----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/48429/#review136881 -----------------------------------------------------------
src/docker/executor.cpp (lines 307 - 314) <https://reviews.apache.org/r/48429/#comment201986> If the `docker inspect` times out, is it possible that we still try to `docker stop` on a container docker does not know about? And is it possible to leak a container if: 1) `docker run ...` starts. 2) `docker inspect` does not complete. 3) Kill is issued. 4) `docker inspect` times out. 5) `_killTask` sends a `TASK_KILLED` and `docker stop`. 6) `docker stop` does not kill the container because it hasn't started yet. - Joseph Wu On June 8, 2016, 9:04 p.m., Benjamin Mahler wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/48429/ > ----------------------------------------------------------- > > (Updated June 8, 2016, 9:04 p.m.) > > > Review request for mesos and Jie Yu. > > > Bugs: MESOS-4279 and MESOS-5195 > https://issues.apache.org/jira/browse/MESOS-4279 > https://issues.apache.org/jira/browse/MESOS-5195 > > > Repository: mesos > > > Description > ------- > > Currently, the docker executor will SIGKILL the 'docker run' > subprocess immediately upon receiving a kill task request. > This is problematic in that the 'docker run' subprocess is > responsible for streaming the stdout and stderr from the > container. When we kill the 'docker run' subprocess, we no > longer stream the output. > > > Diffs > ----- > > src/docker/executor.cpp 1b3a7795b1db83394d4b884c1041c341f88a7df1 > > Diff: https://reviews.apache.org/r/48429/diff/ > > > Testing > ------- > > sudo make check > > > Thanks, > > Benjamin Mahler > >
