[tez] branch master updated: TEZ-4230: LocalContainerLauncher can kill task future too early, causing app hang (László Bodor reviewed by Jonathan Turner Eagles)

[ab...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

abstractdog pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/tez.git


The following commit(s) were added to refs/heads/master by this push:
     new 1fba8f7  TEZ-4230: LocalContainerLauncher can kill task future too early, causing app hang (László Bodor reviewed by Jonathan Turner Eagles)
1fba8f7 is described below

commit 1fba8f71a0f496e1fcb67a223fb32b389bdfb2e0
Author: László Bodor <bo...@gmail.com>
AuthorDate: Tue Sep 8 18:43:23 2020 +0200

    TEZ-4230: LocalContainerLauncher can kill task future too early, causing app hang (László Bodor reviewed by Jonathan Turner Eagles)
    
    Signed-off-by: Laszlo Bodor <bo...@gmail.com>
---
 .../org/apache/tez/dag/app/launcher/LocalContainerLauncher.java  | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/tez-dag/src/main/java/org/apache/tez/dag/app/launcher/LocalContainerLauncher.java b/tez-dag/src/main/java/org/apache/tez/dag/app/launcher/LocalContainerLauncher.java
index a63ce3f..ae30809 100644
--- a/tez-dag/src/main/java/org/apache/tez/dag/app/launcher/LocalContainerLauncher.java
+++ b/tez-dag/src/main/java/org/apache/tez/dag/app/launcher/LocalContainerLauncher.java
@@ -299,10 +299,11 @@ public class LocalContainerLauncher extends DagContainerLauncher {
     if (future == null) {
       LOG.info("Ignoring stop request for containerId: " + event.getContainerId());
     } else {
-      LOG.info(
-          "Stopping containerId: {}",
-          event.getContainerId());
-      future.cancel(true);
+      LOG.info("Stopping containerId: {}, isDone: {}", event.getContainerId(),
+          future.isDone());
+      future.cancel(false);
+      LOG.debug("Stopped containerId: {}, isCancelled: {}", event.getContainerId(),
+          future.isCancelled());
     }
     // Send this event to maintain regular control flow. This isn't of much use though.
     getContext().containerStopRequested(event.getContainerId());