SQL++ Support for Couchbase Transactions

  • Capella Operational
  • concept
    +
    SQL++ offers full support for Couchbase ACID transactions based on optimistic concurrency.

    A transaction is a group of operations that are either committed to the database together, or are all undone from the database if there’s a failure. Refer to Transactions for an overview of Couchbase transactions.

    • Only DML statements are permitted within a transaction: INSERT, UPSERT, DELETE, UPDATE, MERGE, SELECT, EXECUTE FUNCTION, PREPARE, or EXECUTE.

    • The EXECUTE FUNCTION statement is only permitted in a transaction if the user-defined function does not contain any subqueries other than SELECT subqueries.

    • The PREPARE and EXECUTE statements are only permitted in a transaction for the DML statements listed above.

    All statements within a transaction are sent to the same Query node. If your cluster uses private endpoints, the load balancer ensures that the same query node carries out all the steps in the transaction.

    Statements

    SQL++ provides the following statements in support of Couchbase transactions. Refer to the documentation for each statement for more information and examples.

    Settings and Parameters

    The Query service provides settings and parameters in support of Couchbase transactions. Refer to the documentation for each parameter for more information and examples.

    Setting / Parameter Description

    txid request-level parameter

    Specifies the transaction to which a statement belongs.

    tximplicit request-level parameter

    Specifies that a statement is a single transaction.

    txstmtnum request-level parameter

    Specifies the transaction statement number.

    kvtimeout request-level parameter

    Specifies the maximum time to spend on a KV operation within a transaction before timing out.

    durability_level request-level parameter

    Specifies the transactional durability level.

    txtimeout request-level parameter

    Specify the maximum time to spend on a transaction before timing out.

    atrcollection request-level parameter

    Specify where the active transaction record is stored.

    In addition, the scan-consistency request-level parameter is used to specify the transactional scan consistency. Refer to Transactional Scan Consistency for details.

    Query Tools

    To create a Couchbase transaction using SQL++, you can use:

    There are slight differences in the way these tools operate when creating Couchbase transactions. These are explained in the sections below.

    Note that some Couchbase SDKs provide APIs to support Couchbase transactions. For further details, refer to Transactions in the Server documentation.

    Couchbase Transactions with the Query Tab

    • To execute a transaction containing multiple statements, compose the sequence of statements in the Query editor. Each statement must be terminated with a semicolon. After each statement, you must press Shift+Enter to start a new line without executing the query. You can then click Execute to execute the transaction.

    • You do not need to specify the txid parameter or the tximplicit parameter. If you need to specify any other parameters for the Couchbase transaction, you can use the Query Options window.

    Couchbase Transactions with the cbq shell

    • To execute a transaction containing multiple statements, you can create the transaction one statement at a time. Once you have started a transaction, all statements within the cbq shell session are assumed to be part of the same transaction until you rollback or commit the transaction. In this case, you don’t need to set the txid parameter. [1]

    • Alternatively, you can use the tximplicit parameter to run a single statement as a transaction. In this case, you do not need to specify the txid parameter either.

    • You can specify parameters for the Couchbase transaction using the \SET command.

    Monitoring

    You can monitor active Couchbase transactions using the system:transactions catalog. For more information, refer to system:transactions.

    Permissions

    When developing a transaction with an SDK, the transaction may contain a mixture of key-value operations and query statements.

    To execute key-value operations or query statements within a transaction, users must have the relevant cluster access privileges, with permissions on the relevant buckets, scopes and collections.

    Worked Example

    This worked example guides you through a complete Couchbase transaction session using SQL++.

    Preparation

    The worked example assumes that the supplied travel-sample bucket is installed. Refer to Import Sample Data for installation details.

    Context

    For this worked example, set the query context to the tenant_agent_00 scope in the travel sample dataset. For more information, see Query Context.

    The query context drop-down menu, with the tenant_agent_00 scope selected
    Index

    For the purposes of this worked example, you must create a primary index in the keyspace you will be using.

    Create a primary index on `travel-sample`.tenant_agent_00.bookings, so that you can query and update the documents in this keyspace.

