What causes "Schema not empty: " in prestodb/presto?

Drop all tables first: for each table run DROP TABLE memory. .

prestodb/presto · error · PrestoException

SCHEMA_NOT_EMPTY

SCHEMA_NOT_EMPTY

Error message

Schema not empty:

What it means

MemoryMetadata.dropSchema refuses to drop a schema that still contains tables. After confirming the schema exists, it scans the tables map for any table whose schemaName matches; if found, it throws PrestoException SCHEMA_NOT_EMPTY. The schema must be emptied (all tables dropped) before it can be removed.

Source

Thrown at presto-memory/src/main/java/com/facebook/presto/plugin/memory/MemoryMetadata.java:129

    {
        if (schemas.contains(schemaName)) {
            throw new PrestoException(ALREADY_EXISTS, format("Schema [%s] already exists", schemaName));
        }
        schemas.add(schemaName);
    }

    @Override
    public synchronized void dropSchema(ConnectorSession session, String schemaName)
    {
        if (!schemas.contains(schemaName)) {
            throw new PrestoException(NOT_FOUND, format("Schema [%s] does not exist", schemaName));
        }

        boolean tablesExist = tables.values().stream()
                .anyMatch(table -> table.getSchemaName().equals(schemaName));

        if (tablesExist) {
            throw new PrestoException(SCHEMA_NOT_EMPTY, "Schema not empty: " + schemaName);
        }

        verify(schemas.remove(schemaName));
    }

    @Override
    public synchronized ConnectorTableHandle getTableHandle(ConnectorSession session, SchemaTableName schemaTableName)
    {
        Long tableId = tableIds.get(schemaTableName);
        if (tableId == null) {
            return null;
        }
        return tables.get(tableId);
    }

    @Override
    public synchronized ConnectorTableMetadata getTableMetadata(ConnectorSession session, ConnectorTableHandle tableHandle)
    {

View on GitHub (pinned to 55bb57d202)

Solutions

Drop all tables first: for each table run DROP TABLE memory.<schema>.<table>
Use DROP SCHEMA ... CASCADE if the SQL layer in your version supports it for this connector
Query system metadata (memory.information_schema.tables) to enumerate tables to drop
Reorder teardown scripts to drop tables before schemas

Example fix

// before
DROP SCHEMA memory.test_schema; -- SCHEMA_NOT_EMPTY
// after
DROP TABLE memory.test_schema.t1;
DROP TABLE memory.test_schema.t2;
DROP SCHEMA memory.test_schema;

Defensive patterns

Strategy: validation

Validate before calling

// Drop all tables in the schema before dropping the schema
List<String> tables = metadata.listTables(session, new SchemaTableName("memory", schemaName))
        .stream().map(t -> t.getTableName()).collect(toList());
for (String table : tables) {
    metadata.dropTable(session, new SchemaTableName(schemaName, table));
}
metadata.dropSchema(session, schemaName);

Try / catch

try {
    metadata.dropSchema(session, schemaName);
} catch (PrestoException e) {
    if (SCHEMA_NOT_EMPTY.equals(e.getErrorCode())) {
        dropAllTablesThenRetry(schemaName);
    } else throw e;
}

Prevention

Always drop child tables before schemas in teardown scripts
Enumerate tables via information_schema/system metadata first
Fail-safe cleanup: loop until schema lists zero tables
Avoid creating tables you never clean up in shared memory catalogs

When it happens

Trigger: DROP SCHEMA memory.<name> while tables created via CREATE TABLE memory.<name>.<t> still exist in that schema.

Common situations: Trying to tear down test schemas without dropping tables first; teardown scripts that skip table cleanup on failure; forgetting that memory connector has no CASCADE in this code path.

Related errors

AI-assisted analysis of prestodb/presto@55bb57d202 (2026-09-04). Data as JSON: /api/errors/70ce4ee9e33c07fe. Report an issue: GitHub.