GridGain Developers Hub

JDBC Driver

GridGain is shipped with JDBC drivers that allow processing of distributed data using standard SQL statements like SELECT, INSERT, UPDATE or DELETE directly from the JDBC side.

Presently, there are two drivers supported by GridGain: the lightweight and easy to use JDBC Thin Driver described in this document and JDBC Client Driver that interacts with the cluster by means of a client node.

JDBC Thin Driver

The JDBC Thin driver is a default, lightweight driver provided by GridGain. To start using the driver, just add ignite-core-8.9.8.jar to your application’s classpath.

The driver connects to one of the cluster nodes and forwards all the queries to it for final execution. The node handles the query distribution and the result’s aggregations. Then the result is sent back to the client application.

The JDBC connection string may be formatted with one of two patterns: URL query or semicolon:

Connection String Syntax
// URL query pattern

hostAndPortRange := host[:port_from[..port_to]]

params := param1=value1[&param2=value2]...[&paramN=valueN]

// Semicolon pattern
  • host is required and defines the host of the cluster node to connect to.

  • port_from is the beginning of the port range to use to open the connection. 10800 is used by default if this parameter is omitted.

  • port_to is optional. It is set to the port_from value by default if this parameter is omitted.

  • schema is the schema name to access. PUBLIC is used by default. This name should correspond to the SQL ANSI-99 standard. Non-quoted identifiers are not case sensitive. Quoted identifiers are case sensitive. When semicolon format is used, the schema may be defined as a parameter with name schema.

  • <params> are optional.

The name of the driver’s class is org.apache.ignite.IgniteJdbcThinDriver. For instance, this is how you can open a JDBC connection to the cluster node listening on IP address

// Register JDBC driver.

// Open the JDBC connection.
Connection conn = DriverManager.getConnection("jdbc:ignite:thin://");


The following table lists all the parameters that are supported by the JDBC connection string:

Parameter Description Default Value


Username for the SQL Connection. This parameter is required if authentication is enabled on the server. See the Authentication and CREATE user documentation for more details.



Password for SQL Connection. Required if authentication is enabled on the server. See the Authentication and CREATE user documentation for more details.



Whether to execute distributed joins in non-colocated mode.



Whether to enforce join order of tables in the query. If set to true, the query optimizer will not reorder tables in the join.



Set this parameter to true if your SQL statement includes a GROUP BY clause that groups the results by either primary or affinity key. Whenever GridGain executes a distributed query, it sends subqueries to individual cluster members. If you know in advance that the elements of your query selection are colocated together on the same node and you group by a primary or affinity key, then GridGain makes significant performance and network optimizations by grouping data locally on each node participating in the query.



Whether the query contains only replicated tables. This is a hint for potentially more effective execution.



Whether to close server-side cursors automatically when the last piece of a result set is retrieved. When this property is enabled, calling ResultSet.close() does not require a network call, which could improve performance. However, if the server-side cursor is already closed, you may get an exception when trying to call ResultSet.getMetadata(). This is why it defaults to false.



Enables Partition Awareness mode. In this mode, the driver tries to determine the nodes where the data that is being queried is located and send the query to these nodes.



The number of distinct SQL queries that the driver keeps locally for optimization. When a query is executed for the first time, the driver receives the partition distribution for the table that is being queried and saves it for future use locally. When you query this table next time, the driver uses the partition distribution to determine where the data being queried is located to send the query to the right nodes. This local storage with SQL queries invalidates when the cluster topology changes. The optimal value for this parameter should equal the number of distinct SQL queries you are going to perform.



The number of distinct objects that represent partition distribution that the driver keeps locally for optimization. See the description of the previous parameter for details. This local storage with partition distribution objects invalidates when the cluster topology changes. The optimal value for this parameter should equal the number of distinct tables (cache groups) you are going to use in your queries.



Socket send buffer size. When set to 0, the OS default is used.



Socket receive buffer size. When set to 0, the OS default is used.



Whether to use TCP_NODELAY option.



Lazy query execution. By default, GridGain attempts to get and load the whole query result set into memory and then send it to the client. For small and medium result sets, this provides optimal performance and minimizes the duration of internal database locks, thus increasing concurrency. However, if the result set is too big to fit in the available memory, then it can lead to excessive GC pauses and even 'OutOfMemoryError’s. Use this flag to tell GridGain to fetch the result set lazily, thus minimizing memory consumption at the cost of a moderate performance hit.



