apache/pulsar · error · java.lang.UnsupportedOperationException

Another partition exists for [${topicName}].

Error message

Another partition exists for [${topicName}].

What it means

When deleting a partitioned topic, the broker deletes each partition first and then removes the partitioned-topic metadata. Before deleting metadata it checks whether any partition still exists; if at least one partition check returns true, deletion is aborted with UnsupportedOperationException because removing the parent metadata while partitions remain would orphan them. The error protects the invariants of partitioned-topic bookkeeping in metadata storage.

Source

Thrown at pulsar-broker/src/main/java/org/apache/pulsar/broker/service/persistent/PersistentTopic.java:3836

                                    for (int i = 0; i < metadata.partitions; i++) {
                                        persistentTopicExists.add(brokerService.getPulsar()
                                                .getPulsarResources().getTopicResources()
                                                .persistentTopicExists(topicName.getPartition(i)));
                                    }
                                    List<CompletableFuture<Boolean>> unmodifiablePersistentTopicExists =
                                            Collections.unmodifiableList(persistentTopicExists);
                                    return FutureUtil.waitForAll(unmodifiablePersistentTopicExists)
                                            .thenCompose(unused -> {
                                                // make sure all sub partitions were deleted after all future complete
                                                Optional<Boolean> anyExistPartition = unmodifiablePersistentTopicExists
                                                        .stream()
                                                        .map(CompletableFuture::join)
                                                        .filter(topicExist -> topicExist)
                                                        .findAny();
                                                if (anyExistPartition.isPresent()) {
                                                    log.info("Delete topic metadata failed because "
                                                            + "another partition exists");
                                                    throw new UnsupportedOperationException(
                                                            String.format("Another partition exists for [%s].",
                                                                    topicName));
                                                } else {
                                                    try {
                                                        return brokerService.getPulsar().getAdminClient().topics()
                                                                .deletePartitionedTopicAsync(topicName.toString());
                                                    } catch (PulsarServerException e) {
                                                        log.info()
                                                                .exception(e)
                                                                .log("Delete topic metadata failed due to failed to"
                                                                        + "get internal admin client.");
                                                        return CompletableFuture.failedFuture(e);
                                                    }
                                                }
                                            });
                                }))
                            );
                    }

View on GitHub (pinned to 820761864e)

Solutions

1. Wait for partition deletion to fully complete (all partitions removed from all brokers) and retry deletePartitionedTopic
2. Call deletePartitionedTopic with deleteSchema=false as appropriate and ensure no producers/consumers are recreating partitions, then retry
3. Force-delete remaining individual partitions first (topics().delete per partition), then delete the partitioned topic metadata
4. Check broker logs ('Delete topic metadata failed because another partition exists') and clear stale topic list caches if partitions are actually gone

Example fix

// before
admin.topics().deletePartitionedTopic(topic); // may fail if partitions linger
// after
await().atMost(30, SECONDS).until(() -> admin.topics().getPartitionedTopicMetadata(topic).partitions > 0 ? admin.topics().getList(namespace).stream().noneMatch(t -> t.startsWith(topic + PersistentTopicInternalTopics.PARTITIONED_TOPIC_SUFFIX) && !t.equals(topic)) : true);
admin.topics().deletePartitionedTopic(topic);

boolean allPartitionsGone(String partitionedTopic, int partitions) throws PulsarAdminException {
    for (int i = 0; i < partitions; i++) {
        try {
            admin.topics().getStats(partitionedTopic + "-partition-" + i);
            return false; // partition still exists
        } catch (PulsarAdminException.NotFoundException e) { /* gone */ }
    }
    return true;
}

try {
    admin.topics().deletePartitionedTopic(topic);
} catch (UnsupportedOperationException e) {
    if (e.getMessage().contains("Another partition exists")) {
        // wait for partitions to disappear, then retry
        Thread.sleep(5000);
        admin.topics().deletePartitionedTopic(topic);
    } else throw e;
}

When it happens

Trigger: Calling admin topics().deletePartitionedTopic() (or internal metadata delete) after the per-partition existence checks found at least one partition still present — e.g. partial partition deletion left partitions behind, or the existence check futures resolved true due to stale caches/timing.

Common situations: Retrying a previously failed delete where some partitions were removed; topic list/lookup caches briefly reporting the partitions as existing; delete of a partitioned topic whose partitions are being recreated concurrently by a producer; namespace/topic policy propagation delays in geo-replicated clusters.

Related errors

• The topic has a max partition index of %d, the number of par
• Package Management Service is not enabled in the broker.
• entryFilterNames can't be empty. To remove entry filters use
• The offloadPolicies must be specified for namespace offload.
• The driver is not supported, support value: ${supportedDrive

AI-assisted analysis of apache/pulsar@820761864e (2026-09-06). Data as JSON: /api/errors/6d13ab9bbcf68225. Report an issue: GitHub.