Hi Denny, It's nice to get another fella looking at the system side of the force, amongst other _many_ things I do most of the distributed installs on behalf of Entwine and with 1.4 we've moved into rpm for the last mile (we use Jenkins for CI). You probably already know about Lars Kiesow work since he's the one behind the Matterhorn Matterhorn yum repo at Osnabruck. Configuring Matterhorn instances in distributed environment can be somewhat tricky in particular when it comes to the particularities of storage, role wise (admin,worker,etc) as of today it's a matter of dropping the right binaries in the right directory.
At Entwine though when it comes to packaging we went the "one RPM to rule them all" way, for the record when I proposed this on list it was received with mixed feelings since it meant loosing the "hot loading" (i.e. drop a jar in a dir and it's now available to Matterhorn because it's dynamically loaded by Felix). You might want to read that thread so you can make your own opinion, but basically it's about configuring by adding the right binaries in the right place (current setup documented in the community wiki) vs editing a config file (Entwine's approach for production systems). As we went the config file road, the way we do one RPM to serve all profiles is by including all jars in the package and using hard coded $profiles.properties for each available type of node. We are planning to write a series of "how tos" posts on htp://entwinemedia.com/blog including one about installation and configuration when it goes online I'll include the details about how we do this from Jenkins to the creation of the single package and its contents (including how we create the $profile.properties). In terms of Virtual Machines management I'm biased towards Vagrant but Boxgrinder looks quite interesting, ah..one more tool ;-) ************* Jaime Gago Systems Engineer [email protected] @JaimeGagoTech -- Entwine - Knowledge In Motion www.entwinemedia.com @entwinemedia On Sep 28, 2012, at 10:40 PM, Denny wrote: > Hi folks! > > I messed around with using boxgrinder and the new rpm repository to > build MH v-boxen. I don't have any trouble grinding out images, and I'm > tossing puppet and whatnot into the mix as well, so managing > configurations after grinding is easy, etc.. (I'll post the recipes on > github as I iron them out- eventually add openstack y todo, probably). > > I'm just wondering how the different profiles will work from a binary > installation perspective. I'm not sure how they work in general, to be > honest. Is it mostly a matter of turning on and off bundles in felix > and adjusting corresponding service URLs? > > I generate an internal RPM repository as part of my CI stuff, and I'm > just wondering if it is best to have everything in one package, or to > break stuff up somehow (admin, worker, engage, capture?). > > How are people envisioning the binary installation experience? How are > people handling binary installation currently? > > :Denny > > -- > I cannot give you the formula for success, but I can give you the > formula for failure--which is: Try to please everybody. > Herbert Bayard Swope > _______________________________________________ > Matterhorn mailing list > [email protected] > http://lists.opencastproject.org/mailman/listinfo/matterhorn > > > To unsubscribe please email > [email protected] > _______________________________________________ _______________________________________________ Matterhorn mailing list [email protected] http://lists.opencastproject.org/mailman/listinfo/matterhorn To unsubscribe please email [email protected] _______________________________________________
