> On Feb. 3, 2015, 11:23 p.m., Bill Farner wrote: > > Is there a shortcoming in the executor API that prevents this from being an > > independent binary? > > Timothy Chen wrote: > Hi Bill, this is going to be a binary on itself just like command > exexutor. Do you have something else in mind? > > Bill Farner wrote: > Sorry for my late reply, gmail filter mishap on my end. > > It's my understanding that this will not be a standalone binary, but a > part of the slave binary (which, IIUC, is true of the command executor as > well). If this is incorrect, please excuse my ignorance and read no further. > > The distinction worrying me is that this is creating a first-class > executor implementation within mesos itself. I agree that in some respects > this is not dissimilar from the command executor, though i think coupling to > a shell is not as extreme as a third-party container system. This sets a > risky precedent for mesos, as it allows the project to make short-sighted > decisions coupling to the first-class executor, and fail to generalize its > APIs in ways that limits other executors when compared to those built in to > the slave.
Hi Bill, in my mind this isn't changing the executor API, it's implementing the API that works the containerizer system we're integrating with. Unfortunately a container system (or any) has limitations that we will have to overcome, and Mesos being a pluggable containerizer system will have to integrate with the limitations in mind. This isn't really a short-sighted decision as we're not changing the API to cope with Docker, we need to implement a specific implementation same as the Docker containerizer itself which is a implementation of the containerizer interface. And really thinking about the nature of the executor in Mesos, we're only creating the necessary ones that we manage ourselves, which is command executor for the Mesos containerizer, and a docker executor for the Docker containerizer, and we still allow custom executors via the API. - Timothy ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/29329/#review70849 ----------------------------------------------------------- On Jan. 17, 2015, 1:26 a.m., Timothy Chen wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/29329/ > ----------------------------------------------------------- > > (Updated Jan. 17, 2015, 1:26 a.m.) > > > Review request for mesos, Benjamin Hindman and Bernd Mathiske. > > > Repository: mesos > > > Description > ------- > > Add executor for docker containerizer, replaces the usage of command executor > > > Diffs > ----- > > src/Makefile.am 07bea1f > src/docker/executor.cpp PRE-CREATION > > Diff: https://reviews.apache.org/r/29329/diff/ > > > Testing > ------- > > make check > > > Thanks, > > Timothy Chen > >