    CREATE PRIMARY INDEX ON bookings;
    Parameters

    If necessary, set the transaction parameters for this worked example. In particular, you will turn off durability for the purposes of this example, in order to make sure that there are no problems meeting the transaction durability requirements.

    • Query Tab

    • CBQ Shell

    1. Click Query Options to display the Query Options window.

    2. Open the Scan Consistency drop-down list and select not_bounded.

    3. In the Transaction Timeout box, enter 120.

    4. In the Named Parameters section, click the + button to add a named parameter.

    5. When the new named parameter appears, enter durability_level in the name box and "none" (with double quotes) in the value box.

    6. Choose Save to save the preferences and return to the Query tab.

    Enter the following parameters.

    \SET -txtimeout "2m"; (1)
    \SET -scan_consistency "not_bounded"; (2)
    \SET -durability_level "none"; (3)
    1 The transaction timeout.
    2 The transaction scan consistency. No scan consistency is set for individual statements within the transaction; they inherit from the transaction scan consistency.
    3 Durability level of all the mutations within the transaction.

    Transaction

    Example 1. Transaction using the Query tab or cbq shell

    Copy the entire sequence below and paste it into either the Query tab or the cbq shell. Note that you must be using cbq shell version 2.0 or above.

    Transaction
    -- Start the transaction
    BEGIN WORK;
    
    -- Specify transaction settings
    SET TRANSACTION ISOLATION LEVEL READ COMMITTED;
    
    -- Create a booking document
    UPSERT INTO bookings
    VALUES("bf7ad6fa-bdb9-4099-a840-196e47179f03", {
      "date": "07/24/2021",
      "flight": "WN533",
      "flighttime": 7713,
      "price": 964.13,
      "route": "63986"
    });
    
    -- Set a savepoint
    SAVEPOINT s1;
    
    -- Update the booking document to include a user
    UPDATE bookings AS b
    SET b.`user` = "0"
    WHERE META(b).id = "bf7ad6fa-bdb9-4099-a840-196e47179f03";
    
    -- Check the content of the booking and user
    SELECT b.*, u.name
    FROM bookings b
    JOIN users u
    ON b.`user` = META(u).id
    WHERE META(b).id = "bf7ad6fa-bdb9-4099-a840-196e47179f03";
    
    -- Set a second savepoint
    SAVEPOINT s2;
    
    -- Update the booking documents to change the user
    UPDATE bookings AS b
    SET b.`user` = "1"
    WHERE META(b).id = "bf7ad6fa-bdb9-4099-a840-196e47179f03";
    
    -- Check the content of the booking and user
    SELECT b.*, u.name
    FROM bookings b
    JOIN users u
    ON b.`user` = META(u).id
    WHERE META(b).id = "bf7ad6fa-bdb9-4099-a840-196e47179f03";
    
    -- Roll back the transaction to the second savepoint
    ROLLBACK TRAN TO SAVEPOINT s2;
    
    -- Check the content of the booking and user again
    SELECT b.*, u.name
    FROM bookings b
    JOIN users u
    ON b.`user` = META(u).id
    WHERE META(b).id = "bf7ad6fa-bdb9-4099-a840-196e47179f03";
    
    -- Commit the transaction
    COMMIT WORK;

    The results of running the transaction in the Query tab are shown below. If you are using the cbq shell, the results are formatted differently, but contain the same information.

    Results
    [
      {
        "batchQuery": "-- Start the transaction\nBEGIN WORK;",
        "batchQueryResult": [
          {
            "txid": "d3bbf182-1179-42ba-8900-eb20fda69e42" (1)
          }
        ]
      },

    Beginning a transaction returns a unique transaction ID txid.

