----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/70449/#review214556 -----------------------------------------------------------
I didn't look at the test. Also, typo in the description. s/resubscriptions/resubscribes/ src/slave/slave.cpp Lines 5031 (patched) <https://reviews.apache.org/r/70449/#comment300730> Can you expand a bit here on why not publishing the updated executor resources (e.g., some tasks might have terminated while the agent was down) is not an issue? I remember this being safe because of lifecycle of resources vs publish/unpublish. - Vinod Kone On April 10, 2019, 11:36 p.m., Chun-Hung Hsiao wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/70449/ > ----------------------------------------------------------- > > (Updated April 10, 2019, 11:36 p.m.) > > > Review request for mesos, Benjamin Bannier, Greg Mann, and Vinod Kone. > > > Bugs: MESOS-9711 > https://issues.apache.org/jira/browse/MESOS-9711 > > > Repository: mesos > > > Description > ------- > > After an agent failover, an HTTP executor may resubscribe before > resource provider resubscriptions. If that happens and the executor > has tasks consuming resources from an unsubscribed resource provider, > the agent will fail to publish the resources and kill the executor, > which is an undesired behavior. The patch fixes this issue. > > > Diffs > ----- > > src/slave/slave.cpp 794a9c986b266b4916f6fbada670142798245bd1 > src/tests/slave_tests.cpp b1c3a01031b917fb9773c8c890a8f88838870559 > > > Diff: https://reviews.apache.org/r/70449/diff/2/ > > > Testing > ------- > > make check > > > Thanks, > > Chun-Hung Hsiao > >