[4/4] mesos git commit: Terminated driver-based executors if kill arrives before launch task.

[al...@apache.org]

Terminated driver-based executors if kill arrives before launch task.

`ExecutorRegisteredMessage` or `RunTaskMessage` may not be delivered
to a driver-based executor. Since these messages are not retried,
without this patch an executor never starts a task and remains idle,
ignoring kill task request. This patch ensures all built-in driver-
based executors eventually shut down if kill task arrives before
the task has been started.

Review: https://reviews.apache.org/r/64033/


Project: http://git-wip-us.apache.org/repos/asf/mesos/repo
Commit: http://git-wip-us.apache.org/repos/asf/mesos/commit/769108e9
Tree: http://git-wip-us.apache.org/repos/asf/mesos/tree/769108e9
Diff: http://git-wip-us.apache.org/repos/asf/mesos/diff/769108e9

Branch: refs/heads/master
Commit: 769108e94a7c7834c44e01091a9940354eb3f6e4
Parents: b2eddcf
Author: Alexander Rukletsov <ru...@gmail.com>
Authored: Fri Dec 22 12:10:35 2017 +0100
Committer: Alexander Rukletsov <al...@apache.org>
Committed: Fri Dec 22 12:10:35 2017 +0100

----------------------------------------------------------------------
 src/docker/executor.cpp   |  6 ++++++
 src/exec/exec.cpp         | 11 +++++++++++
 src/launcher/executor.cpp |  6 ++++++
 3 files changed, 23 insertions(+)
----------------------------------------------------------------------


http://git-wip-us.apache.org/repos/asf/mesos/blob/769108e9/src/docker/executor.cpp
----------------------------------------------------------------------
diff --git a/src/docker/executor.cpp b/src/docker/executor.cpp
index 3974f20..e4c53d5 100644
--- a/src/docker/executor.cpp
+++ b/src/docker/executor.cpp
@@ -383,6 +383,12 @@ private:
       return;
     }
 
+    // Terminate if a kill task request is received before the task is launched.
+    // This can happen, for example, if `RunTaskMessage` has not been delivered.
+    // See MESOS-8297.
+    CHECK_SOME(run) << "Terminating because kill task message has been"
+                    << " received before the task has been launched";
+
     // TODO(alexr): If a kill is in progress, consider adjusting
     // the grace period if a new one is provided.
 

http://git-wip-us.apache.org/repos/asf/mesos/blob/769108e9/src/exec/exec.cpp
----------------------------------------------------------------------
diff --git a/src/exec/exec.cpp b/src/exec/exec.cpp
index 07163a5..0c76a3f 100644
--- a/src/exec/exec.cpp
+++ b/src/exec/exec.cpp
@@ -347,6 +347,17 @@ protected:
       return;
     }
 
+    // A kill task request is received when the driver is not connected. This
+    // can happen, for example, if `ExecutorRegisteredMessage` has not been
+    // delivered. We do not shutdown the driver because there might be other
+    // still running tasks and the executor might eventually reconnect, e.g.,
+    // after the agent failover. We do not drop ignore the message because the
+    // actual executor may still want to react, e.g., commit suicide.
+    if (!connected) {
+      LOG(WARNING) << "Executor received kill task message for task " << taskId
+                   << " while disconnected from the agent!";
+    }
+
     VLOG(1) << "Executor asked to kill task '" << taskId << "'";
 
     Stopwatch stopwatch;

http://git-wip-us.apache.org/repos/asf/mesos/blob/769108e9/src/launcher/executor.cpp
----------------------------------------------------------------------
diff --git a/src/launcher/executor.cpp b/src/launcher/executor.cpp
index dd76c9f..794bf7c 100644
--- a/src/launcher/executor.cpp
+++ b/src/launcher/executor.cpp
@@ -777,6 +777,12 @@ private:
       return;
     }
 
+    // Terminate if a kill task request is received before the task is launched.
+    // This can happen, for example, if `RunTaskMessage` has not been delivered.
+    // See MESOS-8297.
+    CHECK(launched) << "Terminating because kill task message has been"
+                    << " received before the task has been launched";
+
     // If the task is being killed but has not terminated yet and
     // we receive another kill request. Check if we need to adjust
     // the remaining grace period.