Enables server side updates. When GridGain executes a DML operation, it fetches all the affected intermediate rows and sends them to the query initiator (also known as reducer) for analysis. Then it prepares batches of updated values to be sent to remote nodes. This approach might impact performance and it can saturate the network if a DML operation has to move many entries over it. Use this flag to tell GridGain to perform all intermediate row analysis and updates "in-place" on corresponding remote data nodes. Defaults to false, meaning that the intermediate results are fetched to the query initiator first.



Maximum amount of memory available to each query executed through the current connection. This parameter overrides the memory quota for queries.



Sets the number of seconds the driver will wait for a Statement object to execute. Zero means there is no limits.



Sets the number of milliseconds JDBC client will waits for server to response. Zero means there is no limits.


For the list of security parameters, refer to the Using SSL section.

Connection String Examples

  • jdbc:ignite:thin://myHost - connect to myHost on the port 10800 with all defaults.

  • jdbc:ignite:thin://myHost:11900 - connect to myHost on custom port 11900 with all defaults.

  • jdbc:ignite:thin://myHost:11900;user=ignite;password=ignite - connect to myHost on custom port 11900 with user credentials for authentication.

  • jdbc:ignite:thin://myHost:11900;distributedJoins=true&autoCloseServerCursor=true - connect to myHost on custom port 11900 with enabled distributed joins and autoCloseServerCursor optimization.

  • jdbc:ignite:thin://myHost:11900/myschema; - connect to myHost on custom port 11900 and access to MYSCHEMA.

  • jdbc:ignite:thin://myHost:11900/"MySchema";lazy=false - connect to myHost on custom port 11900 with disabled lazy query execution and access to MySchema (schema name is case sensitive).

Multiple Endpoints

You can enable automatic failover if a current connection is broken by setting multiple connection endpoints in the connection string. The JDBC Driver randomly picks an address from the list to connect to. If the connection fails, the JDBC Driver selects another address from the list until the connection is restored. The Driver stops reconnecting and throws an exception if all the endpoints are unreachable.

The example below shows how to pass three addresses via the connection string:

// Register JDBC Driver.

// Open the JDBC connection passing several connection endpoints.
Connection conn = DriverManager

Partition Awareness

Partition awareness is a feature that makes the JDBC driver "aware" of the partition distribution in the cluster. It allows the driver to pick the nodes that own the data that is being queried and send the query directly to those nodes (if the addresses of the nodes are provided in the driver’s configuration). Partition awareness can increase average performance of queries that use the affinity key.

Without partition awareness, the JDBC driver connects to a single node, and all queries are executed through that node. If the data is hosted on a different node, the query has to be rerouted within the cluster, which adds an additional network hop. Partition awareness eliminates that hop by sending the query to the right node.

To make use of the partition awareness feature, provide the addresses of all the server nodes in the connection properties. The driver will route requests to the nodes that store the data requested by the query.

To enable partition awareness, add the partitionAwareness=true parameter to the connection string and provide the endpoints of multiple server nodes:


Connection conn = DriverManager

Also see the description of the two related parameters: partitionAwarenessSQLCacheSize and partitionAwarenessPartitionDistributionsCacheSize.

Cluster Configuration

In order to accept and process requests from JDBC Thin Driver, a cluster node binds to a local network interface on port 10800 and listens to incoming requests.

Use an instance of ClientConnectorConfiguration to change the connection parameters:

<bean id="ignite.cfg" class="org.apache.ignite.configuration.IgniteConfiguration">
  <property name="clientConnectorConfiguration">
    <bean class="org.apache.ignite.configuration.ClientConnectorConfiguration" />
IgniteConfiguration cfg = new IgniteConfiguration()
    .setClientConnectorConfiguration(new ClientConnectorConfiguration());
This API is not presently available for C#/.NET. You can use XML configuration.
This API is not presently available for C++. You can use XML configuration.

The following parameters are supported:

Parameter Description Default Value


Host name or IP address to bind to. When set to null, binding is made to localhost.



TCP port to bind to. If the specified port is already in use, GridGain tries to find another available port using the portRange property.



Defines the number of ports to try to bind to. E.g. if the port is set to 10800 and portRange is 100, then the server tries to bind consecutively to any port in the [10800, 10900] range until it finds a free port.



Maximum number of cursors that can be opened simultaneously for a single connection.



Number of request-handling threads in the thread pool.

MAX(8, CPU cores)


Size of the TCP socket send buffer. When set to 0, the system default value is used.



Size of the TCP socket receive buffer. When set to 0, the system default value is used.



Whether to use TCP_NODELAY option.



Idle timeout for client connections. Clients are disconnected automatically from the server after remaining idle for the configured timeout. When this parameter is set to zero or a negative value, the idle timeout is disabled.



