Mesos v0.16.0 Release Notes

    • ๐Ÿš€ The primary feature of this release is major refactoring work on the master election and detection process to improve its reliability and flexibility.

    All Issues: ** Sub-task

    • [MESOS-645] - Improve the performance of Future.

    ** ๐Ÿ› Bug

    • [MESOS-403] - CoordinatorTest.TruncateLearnedFill test is flaky
    • [MESOS-455] - ZooKeeperTest.MasterDetectorShutdownNetwork runs forever
    • [MESOS-463] - Detector ZNode creation failure.
    • [MESOS-465] - Failures due to ZooKeeper operation timeouts in the master detector.
    • [MESOS-498] - ZooKeeperTest.MasterDetectorTimedoutSession is flaky
    • [MESOS-536] - GarbageCollectorTest.Unschedule is flaky
    • [MESOS-592] - Don't dump a stack trace from bad --zk flag in the detector, use EXIT(1) instead of LOG(FATAL).
    • [MESOS-624] - Master improperly prints the exit status of the executor
    • [MESOS-641] - Stout killtree / pstree tests fail on Ubuntu 10.04.
    • [MESOS-778] - FaultToleranceTest.ReconcileIncompleteTasks test is flaky
    • [MESOS-782] - Slaves in local cluster should get unique work directories
    • [MESOS-795] - ZooKeeperTest.MasterDetectorTimedoutSession test is flaky
    • [MESOS-800] - CHECK failure in cgroups_isolator.
    • [MESOS-807] - Discard is not propagated in process::dispatch.
    • [MESOS-811] - Group::cancel can return a failed future if the membership is already cancelled
    • [MESOS-822] - AllocatorTest/0.SchedulerFailover is flaky
    • [MESOS-823] - ZooKeeperMasterContenderDetectorTest.ContenderDetectorShutdownNetwork is flaky
    • [MESOS-826] - Bad 'master' flag in slave should not print a stack trace
    • [MESOS-828] - CgroupsIsolator BalloonFramework Test is broken.
    • [MESOS-842] - ZooKeeperMasterContenderDetectorTest.ContenderDetectorShutdownNetwork runs forever
    • [MESOS-844] - Slave should not recover checkpointed data immediately after reboot
    • [MESOS-851] - Scheduler Driver does not guarantee that abort() prevents further calls on the Scheduler.
    • [MESOS-858] - Ignore launch/kill requests in the slave originating from non-leading masters.
    • [MESOS-859] - Cgroup kill should use cgroup.procs, not tasks
    • [MESOS-866] - Pailer popup window is not scrollable in Chrome or Safari
    • [MESOS-867] - ZK Membership IDs are 32 bit signed integers, not 64 bit unsigned integers.
    • [MESOS-870] - Slave http endpoint can crash the slave when no master is detected.
    • [MESOS-871] - GroupTest.RetryableErrors is flaky
    • [MESOS-883] - Group's handling of non-retryable errors and local timeout is incorrect
    • [MESOS-884] - Incorrect asynchronous detection and contention loops in Master
    • [MESOS-889] - Bad 'master' string given by scheduler should not print a stack trace
    • [MESOS-892] - Additional Issues with contender related change
    • [MESOS-935] - Group should tell MasterDetector "no memberships detected" when it locally times out
    • [MESOS-940] - Slave should checkpoint bootid after recovery instead of after registration

    ** ๐Ÿ‘Œ Improvement

    • [MESOS-111] - Add SVN ignore and git ignore info to repository
    • [MESOS-728] - Masters should seppuku using EXIT instead of abort() when leadership is lost.
    • [MESOS-756] - Improve release tooling.
    • [MESOS-760] - Capture memory usage statistics before OOM
    • [MESOS-761] - Export all memory stats from memory.stat via CgroupsIsolator's usage()
    • [MESOS-768] - Executor driver stop() should dispatch stop to executor process instead of terminating it
    • [MESOS-802] - Web UI shows no errors when navigation to slave fails
    • [MESOS-806] - Allowing converting from an Owned to a Shared.
    • [MESOS-818] - Bump up the minimum number threads libprocess creates to accommodate new tests
    • [MESOS-833] - The Status Update Manager should use a back-off mechanism for retried updates.
    • [MESOS-835] - Reduce the minimum amount of CPUs required to make offers
    • [MESOS-849] - As a developer I should be able to set the AUTOMAKE and ACLOCAL environment variables for autoconf to pickup when using the bootstrap script.
    • [MESOS-881] - Tests are slow because the scheduler attempts to authenticate before the master realizes it is elected.
    • [MESOS-900] - Paginate all tables in the web UI