apache/druid · error · IllegalStateException

Result partition information is not ready yet

Error message

Result partition information is not ready yet

What it means

ControllerStageTracker.getResultPartitions returns the ReadablePartitions for a stage's results, but until the stage finishes (or its partition count is known) the internal resultPartitions field is null. Callers that read partitions before the stage has reached a phase where partitions are known trigger this ISE — a guard against reading controller state too early.

Source

Thrown at multi-stage-query/src/main/java/org/apache/druid/msq/kernel/controller/ControllerStageTracker.java:241

  {
    return phase;
  }

  /**
   * Whether partitions for the results of this stage have been set.
   */
  boolean hasResultPartitions()
  {
    return resultPartitions != null;
  }

  /**
   * Partitions for the results of the stage associated with this tracker.
   */
  ReadablePartitions getResultPartitions()
  {
    if (resultPartitions == null) {
      throw new ISE("Result partition information is not ready yet");
    } else {
      return resultPartitions;
    }
  }

  /**
   * @return Partition boundaries for the results of this stage
   */
  ClusterByPartitions getResultPartitionBoundaries()
  {
    if (!getStageDefinition().doesShuffle()) {
      throw new ISE("Result partition information is not relevant to this stage because it does not shuffle");
    } else if (resultPartitionBoundaries == null) {
      throw new ISE("Result partition information is not ready yet");
    } else {
      return resultPartitionBoundaries;
    }
  }

View on GitHub (pinned to 9b90983fd2)

Solutions

1. Check the stage phase before calling: only read partitions once the tracker reports the stage finished (or use a method that returns Optional/null for unknown partitions).
2. Use ControllerQueryKernel's phase-based APIs (e.g. checking resultPrefix/readyForReading) instead of direct partition access during execution.
3. If hit intermittently in logs, add a guard: if (tracker.getResultPartitionsIfKnown() == null) skip and re-check on next controller tick.

Example fix

// before
ReadablePartitions p = tracker.getResultPartitions();

// after
if (kernel.getStagePhase(stageId) == StagePhase.FINISHED) {
  ReadablePartitions p = tracker.getResultPartitions();
}

if (kernel.getStagePhase(stageId) != StagePhase.FINISHED) {
  // partitions not knowable yet — defer the read
  return;
}

try {
  ReadablePartitions p = tracker.getResultPartitions();
} catch (IllegalStateException e) {
  if (e.getMessage().contains("not ready yet")) {
    // re-check on next controller tick
  } else {
    throw e;
  }
}

When it happens

Trigger: Calling getResultPartitions() while the tracked stage is still running (e.g. during READING_INPUT/POST_COMPUTATION phase) before resultPartitions is set by finish()/state advancement; also after controller restart if partition info was not persisted.

Common situations: Controller-side code that checks downstream readiness and assumes partitions are known as soon as a stage starts; custom controller extensions or debugging code polling result partitions mid-stage; race between stage completion callback and partition query.

Understand the failure class

Background: "Invalid state transition" errors: "status must be X, actually Y", "already rejected/charging/uninstalled", "cannot ... while running" — what they mean when a library rejects your call — this error's family across 31 libraries.

Related errors

• Work order for worker[%d] not found for stage[%d]
• Number of partitions not known for [%s] with maxPartitions[%
• Cannot accept duplicate stage numbers
• Work order not present for stage[%s]
• Cannot start the stage: [%s]

AI-assisted analysis of apache/druid@9b90983fd2 (2026-09-07). Data as JSON: /api/errors/eba7a036b7d2ae00. Report an issue: GitHub.