> On March 6, 2006, at 8:57 AM, Duncan McGreggor wrote:
> Allen Lerner was asking about the distributed monitoring feature last
> night, off-list. After thinking about it this morning, I realized we
> could do a very simple implementation right now. It should scale
> well into the hundreds of pymon instances. Beyond that, we'd have to
> do some testing.
>
> Here's what I am thinking:
>
> * we add an XML-RPC over SSL service to each pymon instance
> * each node will need to be named for distributing (configuration)
> * each node will have XML-RPC methods like getName(), getData(),
> getConfiguration()
> * the XML-RPC methods will return standard types like strings, lists,
> and dicts
> * a configurable list of IPs/hostnames for pymon instances we want to
> consilidate with username/password for each instance
> * timed intervals for getting this data from other instances
I've decided to go another route with this. A simpler route :-)
Right now, there is a child resource in the pymon web interface called
"json". This resource is not useful in and of itself, however, it has
three children (for now): "localStates", "combinedStates", and
"localConfig". For now, I have localStates publishing service data as a
JSON data structure at the following pymon URL:
http://localhost:8080/json/localStates
State data will be made available in this manner as opposed to providing
an XML-RPC interface. This JSON mechanism should provide a suitable and
simple means by which distributed pymon instances can present unified
monitoring/service data.
> Also, the web interfaces will have additional breakdown: * the
> current/local instance's summary, details and configuration * a list
> of instances that the local instance has been configured to get
> (maybe with some markup/styling to indicate at a glance the
> overall/average state)
Working on this now :-)
> > Also, the web interfaces will have additional breakdown: * the
> > current/local instance's summary, details and configuration * a list
> > of instances that the local instance has been configured to get
> > (maybe with some markup/styling to indicate at a glance the
> > overall/average state)
>
> Working on this now :-)
Done.