What causes "getShuffleFinishTime() not supported for MapTask" in apache/hadoop?

Branch on task type first: status.getIsMap() (or instanceof ReduceTaskStatus) before calling shuffle APIs

apache/hadoop · error · UnsupportedOperationException

getShuffleFinishTime() not supported for MapTask

Error message

getShuffleFinishTime() not supported for MapTask

What it means

MapTaskStatus models only map-phase timing (mapFinishTime); shuffle and sort finish times exist solely on ReduceTaskStatus because maps never shuffle input. Calling getShuffleFinishTime() on a map task's status throws UnsupportedOperationException by design.

Source

Thrown at hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapred/MapTaskStatus.java:59

    return true;
  }

  /**
   * Sets finishTime. 
   * @param finishTime finish time of task.
   */
  @Override
  void setFinishTime(long finishTime) {
    super.setFinishTime(finishTime);
    // set mapFinishTime if it hasn't been set before
    if (getMapFinishTime() == 0) {
      setMapFinishTime(finishTime);
    }
  }
  
  @Override
  public long getShuffleFinishTime() {
    throw new UnsupportedOperationException("getShuffleFinishTime() not supported for MapTask");
  }

  @Override
  void setShuffleFinishTime(long shuffleFinishTime) {
    throw new UnsupportedOperationException("setShuffleFinishTime() not supported for MapTask");
  }

  @Override
  public long getMapFinishTime() {
    return mapFinishTime;
  }
  
  @Override
  void setMapFinishTime(long mapFinishTime) {
    this.mapFinishTime = mapFinishTime;
  }
  
  @Override

View on GitHub (pinned to 2add963021)

Solutions

Branch on task type first: status.getIsMap() (or instanceof ReduceTaskStatus) before calling shuffle APIs
Use getMapFinishTime() for maps and getShuffleFinishTime()/getSortFinishTime() only for reduces
Teach serializers to skip reduce-only getters for map tasks (view/annotation-based exclusion)

Example fix

// before
long phaseEnd = status.getShuffleFinishTime(); // throws for map attempts

// after
long phaseEnd = status.getIsMap()
    ? status.getMapFinishTime()
    : status.getShuffleFinishTime();

Defensive patterns

Strategy: type-guard

Type guard

static boolean hasShuffleFinishTime(TaskStatus s) {
  return s != null && !s.getIsMap(); // equivalently: s instanceof ReduceTaskStatus
}

Try / catch

if (hasShuffleFinishTime(status)) { shuffleEnd = status.getShuffleFinishTime(); } else { mapEnd = status.getMapFinishTime(); } // guard first; only catch UnsupportedOperationException as a last-resort signal of a type bug

Prevention

Model task phases by task type: map -> mapFinish, reduce -> shuffle/sort/finish
Never call reduce-only TaskStatus methods from generic code paths
Write one adapter (getPhaseFinishTime(TaskStatus)) and use it everywhere

When it happens

Trigger: Generic code iterating task reports calls status.getShuffleFinishTime() without checking task type: monitoring/audit dashboards, JSON/JMX serializers that walk all getters, or code ported from paths that only ever saw ReduceTaskStatus.

Common situations: Custom job-history or monitoring tooling reading timings from every task; reflection-based serializers touching reduce-only getters; libraries shared between map and reduce handling paths.

Related errors

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/9b5a4c909e350a94. Report an issue: GitHub.