HTTP task stuck in SCHEDULED state

[anjkl]

Describe the bug
HTTP task get stuck in SCHEDULED state when external.payload.storage not configured and payload size is in between Soft Barrier and Hard Barrier.

logs related to the issue:

2023-08-04 08:39:56 DEBUG AsyncSystemTaskExecutor: - Task: TaskModel{taskType='HTTP', status=SCHEDULED, inputData={asyncComplete=false, http_request={method=GET, uri=http://server:8080/api/test, accept=application/json}}, referenceTaskName='get_test_info', retryCount=2, seq=3, correlationId='10234113-4020-4a89-b4fd-a63ae943b203', pollCount=0, taskDefName='get_test_info', scheduledTime=1691138396532, startTime=0, endTime=0, updateTime=1691138396485, startDelayInSeconds=0, retriedTaskId='48c8d4bb-001a-4c4b-8248-2738f9a11e18', retried=false, executed=false, callbackFromWorker=true, responseTimeoutSeconds=0, workflowInstanceId='11e5b8a8-eaff-4fed-9854-db1b2e702044', workflowType='generic_workflow', taskId='d2fe143a-fc8b-49ba-977c-3d5172dfdf9b', reasonForIncompletion='null', callbackAfterSeconds=0, workerId='null', outputData={}, workflowTask=get_test_info/get_test_info, domain='null', waitTimeout='0', inputMessage=null, outputMessage=null, rateLimitPerFrequency=0, rateLimitFrequencyInSeconds=0, externalInputPayloadStoragePath='null', externalOutputPayloadStoragePath='null', workflowPriority=0, executionNameSpace='null', isolationGroupId='null', iteration=0, subWorkflowId='null', subworkflowChanged=false} fetched from execution DAO for taskId: d2fe143a-fc8b-49ba-977c-3d5172dfdf9b
2023-08-04 08:39:56 DEBUG AsyncSystemTaskExecutor: - Executing HTTP/d2fe143a-fc8b-49ba-977c-3d5172dfdf9b in SCHEDULED state
...
...
2023-08-04 08:39:58 DEBUG AsyncSystemTaskExecutor: - Finished execution of HTTP/d2fe143a-fc8b-49ba-977c-3d5172dfdf9b-COMPLETED
2023-08-04 08:39:58 DEBUG SystemTaskWorker: - Polling queue:START_WORKFLOW, got 0 tasks
2023-08-04 08:39:59 ERROR ExternalPayloadStorageUtils: - Unable to upload payload to external storage for workflow: 11e5b8a8-eaff-4fed-9854-db1b2e702044
java.lang.NullPointerException: null
        at com.netflix.conductor.core.utils.ExternalPayloadStorageUtils.uploadHelper(ExternalPayloadStorageUtils.java:210) ~[conductor-core-3.13.7.jar!/:3.13.7]
        at com.netflix.conductor.core.utils.ExternalPayloadStorageUtils.verifyAndUpload(ExternalPayloadStorageUtils.java:164) ~[conductor-core-3.13.7.jar!/:3.13.7]
        at com.netflix.conductor.core.dal.ExecutionDAOFacade.externalizeTaskData(ExecutionDAOFacade.java:271) ~[conductor-core-3.13.7.jar!/:3.13.7]
        at com.netflix.conductor.core.dal.ExecutionDAOFacade.updateTask(ExecutionDAOFacade.java:504) ~[conductor-core-3.13.7.jar!/:3.13.7]
        at com.netflix.conductor.core.execution.AsyncSystemTaskExecutor.execute(AsyncSystemTaskExecutor.java:190) ~[conductor-core-3.13.7.jar!/:3.13.7]
        at com.netflix.conductor.core.execution.tasks.SystemTaskWorker.lambda$pollAndExecute$1(SystemTaskWorker.java:135) ~[conductor-core-3.13.7.jar!/:3.13.7]
        at java.util.concurrent.CompletableFuture$AsyncRun.run(CompletableFuture.java:1736) ~[?:?]
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128) ~[?:?]
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628) ~[?:?]
        at java.lang.Thread.run(Thread.java:829) ~[?:?]
2023-08-04 08:39:59 DEBUG SystemTaskWorker: - Polling queue: START_WORKFLOW with 4 slots acquired
2023-08-04 08:39:59 DEBUG SystemTaskWorker: - Polling queue:START_WORKFLOW, got 0 tasks
2023-08-04 08:39:59 DEBUG SystemTaskWorker: - Polling queue: KAFKA_PUBLISH with 4 slots acquired
2023-08-04 08:39:59 DEBUG SystemTaskWorker: - Polling queue:KAFKA_PUBLISH, got 0 tasks
...

