I thought this was an interesting article, especially around syntactic monitoring systems: http://container-solutions.com/monitoring-performance-microservice-archit... _______________________________________________ hawkular-dev mailing list hawkular-dev(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/hawkular-dev

I think we all agree that having an inventory with relationships between components and their dependencies/dependents would be very helpful to perform more intelligent alerting (e.g. if a database's write performance is performing poorly, yes we can alert on the database resource, but any application that uses that database for storage should go "yellow" to alert the user that performance degradation is caused by the bad DB and that storage failures could be on the horizon. Expand this further - you could be monitoring a disk, see it is malfunctioning, and alert on all DBs using that disk, which in turn can alert on all apps using that DB). Hawkular Inventory can already store those relationships in a graph, IF those relationships are known. BUT! As i see it, it is the discovery of those relationships that is the issue. Sure you can possibly auto-discover SOME relationships (e.g. we could look at the <datasource> definitions in WildFly configuration and relate those databases to that WildFly application server), but many relationships are difficult if not impossible to auto-detect. For example, how does the WildFly agent know WHICH database resource to relate to its app server? How does the DB resource in inventory (presumably created by another feed) get known to the WF agent such that THAT DB gets related with THAT WF Server?). In other words, feeds/agents won't know about the full inventory "universe" to be able to create many relationships that you would think should be obvious to them. We have Gary's BTM which could be the answer to some of this (his injected monitoring code could be used to detect these relationships and automatically tell inventory about them - but again, how does it know about the full inventory identities of the resources it encounters to be able to relate them together). I think we need a really good UI story to help users map these relationships manually. The users will be able to tell us about these "hidden" relationships - relationships that cannot be auto-discovered. But it needs to be intuitive and easy to map out these relationships. ----- Original Message ----- > Yes interesting, this is why I think the inventory is being an important > element of monitoring. It's not enough to look at a single resource, we need > to look at how multiple resources are "connected". > > On Tue, Dec 8, 2015 at 4:15 PM, mike thompson < mithomps(a)redhat.com > wrote: > > > > I thought this was an interesting article, especially around syntactic > monitoring systems: > > http://container-solutions.com/monitoring-performance-microservice-archit... > > > > _______________________________________________ > hawkular-dev mailing list > hawkular-dev(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/hawkular-dev > > > > _______________________________________________ > hawkular-dev mailing list > hawkular-dev(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/hawkular-dev > _______________________________________________ hawkular-dev mailing list hawkular-dev(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/hawkular-dev