> On May 23, 2016, 10:03 a.m., Guangya Liu wrote: > > Did not go to detail for now, but post one early comment: It is suggested > > that every patch should be atomic, which means that you should merge your > > patch https://reviews.apache.org/r/47708/ to this one to make sure the > > test will not be failed once this patch merged.
I always post them separately and most shepherds appreciate it so that there's not too much crammed into a single patch. - Kevin ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/47707/#review134329 ----------------------------------------------------------- On May 23, 2016, 12:45 a.m., Kevin Klues wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/47707/ > ----------------------------------------------------------- > > (Updated May 23, 2016, 12:45 a.m.) > > > Review request for mesos, Benjamin Mahler and Jie Yu. > > > Bugs: MESOS-5256 > https://issues.apache.org/jira/browse/MESOS-5256 > > > Repository: mesos > > > Description > ------- > > Previously, the `Containerizer` class contained a static `resources()` > function for enumerating all of the resources available on an agent. > This made it hard to add new resources to this enumeration that were, > for example, containerizer-specific or tied to a new isolator module > loaded at runtime. > > This commit refactors the resource enumeration logic to allow each > containerizer to enumerate resources itself. It does this by adding a > new virtual `resources()` function, whose default implementation > includes the same logic as the old static version of `resources()`. > Individual containerizers simply overwrite this function to do their > own enumeration if they want to. > > Similar logic exists to push resource enumeration down into isolators > as well. As such, the logic for the Nvidia GPU resource enumeration has > been moved down into the Nvidia GPU isolator instead of being > hard-coded into the mesos containerizer itself. > > > Diffs > ----- > > include/mesos/slave/isolator.hpp 4be8c2bb409052e2e07138483408209384f41e23 > src/slave/containerizer/composing.hpp > f3eebd19bc9e6b3b8a969a2ad967b3e2909e0ee4 > src/slave/containerizer/composing.cpp > 15d059f0bbda4e8cb93c65c09327dde1e34d3e7b > src/slave/containerizer/containerizer.hpp > ff78b4d0fd4a3b862f6019fc757c16b7367cd3cf > src/slave/containerizer/containerizer.cpp > faa0c789dda8a6f36fdb6217b0bae270b6b8f2e2 > src/slave/containerizer/mesos/containerizer.hpp > a1a00020668f6da8d611f26e5637afffc87d09ba > src/slave/containerizer/mesos/containerizer.cpp > 75e5a32a3e70ec60a6800e21a621673184ea0956 > src/slave/containerizer/mesos/isolator.hpp > bacd86af42d16cb7c9b6622dfb298dcaa7007b75 > src/slave/containerizer/mesos/isolator.cpp > 253ff3cea8aff3e7a3051fb5a763cc081f455f18 > src/slave/containerizer/mesos/isolators/cgroups/devices/gpus/nvidia.hpp > 502204650192d5ea44aa631eac8eb37e051843f0 > src/slave/containerizer/mesos/isolators/cgroups/devices/gpus/nvidia.cpp > 8f81cb79c10261670efc9eaa8614751854f53806 > src/slave/slave.cpp ce0e7b1f1d17c3b82d835b0a6296ed7b1e9eeac1 > > Diff: https://reviews.apache.org/r/47707/diff/ > > > Testing > ------- > > make -j check ( which fails on one test -- see > https://reviews.apache.org/r/47708/ ) > > > Thanks, > > Kevin Klues > >