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

lidongdai pushed a commit to branch dev
in repository https://gitbox.apache.org/repos/asf/dolphinscheduler.git


The following commit(s) were added to refs/heads/dev by this push:
     new 7a0a2c2a46 Solve the deadlock problem caused by queuing (#13191)
7a0a2c2a46 is described below

commit 7a0a2c2a46a0224d0c9fb1f649c1f901d11e814b
Author: sssqhai <[email protected]>
AuthorDate: Fri Dec 16 19:55:02 2022 +0800

    Solve the deadlock problem caused by queuing (#13191)
    
    * Solve the deadlock problem caused by queuing
    
    * Solve the deadlock problem caused by queuing
    
    * Solve the deadlock problem caused by queuing
    
    * Solve the deadlock problem caused by queuing,move the event to the tail 
by throwing a exception
    
    Co-authored-by: wfs <[email protected]>
---
 ...ntHandler.java => StateEventHandleFailure.java} | 25 ++++++++++------------
 .../server/master/event/StateEventHandler.java     |  3 ++-
 .../event/TaskWaitTaskGroupStateHandler.java       |  8 +++++--
 .../master/runner/WorkflowExecuteRunnable.java     |  8 +++++++
 4 files changed, 27 insertions(+), 17 deletions(-)

diff --git 
a/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/StateEventHandler.java
 
b/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/StateEventHandleFailure.java
similarity index 53%
copy from 
dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/StateEventHandler.java
copy to 
dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/StateEventHandleFailure.java
index 93d4a7d4b3..5e757c7858 100644
--- 
a/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/StateEventHandler.java
+++ 
b/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/StateEventHandleFailure.java
@@ -17,20 +17,17 @@
 
 package org.apache.dolphinscheduler.server.master.event;
 
-import org.apache.dolphinscheduler.common.enums.StateEventType;
-import 
org.apache.dolphinscheduler.server.master.runner.WorkflowExecuteRunnable;
-
-public interface StateEventHandler {
+/**
+ * This exception represent the exception can be recovered, when we get this 
exception,
+ * we will move the event to the fail of the queue.
+ */
+public class StateEventHandleFailure extends Exception {
 
-    /**
-     * Handle a event, if handle success will reture true, else return false
-     *
-     * @param stateEvent given state event.
-     * @throws StateEventHandleException this exception means it can be 
recovered.
-     * @throws StateEventHandleError     this exception means it cannot be 
recovered, so the event need to drop.
-     */
-    boolean handleStateEvent(WorkflowExecuteRunnable workflowExecuteRunnable,
-                             StateEvent stateEvent) throws 
StateEventHandleException, StateEventHandleError;
+    public StateEventHandleFailure(String message) {
+        super(message);
+    }
 
-    StateEventType getEventType();
+    public StateEventHandleFailure(String message, Throwable throwable) {
+        super(message, throwable);
+    }
 }
diff --git 
a/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/StateEventHandler.java
 
b/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/StateEventHandler.java
index 93d4a7d4b3..377ea71f62 100644
--- 
a/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/StateEventHandler.java
+++ 
b/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/StateEventHandler.java
@@ -28,9 +28,10 @@ public interface StateEventHandler {
      * @param stateEvent given state event.
      * @throws StateEventHandleException this exception means it can be 
recovered.
      * @throws StateEventHandleError     this exception means it cannot be 
recovered, so the event need to drop.
+     * @throws StateEventHandleException this means it can be recovered.
      */
     boolean handleStateEvent(WorkflowExecuteRunnable workflowExecuteRunnable,
-                             StateEvent stateEvent) throws 
StateEventHandleException, StateEventHandleError;
+                             StateEvent stateEvent) throws 
StateEventHandleException, StateEventHandleError, StateEventHandleFailure;
 
     StateEventType getEventType();
 }
diff --git 
a/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/TaskWaitTaskGroupStateHandler.java
 
b/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/TaskWaitTaskGroupStateHandler.java
index 02d08a6616..bb0c7b8b70 100644
--- 
a/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/TaskWaitTaskGroupStateHandler.java
+++ 
b/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/event/TaskWaitTaskGroupStateHandler.java
@@ -31,9 +31,13 @@ public class TaskWaitTaskGroupStateHandler implements 
StateEventHandler {
     private static final Logger logger = 
LoggerFactory.getLogger(TaskWaitTaskGroupStateHandler.class);
 
     @Override
-    public boolean handleStateEvent(WorkflowExecuteRunnable 
workflowExecuteRunnable, StateEvent stateEvent) {
+    public boolean handleStateEvent(WorkflowExecuteRunnable 
workflowExecuteRunnable,
+                                    StateEvent stateEvent) throws 
StateEventHandleFailure {
         logger.info("Handle task instance wait task group event, 
taskInstanceId: {}", stateEvent.getTaskInstanceId());
-        return workflowExecuteRunnable.checkForceStartAndWakeUp(stateEvent);
+        if (!workflowExecuteRunnable.checkForceStartAndWakeUp(stateEvent)) {
+            throw new StateEventHandleFailure("Task state event handle failed 
due to robing taskGroup resource failed");
+        }
+        return true;
     }
 
     @Override
diff --git 
a/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/runner/WorkflowExecuteRunnable.java
 
b/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/runner/WorkflowExecuteRunnable.java
index efb4a944c7..ab0af2be5b 100644
--- 
a/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/runner/WorkflowExecuteRunnable.java
+++ 
b/dolphinscheduler-master/src/main/java/org/apache/dolphinscheduler/server/master/runner/WorkflowExecuteRunnable.java
@@ -71,6 +71,7 @@ import 
org.apache.dolphinscheduler.server.master.dispatch.executor.NettyExecutor
 import org.apache.dolphinscheduler.server.master.event.StateEvent;
 import org.apache.dolphinscheduler.server.master.event.StateEventHandleError;
 import 
org.apache.dolphinscheduler.server.master.event.StateEventHandleException;
+import org.apache.dolphinscheduler.server.master.event.StateEventHandleFailure;
 import org.apache.dolphinscheduler.server.master.event.StateEventHandler;
 import 
org.apache.dolphinscheduler.server.master.event.StateEventHandlerManager;
 import org.apache.dolphinscheduler.server.master.event.TaskStateEvent;
@@ -315,6 +316,13 @@ public class WorkflowExecuteRunnable implements 
Callable<WorkflowSubmitStatue> {
                         stateEvent,
                         stateEventHandleException);
                 ThreadUtils.sleep(Constants.SLEEP_TIME_MILLIS);
+            } catch (StateEventHandleFailure stateEventHandleFailure) {
+                logger.error("State event handle failed, will move event to 
the tail: {}",
+                        stateEvent,
+                        stateEventHandleFailure);
+                this.stateEvents.remove(stateEvent);
+                this.stateEvents.offer(stateEvent);
+                ThreadUtils.sleep(Constants.SLEEP_TIME_MILLIS);
             } catch (Exception e) {
                 // we catch the exception here, since if the state event 
handle failed, the state event will still keep
                 // in the stateEvents queue.

Reply via email to