apache/druid · error · IllegalStateException

trying to add taskGroup with id

Error message

trying to add taskGroup with id [%s] to actively reading task groups, but group already exists.

What it means

addTaskGroup (createTaskGroup) inserts a new TaskGroup into activelyReadingTaskGroups with putIfAbsent; a non-null return means a group with that id already exists. Since task group ids derive from partition-group versions, this signals state corruption or a race where two threads/paths tried to create the same group, so an ISE is thrown.

Solutions

Restart the supervisor to rebuild task group state cleanly
Check logs for prior killTaskGroupForPartitions failures; ensure groups are removed before re-creation
Upgrade Druid if hitting a known race-condition bug in group creation

Example fix

null

Defensive patterns

Strategy: try-catch

Validate before calling

if (activelyReadingTaskGroups.containsKey(taskGroupId)) {
  return activelyReadingTaskGroups.get(taskGroupId); // reuse instead of re-creating
}

Try / catch

try {
  group = supervisor.createTaskGroup(taskGroupId, partitions, tasks);
} catch (IllegalStateException e) {
  if (e.getMessage().contains("group already exists")) {
    group = supervisor.getActivelyReadingTaskGroups().get(taskGroupId);
  } else throw e;
}

Prevention

Ensure task groups are removed (killTaskGroupForPartitions) before re-creation with the same id
Avoid running multiple supervisor instances for the same datasource
Upgrade Druid if hitting known group-creation race bugs

When it happens

Trigger: Two concurrent calls creating the same taskGroupId (same partition allocation and same partitionGroups version), or re-creating a group that was never removed from activelyReadingTaskGroups (e.g. after killTaskGroupForPartitions failed to run).

Common situations: Supervisor race conditions during task group recycling; version not incremented after killing groups; supervisor state partially restored after restart.

Understand the failure class

Background: "This is a bug, please report it": internal invariant violations, unreachable panics, and SNH errors explained — this error's family across 47 libraries.

Related errors

AI-assisted analysis of apache/druid@9b90983fd2 (2026-09-07). Data as JSON: /api/errors/a2675c0bbde66590. Report an issue: GitHub.

Appendix: source

Thrown at indexing-service/src/main/java/org/apache/druid/indexing/seekablestream/supervisor/SeekableStreamSupervisor.java:1972

      Set<String> tasks,
      Set<PartitionIdType> exclusiveStartingSequencePartitions,
      @Nullable Map<String, Integer> taskIdToServerPriority
  )
  {
    TaskGroup group = new TaskGroup(
        taskGroupId,
        partitionOffsets,
        null,
        minMsgTime,
        maxMsgTime,
        exclusiveStartingSequencePartitions
    );
    group.tasks.putAll(tasks.stream().collect(Collectors.toMap(x -> x, x -> new TaskData())));
    if (taskIdToServerPriority != null) {
      group.taskIdToServerPriority.putAll(taskIdToServerPriority);
    }
    if (activelyReadingTaskGroups.putIfAbsent(taskGroupId, group) != null) {
      throw new ISE(
          "trying to add taskGroup with id [%s] to actively reading task groups, but group already exists.",
          taskGroupId
      );
    }
    return group;
  }

  @VisibleForTesting
  public TaskGroup addTaskGroupToPendingCompletionTaskGroup(
      int taskGroupId,
      ImmutableMap<PartitionIdType, SequenceOffsetType> partitionOffsets,
      @Nullable DateTime minMsgTime,
      @Nullable DateTime maxMsgTime,
      Set<String> tasks,
      Set<PartitionIdType> exclusiveStartingSequencePartitions,
      @Nullable Map<String, Integer> taskIdToServerPriority
  )
  {

View on GitHub (pinned to 9b90983fd2)