Hi let me emphasise the importance of traditional deployment and update scenarios via packages for real life scenario on a cluster (mean ten’ths to hundret’s) of bare metal machines. Microservices or docker is no solution to the handling problem, IMHO it adresses the dependency hell. If I would have to deploy an app which scales up for hundreds of machines I will go to the solutions of my company or something like Openstack Ironic. Docker does not help deploying.
Even Hadoop itself is an orchestra of many services interconnected to each other, way to much to be seen as micro services. They depend on each other, that’s the strong side of Linux dependency management via packages, but it must be done right. Bigtop is only at the first step right now. Cheers, Olaf
signature.asc
Description: Message signed with OpenPGP using GPGMail
