> On June 2, 2014, 12:38 p.m., Ben Mahler wrote: > > src/slave/containerizer/isolators/cgroups/perf_event.cpp, lines 137-142 > > <https://reviews.apache.org/r/21451/diff/5/?file=599364#file599364line137> > > > > So one single RunState missing the container id causes everything to > > fail? Why is the container id an optional field? > > > > Chatting with Vinod it sounds like it was an artifact of UUID not > > having a default constructor. > > > > If we can make it non-optional that would clean this up.
Current semantics are that recovery is all or nothing. I thought RunState.latest was optional because of: slave/state.cpp:327 {noformat} // Find the latest executor. // It is possible that we cannot find the "latest" executor if the // slave died before it created the "latest" symlink. {noformat} - Ian ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/21451/#review44533 ----------------------------------------------------------- On June 3, 2014, 1:35 p.m., Ian Downes wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/21451/ > ----------------------------------------------------------- > > (Updated June 3, 2014, 1:35 p.m.) > > > Review request for mesos, Benjamin Hindman, Ben Mahler, and Vinod Kone. > > > Bugs: MESOS-1278 > https://issues.apache.org/jira/browse/MESOS-1278 > > > Repository: mesos-git > > > Description > ------- > > Three new slave flags: > 1. perf_events is a list of perf events to sample. This list is validated > through perf during isolation creation. > 2. perf_interval is the time between starting new samples > 3. perf_duration is the duration of each sample > > e.g., you could sample cpu-cycles and cpu-migrations for 10 seconds every 60 > seconds. > > Isolator::usage() will initially return an empty PerfStatistics protobuf > (containing only the timing information) until the first sample is obtained. > After that it will return the most recent sample. > > > Diffs > ----- > > src/slave/containerizer/isolators/cgroups/perf_event.hpp PRE-CREATION > src/slave/containerizer/isolators/cgroups/perf_event.cpp PRE-CREATION > src/slave/containerizer/mesos_containerizer.cpp > 14380240512d29b49c2e8f2831a9ef8ca102a024 > src/slave/flags.hpp 15e5b64fd24a9381074b4833a0403314e1f404bc > src/tests/isolator_tests.cpp b0eff5721b6ea4dcee8ff3748a2a11ade809e30e > src/tests/mesos.cpp 3065ae30bceb8b258d071110c0360f701ff48b64 > src/tests/slave_recovery_tests.cpp 44ffac40b9edc9940f17b5fbe1848d56cf53b69b > > Diff: https://reviews.apache.org/r/21451/diff/ > > > Testing > ------- > > tests: > 1. sampling of the perf event isolator > 2. slave recovery for roll forward > > > Thanks, > > Ian Downes > >