Hi Shao, > I am a student from National University of Singapore and majoring in Computer > Science. Currently, I am looking for a GSoC project and I came across your > org and saw the project on monitoring dashboard and I was chatting with rrolf > regrading this project and he shared with me a mail archive on the discussion > about JMX based monitoring. I understand that opencast is moving towards JMX, > can I roughly know what will be expose? In term of, application data or > system data? Does the system will need to set some trigger if it hit certain > threshold?
Matterhorn in its current state only offers a very low number of JMX metrics, mostly related to database pooling and we know that this needs to change. Depending on when you are coming on board, it will be part of your project to take a closer look at Matterhorn and together with the Matterhorn developers and system engineers decide on the metrics that need to be exposed. Generally speaking, there is a set of "default" metrics that can easily be exposed by any random set of JMX libraries and partly are already exposed by the JVM itself (memory, garbage collection etc.). In this project, we are more interested in coming up with a strategy and implementation around exposing application-level metrics, such das the number of workflows currently being processed, the number of services in the system, the number of online users etc. We have started putting together an initial page with a suggested set of metrics [1] so you can get a better idea of what we are talking about. After an initial verification that this is indeed a usable set, the next step would be to start implementing a subset of them. Matterhorn developers throughout the community are likely to be working with you on this. Last but not least, demonstrating and documenting how to access the statistics and how to take advantage of them would be the last step. Ideally, this would happen in the form of a dashboard as part of the Matterhorn admin ui but could as well be much simpler, depending on where you would like to put the emphasis. In the admin ui, being able to spot values that went above a given threshold would certainly be fantastic! Tobias [1] http://opencast.jira.com/wiki/display/MH/Metrics _______________________________________________ Matterhorn mailing list [email protected] http://lists.opencastproject.org/mailman/listinfo/matterhorn To unsubscribe please email [email protected] _______________________________________________
