in any case, the higher level point is : we might want to discuss wether we want to evolve to support microservices, or we continue supporting traditional VM style architectures, because depending on that decision, the approach to packaging and maintaining things will indeed be quite different ....
On Fri, Jan 30, 2015 at 2:48 PM, Evans Ye <[email protected]> wrote: > > That's also an issue when we'd like to ship docker to production. > Mesos with its frameworks might be a solution and I like it, But, > IMHO, that fixed the infrastructure we supported. It seems docker itself is > trying to provide an orchestration tool > https://github.com/docker/machine > But that still requires additional configuration management... > > jay vyas <[email protected]> 於 2015年1月31日 星期六寫道： > > Roman youre right, a real docker container doesn't do things the way we >> currently do in bigtop. >> >> If we embrace docker, we will be able to really drop alot of weight. >> >> Maybe the experimental bigtop 1x can deploy pure tarballs from apache >> distributions on apache mesos for the orchestration ? >> >> That would be verrrrrrrryyyyyyyyyyyy interesting ..... >> >> On Fri, Jan 30, 2015 at 12:45 PM, Roman Shaposhnik <[email protected]> >> wrote: >> >>> On Thu, Jan 29, 2015 at 7:09 AM, Evans Ye <[email protected]> wrote: >>> > I really like the idea to deploy hadoop stack in docker containers. >>> > Just want to catch up with you guys, are you planning to ship docker >>> images >>> > as a drop and run cluster deployment? or no? Anyhow that was just my >>> guess, >>> > I'm interesting in the challenge of this cool stuff :) >>> >>> That's a $64k question. On one hand, a kosher docker container has just >>> one single service inside of it. We can do that very easily. Then the >>> real >>> questions kick in: how do you orchestrate those, how do you provide >>> them with configuration, etc. Seems like you still need some kind of >>> a system like Amabari, etc. >>> >>> Thanks, >>> Roman. >>> >> >> >> >> -- >> jay vyas >> > -- jay vyas
