Fix TaskGroupQueue will never be wakeup due to wakeup failed at one time #15528

ruanwenjun · 2024-01-28T05:34:06Z

Purpose of the pull request

Fix the Task which using TaskGroup will never be wakeup due to wakeup failed at one time.

Brief change log

In DS, we use TaskGroup to restrict the Task parallelism.
e.g. We have two task TaskA, TaskB, and they belong to the same task group, if we execute these two task at the same time, only one will get the taskGroup slot and the other will wait until be wakeup.

Unfortunately, if the TaskA wakeup TaskB failed, then TaskB will never be wakeup, since we use put mode, the waiting task will only be wakeup by other task, it will not try to pull the slot.

This PR will refactor this.

This PR add a new component TaskGroupCoordinator which use to deal with acquire/release task group slot and notify task.

Once a task need to acquire TaskGroup slot, it should call the TaskGroupCoordinator::acquireTaskGroupSlot, in this method, will insert a in queue TaskGroupQueue which status is WAITING.
Once a task is finished it will call TaskGroupCoordinator::releaseTaskGroupSlot, in this method, will move the TaskGroupQueue out queue.

And TaskGroupCoordinator will loop the slot, if there exist available slot, it will try to wakeup a waiting task.

Verify this pull request

This pull request is code cleanup without any test coverage.

(or)

This pull request is already covered by existing tests, such as (please describe tests).

(or)

This change added tests and can be verified as follows:

(or)

If your pull request contain incompatible change, you should also add it to docs/docs/en/guide/upgrede/incompatible.md

codecov-commenter · 2024-01-28T05:52:46Z

Codecov Report

Attention: 192 lines in your changes are missing coverage. Please review.

Comparison is base (2f66a66) 38.17% compared to head (1cb5335) 38.38%.

❗ Current head 1cb5335 differs from pull request most recent head 3681244. Consider uploading reports for the commit 3681244 to get more accurate results

Files	Patch %	Lines
.../master/runner/taskgroup/TaskGroupCoordinator.java	42.10%	114 Missing and 7 partials ⚠️
.../server/master/runner/WorkflowExecuteRunnable.java	4.00%	22 Missing and 2 partials ⚠️
...aster/rpc/TaskInstanceWakeupOperationFunction.java	0.00%	16 Missing ⚠️
...cheduler/api/service/impl/ExecutorServiceImpl.java	0.00%	11 Missing ⚠️
.../server/master/event/StateEventHandlerManager.java	0.00%	6 Missing ⚠️
...cheduler/dao/repository/impl/TaskGroupDaoImpl.java	61.53%	2 Missing and 3 partials ⚠️
...ler/dao/repository/impl/TaskGroupQueueDaoImpl.java	78.57%	1 Missing and 2 partials ⚠️
...er/api/service/impl/TaskGroupQueueServiceImpl.java	0.00%	1 Missing ⚠️
...heduler/api/service/impl/TaskGroupServiceImpl.java	83.33%	0 Missing and 1 partial ⚠️
...e/dolphinscheduler/server/master/MasterServer.java	0.00%	1 Missing ⚠️
... and 3 more

Additional details and impacted files

@@             Coverage Diff              @@
##                dev   #15528      +/-   ##
============================================
+ Coverage     38.17%   38.38%   +0.20%     
- Complexity     4709     4748      +39     
============================================
  Files          1304     1305       +1     
  Lines         44828    44841      +13     
  Branches       4807     4800       -7     
============================================
+ Hits          17115    17212      +97     
+ Misses        25860    25760     -100     
- Partials       1853     1869      +16

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

...in/java/org/apache/dolphinscheduler/server/master/runner/taskgroup/TaskGroupCoordinator.java

sonarqubecloud · 2024-01-28T07:02:35Z

Please retry analysis of this Pull-Request directly on SonarCloud

...uler-api/src/main/java/org/apache/dolphinscheduler/api/service/impl/ExecutorServiceImpl.java

+        ProcessInstance processInstance = processInstanceDao.queryOptionalById(taskGroupQueue.getProcessId())
+                .orElseThrow(
+                        () -> new ServiceException(Status.PROCESS_INSTANCE_NOT_EXIST, taskGroupQueue.getProcessId()));


zhongjiajie · 2024-01-31T14:58:27Z

...uler-dao/src/main/java/org/apache/dolphinscheduler/dao/repository/impl/TaskGroupDaoImpl.java

+        if (taskGroupId == null) {
+            throw new IllegalArgumentException("taskGroupId cannot be null");
+        }


is there any reason we do not use @NonNull directly for parameter taskGroupId?

