8:06 p.m.
Hi
Maybe your suggestion is the way to go - RTGov still generates 'situations' which
are made available to the alerts engine to determine if a notification should be raised.
So looks like there are two possibilities:
(a) RTGov somehow creates alerts directly, so alerts engine, RTGov, other components, etc
could independently raise alerts that are then used to create notifications.
or
(b) RTGov and other components create their own event to represent a 'situation of
interest', and the alerts engine interprets these to determine if an alert should be
create resulting in notifications.
I'm open to either approach, although (b) has the benefit of allowing the source
component (e.g. RTGov) to simple raises the 'situations' while not worrying about
issues such as dampening, which can be defined with the alerts engine.
Regards
Gary
----- Original Message -----
>
> ----- Original Message -----
> > > 3) Would you expect any component generating notifications to use a
> > > 'client
> > > api' or is direct publishing of the NotificationMessage a valid use
> > > case?
> > >
> > >
> >
> > Not sure if I get the "client api" use case, if you can add an
example I
> > can
> > check if the current architecture supports that use case.
> >
> > At the moment, the idea is that the plugin should listen from the bus,
> > but
> > you can create an alternative notifications implementation without having
> > that dependency from the bus (i.e. some thirdparty likes metrics and
> > alerts
> > components and they want to hard integrate them into other code, so they
> > won't use the bus but the API defined for the notifications can be
> > overrided
> > it with additional/alternative implementations).
> >
>
> The question was more about how the alerts/notification messages are raised
> -
> so instead of the alert engine, it may be rtgov policies identifying a
> 'situation' an needing to create an 'alert'. In this case, would
rtgov use
> some form of client API to report the alert, or simply create the alert and
> publish it on the bus?
>
In our present design the hawkular-alerts-engine should be responsible to
generate the Alerts from Events/Data.
The alerts engine is also responsible to manage the persistence layer of
alerts and rules definitions.
If I am understanding your scenario, it is like if we can "decouple" the
alerts engine, because we have a third party element that can raise alerts
(i.e. RTGov policies) and it can be interesting to create an Alert that can
have attached a list of actions/notifications.
It sounds not clear for me (at a first thought) I guess you can accomplish
same feature just letting RTGov to generate a simple data and attaching a
definition from the alerts engine, that sounds more natural.
But anyway it is interesting to understand this use case to see what is the
best way to implement it.
Thx,
Lucas
> Regards
> Gary
>