What causes "One or more partitions already exist for table '%s.%s'" in prestodb/presto?

Use ADD IF NOT EXISTS or check SHOW PARTITIONS before adding

prestodb/presto · error · PrestoException

ALREADY_EXISTS

ALREADY_EXISTS

Error message

One or more partitions already exist for table '%s.%s'

What it means

If the Hive metastore throws AlreadyExistsException during addPartitions, ThriftHiveMetastore maps it to a PrestoException with code ALREADY_EXISTS stating that one or more of the requested partitions already exist for the table. The whole batch add is aborted rather than silently skipping duplicates (unless IF NOT EXISTS semantics are used upstream).

Source

Thrown at presto-hive-metastore/src/main/java/com/facebook/presto/hive/metastore/thrift/ThriftHiveMetastore.java:1319

        if (partitions.isEmpty()) {
            return;
        }
        try {
            retry()
                    .stopOn(AlreadyExistsException.class, InvalidObjectException.class, MetaException.class, NoSuchObjectException.class, PrestoException.class)
                    .stopOnIllegalExceptions()
                    .run("addPartitions", stats.getAddPartitions().wrap(() ->
                            getMetastoreClientThenCall(metastoreContext, client -> {
                                int partitionsAdded = client.addPartitions(partitions);
                                if (partitionsAdded != partitions.size()) {
                                    throw new PrestoException(HIVE_METASTORE_ERROR,
                                            format("Hive metastore only added %s of %s partitions", partitionsAdded, partitions.size()));
                                }
                                return null;
                            })));
        }
        catch (AlreadyExistsException e) {
            throw new PrestoException(ALREADY_EXISTS, format("One or more partitions already exist for table '%s.%s'", databaseName, tableName), e);
        }
        catch (NoSuchObjectException e) {
            throw new TableNotFoundException(new SchemaTableName(databaseName, tableName));
        }
        catch (TException e) {
            throw new PrestoException(HIVE_METASTORE_ERROR, e);
        }
        catch (Exception e) {
            throw propagate(e);
        }
    }

    private <V> V getMetastoreClientThenCall(MetastoreContext metastoreContext, MetastoreCallable<V> callable)
            throws Exception
    {
        if (!impersonationEnabled) {
            try (HiveMetastoreClient client = clientProvider.createMetastoreClient(Optional.empty())) {
                return callable.call(client);

View on GitHub (pinned to 55bb57d202)

Solutions

Use ADD IF NOT EXISTS or check SHOW PARTITIONS before adding
Drop the existing partition first if re-creation is intended
Make the pipeline idempotent: skip partitions already registered
Coordinate concurrent writers so the same partition isn't added twice

Example fix

-- before
ALTER TABLE hive.db.tbl ADD PARTITION (dt='2026-09-01');
-- after
ALTER TABLE hive.db.tbl ADD IF NOT EXISTS PARTITION (dt='2026-09-01');

Defensive patterns

Strategy: validation

Validate before calling

-- check existing partitions before adding
SHOW PARTITIONS FROM hive.<db>.<table>;
-- add only if missing, or use IF NOT EXISTS
ALTER TABLE hive.<db>.<table> ADD IF NOT EXISTS PARTITION (dt='2026-09-01');

Try / catch

try {
    addPartitions(...);
} catch (PrestoException e) {
    if ("ALREADY_EXISTS".equals(e.getErrorCode().getName())) {
        // idempotent path: treat as success or skip existing partitions
    } else throw e;
}

Prevention

Use ADD IF NOT EXISTS for idempotent pipelines
Coordinate concurrent writers to the same partitions
Make retries idempotent after failed inserts
Check SHOW PARTITIONS before re-running partition DDL

When it happens

Trigger: addPartitions is invoked with a batch containing a partition (or partitions) already registered in the metastore — e.g. re-running an INSERT that created partitions, or ADD PARTITION without IF NOT EXISTS on an existing partition.

Common situations: Idempotent retries of a failed INSERT; two jobs writing the same partitions concurrently; re-running ALTER TABLE ADD PARTITION after a prior partial success; external tools (Hive/Spark) already registered the partition.

Related errors

AI-assisted analysis of prestodb/presto@55bb57d202 (2026-09-04). Data as JSON: /api/errors/d6a8dd02cfdb19d2. Report an issue: GitHub.