I think we should be using existing message broker implementations instead of re-inventing it. For ActiveMQ we can use its failover transport; http://activemq.apache.org/failover-transport-reference.html
On Wed, May 28, 2014 at 10:22 AM, Imesh Gunaratne <im...@apache.org> wrote: > Hi, > > Please find my comments in-line: > > >> - We got to have a health check mechanism. May be with low frequency if >> what nirmal claims is true. >> > Yes we could make the frequency configurable. >> > > >> - What prevent us from publish/subscribe both in the health checker.. >> > With the current implementation we have not added a health checker to > publisher as it detects a connection drop at the time it tries to publish a > message. > > >> - I am not convinced about having consumer in SM. >> > The problem here is there could be N number of subscribers for a given > topic. Each subscriber has a health checker and each sends ping messages to > the ping topic (may be we could call this "keepalive"). > > So if we introduce a ping message consumer it will receive all the ping > messages sent by N subscribers. Now the question is, are we going to > introduce a ping message consumer for each health checker instance or are > we going to have one ping message consumer in a Stratos component? > > >> - health check should be independent of other system components. >> > Agree >> >> Thanks > > > -- > Imesh Gunaratne > > Technical Lead, WSO2 > Committer & PPMC Member, Apache Stratos > -- Best Regards, Nirmal Nirmal Fernando. PPMC Member & Committer of Apache Stratos, Senior Software Engineer, WSO2 Inc. Blog: http://nirmalfdo.blogspot.com/
