The type property should really specify the type - as all these are registered as the same mbean with the same attributes, I don't think we should a HC allow to specify the type.
I think using the service ranking approach in combination with the name, is the better option - it's used nearly everywhere else in same situations and it allows you to overwrite a HC e.g. by deploying a better version coming from a different source. We can add a nice log message for these cases. (and have a HC checking this :) ) Regards Carsten 2013/8/14 Bertrand Delacretaz <[email protected]> > On Wed, Aug 14, 2013 at 4:18 PM, Carsten Ziegeler <[email protected]> > wrote: > > Ok, and the value of the hc.mbean.name is used as the property "name" > for > > the object name, right? > > Yes > > > So all HCs get the same fixed domain and type, they just differ by name. > > If you configure a name like foo/bar foo becomes the type, so you can set > it. > > Might be better to have an explicit hc.mean.type service property. > > > > > ...I already added a TODO to the jmx bridge, that we have to handle the > case > > where two HCs use the same "name". I think we should solve this by using > > service ranking and only registering the one with the highest ranking.... > > Isn't that hard to troubleshoot? Generating a unique suffix for the > second, instead of not registering it, might make it easier to detect > the error. > > > > > If no one disagrees, I'll do the changes in the next days, along with > using > > service properties. > > works for me. > -Bertrand > -- Carsten Ziegeler [email protected]
