@Guillaume: Don't worry there is nothing to take personally. +1 On creating a thread and start a discussion about the features we want our clustering engine to have, In fact I will create one asap.
I will disagree with Guillaume on that its not usable in real life, as I initially implemented that to solve some "real life" problems of mine and to be more exact: a) Automatic Discovery of nodes b) Configuration replication c) Feature repository & feature state replication. etc. imho these are "real" problems, they do cover the needs that have been discussed and that have been around the last months (at least as expressed via https://issues.apache.org/jira/browse/KARAF-376). Any feedback and/or additional requirements/features are more than welcome and I would gladly work on them. Also note, that the statement "all nodes are roughly the same" is not very accurate, as each node can be configured to have different role inside the cluster. a) A node can be set to be just a receiver of events b) A node can be set to be just a transmitter of events c) A node can set a white list/black list by type of event and/or event id (e.g. a node could disable config replication, or even exclude a specifc PID from the list of replicated PIDs) On Wed, Apr 13, 2011 at 3:22 PM, Jamie G. <jamie.goody...@gmail.com> wrote: > Would not all nodes involved in a particular cluster be roughly the > same? Yes, a separate thread should be started to discuss the > technical issues. > > Jamie > > On Wed, Apr 13, 2011 at 9:44 AM, Guillaume Nodet <gno...@gmail.com> wrote: > > (Ioannis, please don't take that personally) > > > > We can start a discussion around clustering, but I strongly disagree > > with several design > > decisions that have been made by Ioannis while working on cellar. > > Mainly the assumption > > that all nodes are roughly the same. > > > > I think the cellar is a nice experiment, but I'm not so sure it's > > really so useful in real life. > > So, if this is to be imported in Karaf, I'd like that not to be in > > trunk for now until we can > > start a real discussion on what problems we want to solve and what's > > the best way to > > solve them. > > > > > > On Wed, Apr 13, 2011 at 10:19, Jean-Baptiste Onofré <j...@nanthrax.net> > wrote: > >> Hi all, > >> > >> Ioannis made a great work by using Hazelcast to create a cluster of > Karaf > >> nodes. > >> > >> He has already blogged around: > >> > http://iocanel.blogspot.com/2011/03/karaf-clustering-using-hazelcast.html > >> > >> I like the design and I think it could be very interesting to embed this > >> feature in Karaf 3.0.0. > >> > >> Ioannnis has kindly accepted to donate this work to Karaf. > >> > >> The next steps are: > >> - think about an assemblies extension to provide cluster ready > distribution > >> - document Karaf clustering in manual (mostly based on the Ioannis' > blog) > >> - add an example of Karaf cluster usage. > >> I propose to work with Ioannis around these steps. > >> > >> WDYT ? > >> Are you OK to ship Karaf clustering using Hazelcast in Karaf ? > >> Do you think that we need a kind of "agnostic" layer around clustering > >> (especially if some of us already started to work around a different > kind of > >> Karaf clusters) ? > >> > >> Thanks > >> Regards > >> JB > >> > > > > > > > > -- > > Cheers, > > Guillaume Nodet > > ------------------------ > > Blog: http://gnodet.blogspot.com/ > > ------------------------ > > Open Source SOA > > http://fusesource.com > > > -- *Ioannis Canellos* * http://iocanel.blogspot.com Apache Karaf <http://karaf.apache.org/> Committer & PMC Apache ServiceMix <http://servicemix.apache.org/> Committer *