      {
        "batchQuery": "\n\n-- Specify transaction settings\nSET TRANSACTION ISOLATION LEVEL READ COMMITTED;",
        "batchQueryResult": []
      },
      {
        "batchQuery": "\n\n-- Create a booking document\nUPSERT INTO bookings\nVALUES(\"bf7ad6fa-bdb9-4099-a840-196e47179f03\", {\n  \"date\": \"07/24/2021\",\n  \"flight\": \"WN533\",\n  \"flighttime\": 7713,\n  \"price\": 964.13,\n  \"route\": \"63986\"\n});",
        "batchQueryResult": []
      },
      {
        "batchQuery": "\n\n-- Set a savepoint\nSAVEPOINT s1;",
        "batchQueryResult": []
      },
      {
        "batchQuery": "\n\n-- Update the booking document to include a user\nUPDATE bookings AS b\nSET b.`user` = \"0\"\nWHERE META(b).id = \"bf7ad6fa-bdb9-4099-a840-196e47179f03\";",
        "batchQueryResult": []
      },
      {
        "batchQuery": "\n\n-- Check the content of the booking and user\nSELECT b.*, u.name\nFROM bookings b\nJOIN users u\nON b.`user` = META(u).id\nWHERE META(b).id = \"bf7ad6fa-bdb9-4099-a840-196e47179f03\";",
        "batchQueryResult": [
          {
            "date": "07/24/2021",
            "flight": "WN533",
            "flighttime": 7713,
            "name": "Keon Hoppe",
            "price": 964.13,
            "route": "63986",
            "user": "0" (1)
          }
        ]
      },

    Before setting the second savepoint, the booking document has user "0", name "Keon Hoppe".

      {
        "batchQuery": "\n\n-- Set a second savepoint\nSAVEPOINT s2;",
        "batchQueryResult": []
      },
      {
        "batchQuery": "\n\n-- Update the booking documents to change the user\nUPDATE bookings AS b\nSET b.`user` = \"1\"\nWHERE META(b).id = \"bf7ad6fa-bdb9-4099-a840-196e47179f03\";",
        "batchQueryResult": []
      },
      {
        "batchQuery": "\n\n-- Check the content of the booking and user\nSELECT b.*, u.name\nFROM bookings b\nJOIN users u\nON b.`user` = META(u).id\nWHERE META(b).id = \"bf7ad6fa-bdb9-4099-a840-196e47179f03\";",
        "batchQueryResult": [
          {
            "date": "07/24/2021",
            "flight": "WN533",
            "flighttime": 7713,
            "name": "Rigoberto Bernier",
            "price": 964.13,
            "route": "63986",
            "user": "1" (1)
          }
        ]
      },

    After setting the second savepoint and performing an update, the booking document has user "1", name "Rigoberto Bernier".

      {
        "batchQuery": "\n\n-- Roll back the transaction to the second savepoint\nROLLBACK TRAN TO SAVEPOINT s2;",
        "batchQueryResult": []
      },
      {
        "batchQuery": "\n\n-- Check the content of the booking and user again\nSELECT b.*, u.name\nFROM bookings b\nJOIN users u\nON b.`user` = META(u).id\nWHERE META(b).id = \"bf7ad6fa-bdb9-4099-a840-196e47179f03\";",
        "batchQueryResult": [
          {
            "date": "07/24/2021",
            "flight": "WN533",
            "flighttime": 7713,
            "name": "Keon Hoppe",
            "price": 964.13,
            "route": "63986",
            "user": "0" (1)
          }
        ]
      },
      {
        "batchQuery": "\n\n-- Commit the transaction\nCOMMIT WORK;",
        "batchQueryResult": []
      }
    ]

    After rolling back to the second savepoint, the booking document again has user "0", name "Keon Hoppe".

    Example 2. Check the results of Example 1

    Check the result of committing the transaction.

    Query
    SELECT b.*, u.name
    FROM bookings b
    JOIN users u
    ON b.`user` = META(u).id
    WHERE META(b).id = "bf7ad6fa-bdb9-4099-a840-196e47179f03";
    Results
    {
      "date": "07/24/2021",
      "flight": "WN533",
      "flighttime": 7713,
      "name": "Keon Hoppe",
      "price": 964.13,
      "route": "63986",
      "user": "0" (1)
    }

    The booking document has been added with the attributes that were present when the transaction was committed.


    1. You must be using cbq shell version 2.0 or above to use the automatic transaction ID functionality.