On Saturday, May 20, 2017 at 8:24:20 AM UTC+1, Timo Reimann wrote: > On Friday, May 19, 2017 at 2:29:49 PM UTC+2, Rodrigo Campos wrote: > > On Friday, May 19, 2017, <morph...@gmail.com> wrote: > > Basically, this app can be a master Pod of sorts and that creates a > > multi-container pod through API on the cluster. > > > > > > You can use the API, but doesn't seem like a good idea to do that at first > > glance. > > Speaking very generally, I don't think that using the Kubernetes API from a > Pod is a bad idea per se. That's how most controllers operate, including > third-party ones. > > What the OP describes seems to be a kind of deployment service which, in my > opinion, could be designed and implemented in terms of the controller (or > possibly even operator) pattern.
Apologies for not giving the proper context. What I am trying to achieve here is, In a typical CQRS scenario, we have our read and write models separately. So from event log, we process the data and update the read tables in the db. We use kafka and we would have Kafka Streams Job running constantly, reading from topic and writing into topic after processing. Topic --> read & process --> Topic The way, kafka streams work is that, we deploy multiple instances of such job(jar) to load balance depending on the partitions. If you're new to this, think the streams job to be a stateless application packed in containers and we increase to scale up. lets assume, we do three steps in a sequential way..(for each read model) (consider a read model to be different tables in db - accounts, users, clicksByUser etc.) 1. preprocessors 2. processors 3. postprocessors So the data processing team would run a script which would build a master app with jars(above steps) and create a deployment in k8s. Master app(pod) to be deployed which would do these steps on by one, rollback if something fails, manage a bunch of these pre/post processors and finally deploy the read model. I use multiple master apps for each read model, or have a single REST layer which maintains multiple read model pods, either way, I would like to dockerize my application that is making the call to k8s. ^if the above does not make sense, because it is quite complex, all I want to know is, if its fine to have a master POD of sorts that will create multiple PODs using the k8s API. -- You received this message because you are subscribed to the Google Groups "Kubernetes user discussion and Q&A" group. To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-users+unsubscr...@googlegroups.com. To post to this group, send email to kubernetes-users@googlegroups.com. Visit this group at https://groups.google.com/group/kubernetes-users. For more options, visit https://groups.google.com/d/optout.
