> On Feb. 2, 2015, 9:20 p.m., Ian Downes wrote: > > src/slave/containerizer/mesos/containerizer.hpp, lines 221-226 > > <https://reviews.apache.org/r/30510/diff/1/?file=843873#file843873line221> > > > > In my current working code I've added an Option<string> rootfs to > > struct Container. Would that be sufficient?
Sounds good to me! Can you pull this change out and we can commit that first? > On Feb. 2, 2015, 9:20 p.m., Ian Downes wrote: > > src/slave/containerizer/mesos/containerizer.cpp, line 1196 > > <https://reviews.apache.org/r/30510/diff/1/?file=843874#file843874line1196> > > > > There ordering considerations here for container paths which have > > common ancestory beneath the container. > > > > > > this won't work: > > /bar -> container/foo/bar # fails because foo doesn't yet exist > > /foo -> container/foo > > > > this works: > > /foo -> container/foo > > /bar -> container/foo/bar # succeeds, symlink is created inside /foo > > > > The filesystem/shared isolator explicitly chose to not implement > > re-ordering but it does check whether this is attempted and returns an > > error stating such. I'll disable all nested symlinks in master. I'll add a NOTE here. > On Feb. 2, 2015, 9:20 p.m., Ian Downes wrote: > > src/slave/containerizer/mesos/containerizer.cpp, line 1213 > > <https://reviews.apache.org/r/30510/diff/1/?file=843874#file843874line1213> > > > > This requires the parent of the joined target to exist? > > > > "ln -s /persistent1 /container/foo/persistent" will fail if > > /container/foo doesn't exist I'll disable nested symlinks. I'll add a NOTE here too. > On Feb. 2, 2015, 9:20 p.m., Ian Downes wrote: > > src/slave/containerizer/mesos/containerizer.cpp, line 1215 > > <https://reviews.apache.org/r/30510/diff/1/?file=843874#file843874line1215> > > > > I presume this always returns an absolute path? Yes, it is (if not, we need to make sure it is). > On Feb. 2, 2015, 9:20 p.m., Ian Downes wrote: > > src/slave/containerizer/mesos/containerizer.cpp, lines 1161-1167 > > <https://reviews.apache.org/r/30510/diff/1/?file=843874#file843874line1161> > > > > How can we ensure that such a filesystem isolator is being used? > > Shouldn't all of this functionality be in one place!? For now, in master, we will only allow relative non-nested container path. - Jie ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/30510/#review70617 ----------------------------------------------------------- On Feb. 2, 2015, 6:36 p.m., Jie Yu wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/30510/ > ----------------------------------------------------------- > > (Updated Feb. 2, 2015, 6:36 p.m.) > > > Review request for mesos, Ben Mahler, Ian Downes, Michael Park, and Vinod > Kone. > > > Bugs: MESOS-2031 > https://issues.apache.org/jira/browse/MESOS-2031 > > > Repository: mesos > > > Description > ------- > > Allowed Mesos containerizer to prepare and update volumes. > > Mesos containerizer setup volumes for the container. > > > Diffs > ----- > > src/slave/containerizer/mesos/containerizer.hpp > 802988c90ac872b0cefa5e28f06e6fec98e8d032 > src/slave/containerizer/mesos/containerizer.cpp > d712278428889ebdfd598537690138329d8464f0 > > Diff: https://reviews.apache.org/r/30510/diff/ > > > Testing > ------- > > make check > > > Thanks, > > Jie Yu > >