@nonnull Cannot write error message. If we want to use a specific error message ,we cannot use @nonnull.

zhongjiajie · 2024-01-31T15:03:02Z

dolphinscheduler-dao/src/main/resources/sql/upgrade/3.3.0_schema/mysql/dolphinscheduler_ddl.sql

+-- create idx_t_ds_task_group_queue_in_queue on t_ds_task_group_queue
+DROP PROCEDURE IF EXISTS create_idx_t_ds_task_group_queue_in_queue;
+delimiter d//
+CREATE PROCEDURE create_idx_t_ds_task_group_queue_in_queue()
+BEGIN
+    DECLARE index_exists INT DEFAULT 0;
+    SELECT COUNT(*) INTO index_exists FROM information_schema.statistics WHERE table_schema = (SELECT DATABASE()) AND  table_name = 't_ds_task_group_queue' AND index_name = 'idx_t_ds_task_group_queue_in_queue';
+    IF index_exists = 0 THEN CREATE INDEX idx_t_ds_task_group_queue_in_queue ON t_ds_task_group_queue(in_queue);
+END IF;
+END;
+d//
+delimiter ;
+CALL create_idx_t_ds_task_group_queue_in_queue;
+DROP PROCEDURE create_idx_t_ds_task_group_queue_in_queue;


I think we could directly remove the related 3.3.0 upgrade. I would merge almost all upgrade sql from 3.3.0 to 3.2.1. and we will not release 3.3.0 until we close or nearly close 3.2.x

Yes， we can remove the upgrade 3.3.0. We can use another PR to do this.

caishunfeng

LGTM

sonarqubecloud · 2024-02-01T05:22:02Z

Quality Gate failed

Failed conditions

42.1% Coverage on New Code (required ≥ 60%)

See analysis details on SonarCloud

Gallardot

Good job! Rest LGTM.

Gallardot · 2024-02-01T07:48:47Z

...in/java/org/apache/dolphinscheduler/server/master/runner/taskgroup/TaskGroupCoordinator.java

+                log.warn("The TaskInstance: {} state: {} finished, will release the TaskGroupQueue: {}",
+                        taskInstance.getName(), taskInstance.getState(), taskGroupQueue);
+                releaseTaskGroupQueueSlot(taskGroupQueue);
+                continue;


Useless code

Gallardot · 2024-02-01T08:08:43Z

...in/java/org/apache/dolphinscheduler/server/master/runner/taskgroup/TaskGroupCoordinator.java

+                    taskGroupQueue.setInQueue(Flag.YES.getCode());
+                    taskGroupQueue.setStatus(TaskGroupQueueStatus.ACQUIRE_SUCCESS);
+                    taskGroupQueue.setUpdateTime(new Date());
+                    taskGroupQueueDao.updateById(taskGroupQueue);


There is some repetitive code in the state change that might be suitable for turning into a function.

Good suggestion, I will use another PR to move these kind of code in the whole project into function.

