'd like to write a UI that allows monitoring the CPU load, free memory,
etc. of many (64+) hosts. Therefore, I wanted to use TPTP for collecting
the data, and extend it by a view that is presenting the data I received
from the agents.
My current idea is to link this view to the monitor (like it does the
statistical view for single hosts). The problem now is, how can i set up
many hosts at once (do a configuration for every host and to start up
one by one is very uncomfortable)? Is there any scripting capability or
a programmatic way to set them up?
Or has somebody another idea than using the monitor as hook for my view?
Thanks a lot,
|Re: Extended monitoring [message #16961 is a reply to message #16899]
||Tue, 26 April 2005 11:50
| George Christelis
Registered: July 2009
I guess there are two problems here. Firstly you require the ability to
launch multple agents from a single launch configuration. This can be done
by writing a custom launch configuration that would take a list of hosts and
then launch agents on all these hosts. All the agents would be presented in
the Profiling Monitor. I don't see a need for you to write a new custom
view, as the current statistical view is capable of concurrently displaying
multiple graphs and this should meet your needs.
Currently, there is no scripting to set up multiple agents, but this should
not be too difficult to do programmatically. If you have a look at the
org.eclipse.hyades.perfmon plugin source, in the package
org.eclipse.hyades.perfmon.perfmon.internal.launchconfig you will see the
launch configuration code that the perfmon agent uses.The PerfmonLauncher
class is responsible for launching the perfmon agent, and it is this that
you would have to programatically iterate for each host.
The second issue is the runtime configuration of the agents. Essentially,
you wish to query a set of counters on all the agents without having to
configure each agent seperately. This is tricky as the counter IDs that are
sent back from each host are not guaranteed to be identical. Certain
counters will have the same IDS across all hosts, but some are dependent on
variable entities such as processes. One way to do this is to construct a
template that you wish to apply to each agent. Such a template would perform
matching against counters the agent sends back by name, and not by ID. This
template could be stored in the launch configuration and would be applied
once all the agents have launched succesfully.
The perfmon agent received commands from the workbench through the
DCAgentCommandUtil class in the org.eclipse.hyades.perfmon.utils.internal
package. Common commands are "start tracing a counter", "get the children of
this counter" etc.. Essentially your application of the template would
perform calls similar to these to query the required counters from the
Feel free to email me directly if you require more in depth information.
"Thomas Kuhn" <ng@MYFIRSTNAMEkuhn.ch> wrote in message
> I'd like to write a UI that allows monitoring the CPU load, free memory,
> etc. of many (64+) hosts. Therefore, I wanted to use TPTP for collecting
> the data, and extend it by a view that is presenting the data I received
> from the agents.
> My current idea is to link this view to the monitor (like it does the
> statistical view for single hosts). The problem now is, how can i set up
> many hosts at once (do a configuration for every host and to start up
> one by one is very uncomfortable)? Is there any scripting capability or
> a programmatic way to set them up?
> Or has somebody another idea than using the monitor as hook for my view?
> Thanks a lot,
Powered by FUDForum
. Page generated in 1.35001 seconds