----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/16054/#review30433 -----------------------------------------------------------
Ship it! Sorry for the delay on this review, some comments inline, none of which I consider ship-blockers. src/main/java/com/twitter/aurora/scheduler/SchedulerLifecycle.java <https://reviews.apache.org/r/16054/#comment58221> Is this true? Not clear to me how we get that from InterruptedException (and if it is true why isn't it the same phrasing as above?) src/main/java/com/twitter/aurora/scheduler/SchedulerLifecycle.java <https://reviews.apache.org/r/16054/#comment58222> You add e.toString() in one message, but not the other. Mind standardizing here? src/main/java/com/twitter/aurora/scheduler/SchedulerLifecycle.java <https://reviews.apache.org/r/16054/#comment58223> Should storage teardown be moved into the finally block? I don't see an obvious reason to not attempt storage teardown when driver teardown fails. src/main/java/com/twitter/aurora/scheduler/SchedulerLifecycle.java <https://reviews.apache.org/r/16054/#comment58224> Should storage teardown happen after this? AFAICT the thrift and UI components can handle the case of the driver being shutdown gracefully (since they have to handle a disconnected master anyway) and doing it in this order just generates ugly user-facing error messages. But I might be missing something, and as this is the current behavior I'd be okay with a TODO punting this to another review. src/test/java/com/twitter/aurora/scheduler/SchedulerLifecycleTest.java <https://reviews.apache.org/r/16054/#comment58225> Drop Twitter copyright src/test/java/com/twitter/aurora/scheduler/SchedulerLifecycleTest.java <https://reviews.apache.org/r/16054/#comment58227> Mind adding a brief prose description of what's expected here? - Kevin Sweeney On Dec. 13, 2013, 12:09 p.m., Bill Farner wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/16054/ > ----------------------------------------------------------- > > (Updated Dec. 13, 2013, 12:09 p.m.) > > > Review request for Aurora, Kevin Sweeney and Zameer Manji. > > > Repository: aurora > > > Description > ------- > > This change addresses two issues: > - Ensure leadership is canceled whenever onDefeated is called > - Scheduler should wait for registered to be called before attempting to > invoke driver > > Some additional structural changes were made: > - Driver.run() is no longer used. Instead, we invoke Driver.start() > (non-blocking), and the lifecycle uses Driver.join() to await exit. > This allows us to avoid jumping through thread-safety hoops in unit tests. > - A shim interface (DelayedActions) was added to SchedulerLifecycle to make > testing easier when capturing delayed closures. > > > Diffs > ----- > > src/main/java/com/twitter/aurora/scheduler/Driver.java > e8fe170b2d9e1a752b152cedc0e006f10eaa5a11 > src/main/java/com/twitter/aurora/scheduler/SchedulerLifecycle.java > 346d52acc5fc9e4841e2dc8b424fc6d46d2cdc8c > src/main/java/com/twitter/aurora/scheduler/SchedulerModule.java > bd7929d631cf45b4c2c7f39177bbafbd8f659071 > src/main/java/com/twitter/aurora/scheduler/app/SchedulerMain.java > ec99e01dd4f982ed5b794ff86b65e6f58d2a0e27 > > src/main/java/com/twitter/aurora/scheduler/storage/testing/StorageTestUtil.java > ceef9d3bd4b43d2dcac0ab9129ae2b624ab654cf > src/test/java/com/twitter/aurora/scheduler/DriverTest.java > 5609b0b8b9a26a737558316e8e4fab0235704cb8 > src/test/java/com/twitter/aurora/scheduler/SchedulerLifecycleTest.java > PRE-CREATION > src/test/java/com/twitter/aurora/scheduler/app/SchedulerIT.java > 4c381b946c8a3c7bbe1757d384d5d43dd74bb4d0 > > Diff: https://reviews.apache.org/r/16054/diff/ > > > Testing > ------- > > gradle clean build > > > Thanks, > > Bill Farner > >