Details
Conductor version: 3.13.7
Persistence implementation: MySQ
Queue implementation: MySQL
Lock: no

To Reproduce
Steps to reproduce the behavior:

• Create some http server that generates a big response. For example

cd $(mktemp -d); python3 -c 'print("a" * 5 * 1024 * 1024)' > 1; python3 -m http.server 8888

• Create a simple workflow with a single HTTP task

{
  "accessPolicy": {},
  "name": "stuck_http_test",
  "description": "test",
  "version": 1,
  "tasks": [
    {
      "name": "get_test_info",
      "taskReferenceName": "get_test_info",
      "inputParameters": {
        "http_request": {
          "uri": "http://localhost:8888/1",
          "method": "GET",
          "accept": "application/json"
        },
        "asyncComplete": false
      },
      "type": "HTTP",
      "startDelay": 0,
      "optional": true,
      "asyncComplete": false
    }
  ],
  "inputParameters": [],
  "outputParameters": {},
  "schemaVersion": 2,
  "restartable": true,
  "workflowStatusListenerEnabled": false,
  "ownerEmail": "example@email.com",
  "timeoutPolicy": "ALERT_ONLY",
  "timeoutSeconds": 0,
  "variables": {},
  "inputTemplate": {}
}

• Run workflow

curl 'http://localhost:8080/api/workflow/stuck_http_test' -X POST -H 'Content-Type: application/json' --data-raw '{}'

• Check workflow status and conductor logs (task stuck, you'll see NullPointerException for ExternalPayloadStorageUtils in conductor logs)

Expected behavior
HTTP task failed

Probably related to #1153

[anjkl]

Few additional notes. The similar issue can be reproduced by other kind of tasks like INLINE task.

In case of INLINE task it may different results but the root cause is the same (NullPointerException while uploading payload to external storage):

• task get stuck in IN_PROGRESS state
• workflow start fails with error:

{
  "status": 500,
  "message": "Unable to upload payload to external storage for workflow: 2031725a-b1ff-4794-828f-7f7de8703b84",
  "instance": "laptop",
  "retryable": false
}

Example of INLINE tasks that reproduce the same bug but with different results:

• workflow start fails with 500 error

{
  "accessPolicy": {},
  "name": "test_workflow_execution_error",
  "version": 1,
  "tasks": [
    {
      "name": "test",
      "taskReferenceName": "test",
      "inputParameters": {
        "evaluatorType": "javascript",
        "expression": "'a'.repeat(${workflow.input.repeats});"
      },
      "type": "INLINE",
      "startDelay": 0,
      "optional": true,
      "asyncComplete": false
    }
  ],
  "inputParameters": [
    "repeats"
  ],
  "outputParameters": {},
  "schemaVersion": 2,
  "ownerEmail": "example@email.com",
  "timeoutPolicy": "ALERT_ONLY",
  "timeoutSeconds": 0
}

• task get stuck in IN_PROGRESS state

{
  "accessPolicy": {},
  "name": "test_inline_task_stuck_in_progress",
  "version": 1,
  "tasks": [
    {
      "name": "wait",
      "taskReferenceName": "wait",
      "inputParameters": {
        "duration": "30s"
      },
      "type": "WAIT",
      "startDelay": 0,
      "optional": false,
      "asyncComplete": false
    },
    {
      "name": "test",
      "taskReferenceName": "test",
      "inputParameters": {
        "evaluatorType": "javascript",
        "expression": "'a'.repeat(${workflow.input.repeats});"
      },
      "type": "INLINE",
      "startDelay": 0,
      "optional": true,
      "asyncComplete": false
    }
  ],
  "inputParameters": [
    "repeats"
  ],
  "outputParameters": {},
  "schemaVersion": 2,
  "ownerEmail": "example@email.com",
  "timeoutPolicy": "ALERT_ONLY",
  "timeoutSeconds": 0
}

To trigger workflow with the issue use input:

{
  "repeats": 7096000
}

[anjkl]

And workaround is setting the barriers to the same value:

conductor:
  app:
    workflowInputPayloadSizeThreshold: 5120
    maxWorkflowInputPayloadSizeThreshold: 5120
    workflowOutputPayloadSizeThreshold: 5120
    maxWorkflowOutputPayloadSizeThreshold: 5120
    taskInputPayloadSizeThreshold: 3072
    maxTaskInputPayloadSizeThreshold: 3072
    taskOutputPayloadSizeThreshold: 3072
    maxTaskOutputPayloadSizeThreshold: 3072

[saksham2105]

This Commit should fix this, I think this should work now, You can let us know if it doesn't work for you

[anjkl]

Thanks a lot for the fix.

Just a note: it may lead to unexpected behavior in case of running few instances of conductor with DummyPayloadStorage. So it shouldn't be run on production. But probably it's OK for dummy storage.

[saksham2105]

Yeah We have a test env property which is creating conditional bean for DummyPayloadStorage, for production we have different storage classes like S3 etc.