That's good. we need a new introduction/tutorial and documentation for eagle 0.5 when existing cases are ported to new arch.
On Sat, Aug 13, 2016 at 8:12 AM, Edward Zhang <yonzhang2...@apache.org> wrote: > Hi Eagle Community, > > There are some on-going significant changes in Eagle 0.5 regarding code > structure and features. These changes happen in develop branch, so we see > many PRs recently. > > One heads up is those changes may break the existing way of you running an > Eagle application. It needs one week or so to make all applications to run > in a new way consistently. > > Those significant changes are as follows: > > 1. Alert engine is decoupled into a separate component. Before Eagle 0.5, > application uses alert engine as a java library. In Eagle 0.5, Alert engine > becomes a multi-tenant platform where data from different applications can > potentially go into one single (or multiple) storm topology to be evaluated > based on some CEP rules. Also, Alert engine will support correlation of > different streams of events by declaring multiple input streams in one > policy. > > This decoupling brings extra benefit e.g. any data in Kafka can go to alert > engine directly for alerting purposing without having to write code if the > data in Kafka is already ready for setting up rule. > > 2. Application configuration and lifecycle management. Before Eagle 0.5, > application is one storm topology which we use local configuration file to > manage, start, stop in a linux box. But Eagle in production normally > consists of many topologies which process different type of data sources. > Operational cost is very big without Eagle to manage those applications' > configuration and lifecycle. > > In Eagle 0.5, application constitutes 2 parts, i.e. managed storm topology > and web resources. Both parts need configurations and those configurations > are managed by Eagle API. By this way, operations can start/stop/status > check storm topology remotely and manage multiple sites(datacenter) in one > organization. > > Another benefit of application management is that it is very easy to > onboard a new Eagle monitoring application because processing(topology) and > web(UI/API) are both managed by Eagle platform. > > 3. More features. Besides existing security features, Eagle community has > integrated more Hadoop monitoring related features into Eagle, including > Hadoop running job, history job, Yarn queue, JMX metrics, Namenode log, GC > log, bad node check etc. > > Because of #1 and #2, Eagle platform now becomes more open and friendly to > add any new monitoring features. > > Thanks > Edward >