Whether access through JDBC is enabled.



Whether access through thin client is enabled.



If SSL is enabled, only SSL client connections are allowed. The node allows only one mode of connection: SSL or plain. A node cannot receive both types of client connections. But this option can be different for different nodes in the cluster.



Whether to use SSL context factory from the node’s configuration (see IgniteConfiguration.sslContextFactory).



Whether client authentication is required.



The class name that implements Factory<SSLContext> to provide node-side SSL. See this for more information.


Using SSL

You can configure the JDBC Thin Driver to use SSL to secure communication with the cluster. SSL must be configured both on the cluster side and in the JDBC Driver. Refer to the SSL for Thin Clients and JDBC/ODBC section for the information about cluster configuration.

To enable SSL in the JDBC Driver, pass the sslMode=require parameter in the connection string and provide the key store and trust store parameters:


String keyStore = "keystore/node.jks";
String keyStorePassword = "123456";

String trustStore = "keystore/trust.jks";
String trustStorePassword = "123456";

try (Connection conn = DriverManager.getConnection("jdbc:ignite:thin://"
        + "&sslClientCertificateKeyStoreUrl=" + keyStore + "&sslClientCertificateKeyStorePassword="
        + keyStorePassword + "&sslTrustCertificateKeyStoreUrl=" + trustStore
        + "&sslTrustCertificateKeyStorePassword=" + trustStorePassword)) {

    ResultSet rs = conn.createStatement().executeQuery("select 10");;
} catch (Exception e) {

The following table lists all parameters that affect SSL/TLS connection:



Default Value


Enables SSL connection. Available modes:

  • require: SSL protocol is enabled on the client. Only SSL connection is available.

  • disable: SSL protocol is disabled on the client. Only plain connection is supported.



Protocol name for secure transport. Protocol implementations supplied by JSSE: SSLv3 (SSL), TLSv1 (TLS), TLSv1.1, TLSv1.2



The Key manager algorithm to be used to create a key manager. Note that in most cases the default value is sufficient. Algorithms implementations supplied by JSSE: PKIX (X509 or SunPKIX), SunX509.



URL of the client key store file. This is a mandatory parameter since SSL context cannot be initialized without a key manager. If sslMode is require and the key store URL isn’t specified in the GridGain properties, the value of the JSSE property will be used.

The value of the system property.


Client key store password.

If sslMode is require and the key store password isn’t specified in the GridGain properties, the JSSE property will be used.

The value of the keyStorePassword system property.


Client key store type used in context initialization.

If sslMode is require and the key store type isn’t specified in the GridGain properties, the JSSE property will be used.

The value of the system property. If the system property is not defined, the default value is JKS.


URL of the trust store file. This is an optional parameter; however, one of these properties must be set: sslTrustCertificateKeyStoreUrl or sslTrustAll

If sslMode is require and the trust store URL isn’t specified in the GridGain properties, the JSSE property will be used.

The value of the system property.


Trust store password.

If sslMode is require and the trust store password isn’t specified in the GridGain properties, the JSSE property will be used.

The value of the system property


Trust store type.

If sslMode is require and the trust store type isn’t specified in the GridGain properties, the JSSE property will be used.

The value of the system property. If the system property is not defined the default value is JKS


Disables server’s certificate validation. Set to true to trust any server certificate (revoked, expired, or self-signed SSL certificates).



Class name of the custom implementation of the Factory<SSLSocketFactory>.

If sslMode is require and a factory is specified, the custom factory is used instead of the JSSE socket factory. In this case, other SSL properties are ignored.


The default implementation is based on JSSE, and works through two Java keystore files:

  • sslClientCertificateKeyStoreUrl - the client certificate keystore holds the keys and certificate for the client.

  • sslTrustCertificateKeyStoreUrl - the trusted certificate keystore contains the certificate information to validate the server’s certificate.

The trusted store is an optional parameter, however one of the following parameters: sslTrustCertificateKeyStoreUrl or sslTrustAll must be configured.

If you want to use your own implementation or method to configure the SSLSocketFactory, you can use JDBC Driver’s sslFactory parameter. It is a string that must contain the name of the class that implements the interface Factory<SSLSocketFactory>. The class must be available for JDBC Driver’s class loader.

Handling Disconnects

Auto-recovery logic applies only to idempotent requests (like SELECT).

If the connection cannot be re-established, the current statement is forced to close, and an SQLException with SQL state 08006 is raised. Once that happens, you can retry the operation at your own risk.

The connection is not re-established if MVCC is used. In this case, the connection is forced to close.

