----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/43015/#review118285 -----------------------------------------------------------
I realized a tricky part in the recovery path while dealing with potential conflicts with MesosContainerizer (LinuxFilesystemIsolator). For instance, will the LinuxFilesystemIsolator try to umount some persistent volume mounts that belongs to docker containerizer because it'll treat them as orphans. After checking the code, we're lucky here because of the way LinxuFilesystemIsolator detects orphans in LinuxFilesystemIsolator. There will be a mount from 'directory' (e.g., /var/lib/mesos/...) -> 'sandbox' (e.g., rootfs/mnt/mesos/sandbox). If we cannot find such a mount in LinuxFilesystemIsolator, it'll not try to umount volumes under 'directory'. However, I don't see any cleanup code during recovery in this patch. Do we need to cleanup volume mounts for orphan docker containers? - Jie Yu On Jan. 30, 2016, 10:43 p.m., Timothy Chen wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/43015/ > ----------------------------------------------------------- > > (Updated Jan. 30, 2016, 10:43 p.m.) > > > Review request for mesos and Jie Yu. > > > Repository: mesos > > > Description > ------- > > Fixed persistent volumes with docker tasks. > > > Diffs > ----- > > src/slave/containerizer/docker.hpp 77a50d80179672cf3c270cbdd7fa003c7d9ad6f3 > src/slave/containerizer/docker.cpp 2887cb4a01febbbf276026e584ffc466289e10c9 > src/tests/containerizer/docker_containerizer_tests.cpp > 645bdcf095145097d8b8c65d592c787417883145 > > Diff: https://reviews.apache.org/r/43015/diff/ > > > Testing > ------- > > make check > > > Thanks, > > Timothy Chen > >