apache/shardingsphere · error · OverallConnectionNotEnoughException

0

0

Error message

Can not get %d connections one time, partition succeed connection(%d) have released. Please consider increasing the 'maxPoolSize' of the data sources or decreasing the 'max-connections-size-per-query' in properties.

What it means

OverallConnectionNotEnoughException (converted to SQLException) thrown by DriverDatabaseConnectionManager.createConnections when it cannot obtain the requested number of connections from a datasource. Any SQLException from dataSource.getConnection() aborts the loop: all connections already acquired in this batch are closed (partition success released) and the aggregate error is thrown. The message tells you the two knobs: raise the pool's maxPoolSize or lower ShardingSphere's max-connections-size-per-query.

Source

Thrown at jdbc/src/main/java/org/apache/shardingsphere/driver/jdbc/core/connection/DriverDatabaseConnectionManager.java:401

        }
        synchronized (dataSource) {
            return createConnections(databaseName, dataSourceName, dataSource, connectionSize, connectionContext.getTransactionContext());
        }
    }
    
    private List<Connection> createConnections(final String databaseName, final String dataSourceName, final DataSource dataSource, final int connectionSize,
                                               final TransactionConnectionContext transactionConnectionContext) throws SQLException {
        List<Connection> result = new ArrayList<>(connectionSize);
        for (int i = 0; i < connectionSize; i++) {
            try {
                Connection connection = createConnection(databaseName, dataSourceName, dataSource, transactionConnectionContext);
                methodInvocationRecorder.replay(connection);
                result.add(connection);
            } catch (final SQLException ex) {
                for (Connection each : result) {
                    each.close();
                }
                throw new OverallConnectionNotEnoughException(connectionSize, result.size(), ex).toSQLException();
            }
        }
        return result;
    }
    
    private Connection createConnection(final String databaseName, final String dataSourceName, final DataSource dataSource,
                                        final TransactionConnectionContext transactionConnectionContext) throws SQLException {
        Optional<Connection> connectionInTransaction = getConnectionTransaction().getConnection(databaseName, dataSourceName, transactionConnectionContext);
        return connectionInTransaction.isPresent() ? connectionInTransaction.get() : dataSource.getConnection();
    }
    
    @Override
    public void close() throws SQLException {
        clearCachedConnections();
    }
}

View on GitHub (pinned to e952770a21)

Solutions

1. Increase the underlying datasource pool size (e.g. HikariCP maximumPoolSize) so connectionSize per query can be satisfied.
2. Lower max-connections-size-per-query in ShardingSphere props so a query reuses fewer connections per shard (memory restriction mode).
3. Reduce concurrent heavy queries or shard count involved per statement.
4. Check DB-side max_connections and network health if the underlying getConnection error indicates refusal rather than pool exhaustion.

Example fix

# before
props:
  max-connections-size-per-query: 8 # too high for pool size 10 with 4 shards
datasource:
  ds_0: maxPoolSize: 10
# after
props:
  max-connections-size-per-query: 2
datasource:
  ds_0: maximumPoolSize: 32

int poolMax = hikariConfig.getMaximumPoolSize();
int perQuery = configuredMaxConnectionsPerQuery;
if (shardCount * perQuery > poolMax) warn("pool too small");

catch (SQLException ex) { if (ex.getMessage().contains("maxPoolSize")) adjustPoolOrThrottle(); rethrow; }

When it happens

Trigger: A single query routed to many tables/shards needs connectionSize connections from one datasource, and the pool is exhausted or the DB refuses new connections mid-batch (maxPoolSize reached, DB max_connections hit, network failure).

Common situations: High sharding counts with default pool sizes (each query may need several connections per shard); OLAP/large result queries over many shards; concurrent queries exceeding HikariCP maximumPoolSize; DB-side connection limits; long transactions holding pool connections.

Related errors

• 13000
• 5
• Column index '%d' is out of range.
• Unsupported BLR version: %s
• Expected blr_begin

AI-assisted analysis of apache/shardingsphere@e952770a21 (2026-08-14). Data as JSON: /api/errors/b4a582e1aaf2871a. Report an issue: GitHub.