[fix-#13956] [Master]taskId is null cause NPE

[fuchanghai]

Purpose of the pull request

this close

• [Bug] [master] java.lang.NullPointerException: null #13956
• [Bug] [dolphinscheduler master] Infinite loop process_instance when status is prepare to stop #12440
• [Bug] [Module Name] Master server logs constantly flushing empty pointer logs #13942

Brief change log

This judgment seems to be on the surface if taskInstacne is null, only to enter,
after task insert failed , id is null

I remember that id used to be of type int, but it was later changed to type Integer. My personal opinion is that the above reason caused the exception.

Verify this pull request

[fuchanghai]

@caishunfeng @SbloodyS @ruanwenjun PTAL

[caishunfeng]

LGTM

[codecov-commenter]

Codecov Report

Merging #13980 (e87d7e4) into dev (26c5f3c) will increase coverage by 0.01%.
The diff coverage is 0.00%.

❗ Current head e87d7e4 differs from pull request most recent head ef81911. Consider uploading reports for the commit ef81911 to get more accurate results

@@             Coverage Diff              @@
##                dev   #13980      +/-   ##
============================================
+ Coverage     38.91%   38.93%   +0.01%     
- Complexity     4469     4472       +3     
============================================
  Files          1164     1164              
  Lines         42444    42445       +1     
  Branches       4761     4761              
============================================
+ Hits          16518    16526       +8     
+ Misses        24106    24096      -10     
- Partials       1820     1823       +3     

Impacted Files	Coverage Δ
.../server/master/runner/WorkflowExecuteRunnable.java	9.76% <0.00%> (-0.01%)	⬇️

... and 4 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[sonarcloud]

Kudos, SonarCloud Quality Gate passed!   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
0 Code Smells

0.0% Coverage
0.0% Duplication

[zhuangchong]

The taskId cannot be set to 0 here, because the task instance will be submitted and written to the database before this, thereby generating the id.
https://github.com/fuchanghai/dolphinscheduler/blob/ef81911fe55f19faed00a4e908c0ce1c3d7e39ef/dolphinscheduler-service/src/main/java/org/apache/dolphinscheduler/service/process/ProcessServiceImpl.java#L1159-L1165

[zuzuviewer]

I think bug is from here, created new task instance but not insert to db and task instance id is null

[zhuangchong]

@zuzuviewer Yes, the specific reason is that the task instance creation failed because the obtained resource data is empty.

[zuzuviewer]

@zuzuviewer Yes, the specific reason is that the task instance creation failed because the obtained resource data is empty.

@zhuangchong what resource data

[zuzuviewer]

How about this fix

[HomminLee]

@fuchanghai
I reproduce this BUG, according to #12440 (comment). I'm sure this cause by resource id。

After debugger, I found the problem lies in three steps：

1. The resource files referenced by offline workflow could be deleted in Resource center, but the id of deleted resource files still be used when the workflow is edited and saved.
   

2. Before submitTaskInstanceToDB, will update resource info in task instance with no check exists, so throw NPE.
   

3. submitTaskExec will catch the exception, judge this is due to submit failed. Code go on, but task instance has not save to BD, so there are loop NPE logs.
   

[zhuangchong]

Version 3.1.6 will fix this issue.
#13878

[HomminLee]

Version 3.1.6 will fix this issue. #13878

Does the UI also need to be fixed?

[zhuangchong]

Version 3.1.6 will fix this issue. #13878

Does the UI also need to be fixed?

The ui part has not been fixed yet, but it has little impact. It would be even better if some front-end students are willing to fix it.