…ime (#15528) (cherry picked from commit 8974233)

ruanwenjun requested review from SbloodyS, caishunfeng and zhongjiajie as code owners January 28, 2024 05:34

ruanwenjun added bug Something isn't working 3.2.1 labels Jan 28, 2024

github-actions bot added the backend label Jan 28, 2024

github-advanced-security bot found potential problems Jan 28, 2024

View reviewed changes

...in/java/org/apache/dolphinscheduler/server/master/runner/taskgroup/TaskGroupCoordinator.java Fixed Show fixed Hide fixed

ruanwenjun force-pushed the dev_wenjun_fixTaskGroupQueueWillNeverWakeup branch from 38ff493 to 212ba8d Compare January 28, 2024 06:36

github-actions bot added the CI&CD label Jan 28, 2024

ruanwenjun force-pushed the dev_wenjun_fixTaskGroupQueueWillNeverWakeup branch 5 times, most recently from 2d17d56 to 98a8d8f Compare January 28, 2024 14:59

caishunfeng assigned ruanwenjun Jan 29, 2024

caishunfeng added this to the 3.2.1 milestone Jan 29, 2024

ruanwenjun force-pushed the dev_wenjun_fixTaskGroupQueueWillNeverWakeup branch from 98a8d8f to 3cfd45c Compare January 29, 2024 03:50

github-advanced-security bot found potential problems Jan 29, 2024

View reviewed changes

ruanwenjun force-pushed the dev_wenjun_fixTaskGroupQueueWillNeverWakeup branch 10 times, most recently from 59d46b2 to 82c1bce Compare January 30, 2024 09:01

ruanwenjun force-pushed the dev_wenjun_fixTaskGroupQueueWillNeverWakeup branch 2 times, most recently from ec952bc to 628f690 Compare January 30, 2024 10:34

ruanwenjun added major refactor labels Jan 30, 2024

ruanwenjun force-pushed the dev_wenjun_fixTaskGroupQueueWillNeverWakeup branch 2 times, most recently from e36844f to b9e452e Compare January 31, 2024 14:37

zhongjiajie reviewed Jan 31, 2024

View reviewed changes

caishunfeng approved these changes Feb 1, 2024

View reviewed changes

ruanwenjun added the ready-to-merge label Feb 1, 2024

Fix TaskGroupQueue will never be wakeup due to wakeup failed at one time

3681244

ruanwenjun force-pushed the dev_wenjun_fixTaskGroupQueueWillNeverWakeup branch from b9e452e to 3681244 Compare February 1, 2024 04:56

Gallardot reviewed Feb 1, 2024

View reviewed changes

ruanwenjun merged commit 8974233 into apache:dev Feb 1, 2024

ruanwenjun deleted the dev_wenjun_fixTaskGroupQueueWillNeverWakeup branch February 1, 2024 09:14

zhongjiajie pushed a commit that referenced this pull request Feb 6, 2024

Fix TaskGroupQueue will never be wakeup due to wakeup failed at one t…

c8f120a

…ime (#15528) (cherry picked from commit 8974233)

Fix TaskGroupQueue will never be wakeup due to wakeup failed at one time #15528

Fix TaskGroupQueue will never be wakeup due to wakeup failed at one time #15528

Uh oh!

Conversation

ruanwenjun commented Jan 28, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Purpose of the pull request

Brief change log

Verify this pull request

Uh oh!

codecov-commenter commented Jan 28, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Uh oh!

sonarqubecloud bot commented Jan 28, 2024

Uh oh!

Check notice

zhongjiajie Jan 31, 2024

Choose a reason for hiding this comment

Uh oh!

ruanwenjun Feb 1, 2024

Choose a reason for hiding this comment

Uh oh!

zhongjiajie Jan 31, 2024

Choose a reason for hiding this comment

Uh oh!

ruanwenjun Feb 1, 2024

Choose a reason for hiding this comment

Uh oh!

caishunfeng left a comment

Choose a reason for hiding this comment

Uh oh!

sonarqubecloud bot commented Feb 1, 2024

Quality Gate failed

Uh oh!

Gallardot left a comment

Choose a reason for hiding this comment

Uh oh!

Gallardot Feb 1, 2024

Choose a reason for hiding this comment

Uh oh!

Gallardot Feb 1, 2024

Choose a reason for hiding this comment

Uh oh!

ruanwenjun Feb 1, 2024

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

ruanwenjun commented Jan 28, 2024 •

edited

Loading

codecov-commenter commented Jan 28, 2024 •

edited

Loading