This is what a disconnection looks like in the Java code:

catch(SQLException ex) {
  if (!conn.isClosed() && "08006".equals(ex.getSQLState()))

The following example showcases a method for repeating an operation.

try (Connection conn = DriverManager.getConnection(URL)) {
        executeWithRetries(conn, 3, () -> {
            try (Statement stmt = conn.createStatement()) {
                return stmt.executeUpdate("INSERT INTO TEST VALUES (1, 1)");

<T> T executeWithRetries(Connection conn, int retryCount, Callable<T> sqlCall) throws SQLException {
    for (; ; ) {
        try {
        catch (SQLException e) {
            if (--retryCount == 0) {
                // The number of retries exceeded.
                throw e;
            if (!"08006".equals(e.getSQLState())) {
                // Other kind of error has occurred.
                throw e;
            if (conn.isClosed()) {
                // Must re-establish connection manually.
                throw e;
        catch (Exception e) {
            throw (RuntimeException)e;

Ignite DataSource

The DataSource object is used as a deployed object that can be located by logical name via the JNDI naming service. JDBC Driver’s org.apache.ignite.IgniteJdbcThinDataSource implements a JDBC DataSource interface allowing you to utilize the DataSource interface instead.

In addition to generic DataSource properties, IgniteJdbcThinDataSource supports all the Ignite-specific properties that can be passed into a JDBC connection string. For instance, the distributedJoins property can be (re)set via the IgniteJdbcThinDataSource#setDistributedJoins() method.

Refer to the JavaDocs for more details.


To start processing the data located in the cluster, you need to create a JDBC Connection object via one of the methods below:

// Open the JDBC connection via DriverManager.
Connection conn = DriverManager.getConnection("jdbc:ignite:thin://");


// Or open connection via DataSource.
IgniteJdbcThinDataSource ids = new IgniteJdbcThinDataSource();

Connection conn = ids.getConnection();

Then you can execute SQL SELECT queries as follows:

// Query people with specific age using prepared statement.
PreparedStatement stmt = conn.prepareStatement("select name, age from Person where age = ?");

stmt.setInt(1, 30);

ResultSet rs = stmt.executeQuery();

while ( {
    String name = rs.getString("name");
    int age = rs.getInt("age");
    // ...

You can also modify the data via DML statements.


// Insert a Person with a Long key.
PreparedStatement stmt = conn
        .prepareStatement("INSERT INTO Person(_key, name, age) VALUES(CAST(? as BIGINT), ?, ?)");

stmt.setInt(1, 1);
stmt.setString(2, "John Smith");
stmt.setInt(3, 25);



// Merge a Person with a Long key.
PreparedStatement stmt = conn
        .prepareStatement("MERGE INTO Person(_key, name, age) VALUES(CAST(? as BIGINT), ?, ?)");

stmt.setInt(1, 1);
stmt.setString(2, "John Smith");
stmt.setInt(3, 25);



// Update a Person.
  executeUpdate("UPDATE Person SET age = age + 1 WHERE age = 25");


conn.createStatement().execute("DELETE FROM Person WHERE age = 25");


JDBC Driver allows streaming data in bulk using the SET command. See the SET command documentation for more information.

Error Codes

The JDBC drivers pass error codes in the java.sql.SQLException class, used to facilitate exception handling on the application side. To get an error code, use the java.sql.SQLException.getSQLState() method. It returns a string containing the ANSI SQLSTATE error code defined:

PreparedStatement ps;

try {
    ps = conn.prepareStatement("INSERT INTO Person(id, name, age) values (1, 'John', 'unparseableString')");
} catch (SQLException e) {
    switch (e.getSQLState()) {
    case "0700B":
        System.out.println("Conversion failure");

    case "42000":
        System.out.println("Parsing error");

        System.out.println("Unprocessed error: " + e.getSQLState());

The table below lists all the ANSI SQLSTATE error codes currently supported by GridGain. Note that the list may be extended in the future.

Code Description


Conversion failure (for example, a string expression cannot be parsed as a number or a date).


Invalid transaction isolation level.


The driver failed to open a connection to the cluster.


The connection is in the closed state. Happened unexpectedly.


The connection was rejected by the cluster.


I/O error during communication.


Null value not allowed.


Unsupported parameter type.


Data integrity constraint violation.


Invalid result set state.


Requested operation is not supported.


Concurrent update conflict. See Concurrent Updates.


Query parsing exception.


Internal error. The code is not defined by ANSI and refers to an Ignite specific error. Refer to the java.sql.SQLException error message for more information.