The properties of transactions are ACID

Transactions are the basic unit of concurrency control. A transaction is a sequence of SQL statement operations that are either executed or not executed. It is an indivisible unit of work.

For example, in a bank transfer, a debit is deducted from one account and a debit is added to another account. Both operations are performed or neither is performed.

Transactions for a database must be ACID, which stands for Atomic, Consistensy, Isolation, and persistence.

1. Atomicity

All operations in a transaction either complete or do not complete and do not end up somewhere in the middle. If an error occurs during the execution of a transaction, it is rolled back to the state before the transaction began, as if the transaction had never been executed.

2. Consistency

The database must be in a consistent state before and after a transaction.

If the transaction completes successfully, all changes to the database take effect.

If a transaction execution error occurs, all changes to the database are rolled back (undone) and returned to the original state.

3. Persistence:

When a transaction is committed, its changes to the database are permanent, and subsequent database failures have no impact on the data.

4. Isolation

When multiple users concurrently access the database, the transactions of one user cannot be disturbed by the transactions of other users, and multiple concurrent transactions must be isolated from each other.

Multiple transaction events are independent of each other and cannot interfere with each other.

Isolate the problem

Without considering the isolated nature of the transaction, the following problems occur due to the concurrency of the transaction:

1, dirty read – the most serious, prevent the occurrence

2. Do not read repeatedly

3. Phantom reading (virtual reading)

Dirty read: a transaction reads another transactionUncommitted data

Suppose that A transfers 100 yuan to B, the corresponding SQL statement is as follows:

Update account set money=money-100 where name= ‘a’; Update account set money=money+100 where nam= ‘b’;

After the two SQL statements are executed, B queries his own account and has 100 yuan more.

After B leaves, A rolls back the transaction, so B loses $100.

A transaction reads data that was not committed by another transaction. Very serious. Dirty reading should be avoided whenever possible.

Non-repeatable reads: data from a table is read multiple times within a transaction, with different results.

The difference between a non-repeatable read and a dirty read is that the committed data is read

L For example, the bank wants to check the balance of account A, and the result of the first check is 200 yuan, and A deposits another 100 yuan into the account. Right now, the result that the bank inquires again became 300 yuan. When the results of two queries are inconsistent, the bank is confused as to which one should prevail.

L Different from dirty reads, dirty reads read uncommitted data of the previous transaction, while unrepeatable reads read committed data of the previous transaction.

L Many people think there is nothing to be confused about, it must be the result of the following. We need to consider such A situation: to query the balance of account A, one is printed to the console and the other is output to the hard disk. In the same transaction, only the order is different, and the results of the two queries are inconsistent. Will you be confused which one shall prevail?

The current transaction query account A balance of $100, another transaction update balance of $300 and commit, resulting in the current transaction using the same query result turned into $300.

Phantom reading (virtual reading)

The insertion or deletion of data records by another transaction in one transaction, resulting in inconsistent data reads by the current transaction.

Third deposit 100 yuan but did not submit, at this moment the bank does statistic of report form total amount is 500 yuan, third will business is submitted, the bank became 600 yuan again statistic, two statistic result is inconsistent, the bank can be at a loss.

One transaction reads data that another transaction has committed, emphasizing changes in the number of records, often with SQL types of INSERT and DELETE.

The difference between virtual and non-repeatable reads:

Virtual reads emphasize changes in the number of data table records, primarily in INSERT and DELETE statements.

Non-repeatable reads emphasize changes to the contents of the data table, mainly update statements.

Isolation level of the database

The database defines four isolation levels (from high to low restriction, from low to high performance) :

Serializable: Avoids dirty read, unrepeatable read, and virtual read.

Repeatable Read: avoid dirty read, unrepeatable read and virtual read. Mysql uses repeatable reads.

Read COMMITTED: Dirty reads can be avoided, unrepeatable reads and virtual reads cannot be avoided. Oracle uses read commit.

Read Uncommitted: Dirty reads, unrepeatable reads, and virtual reads are unavoidable.

Query the isolation level of the current database:

select @@tx_isolation;

Set the isolation level of the transaction:

Set Session Transaction Isolation Level Transaction isolation level;

Isolation level performance issues

Performance comparison:

Serializable has the worst performance, with multiple transactions queued.

serializable < repeatable read < read committed < read uncommitted

Safety comparison:

Serializable has the best security performance and all problems can be avoided

serializable > repeatable read > read committed > read uncommitted

Analysis:

Serializable performance is poor

Read uncommitted Dirty reads cannot be avoided, which is a serious problem

Summary: The database vendor has chosen this scenario: read committed and repeatable reads

Repeatable read – mysql select repeatable read

Oracle selects read Committed

Transaction management in Java

Java JDBC operations default a command to a transaction, and transactions are committed automatically so you must set it to not commit automatically if you want to use transactions

Start transaction: connection.setautoCommit (false)
Execute groups of SQL statements
Rollback () Transaction end connection.rollback (); Connection.mit ();

Add the transaction

Demand: A transfers 100 yuan to B, simulation is abnormal. (Emphasize the importance of transactions: all or nothing.)Copy the code
	@Test
	public void demo2(a) {
		// Demand: A transfers 100 yuan to B, simulation is abnormal. (Emphasize the importance of transactions: all or nothing.)
		Connection conn = null;
		PreparedStatement stmt = null;

		try {
			// int j = 1 / 0;
			// Get the connection
			conn = JDBCUtils.getConnection();
			// Start the transaction
			conn.setAutoCommit(false);

			// a-100
			// Get the object to send the SQL
			String sql = "update account set money=money-100 where name='a'";
			stmt = conn.prepareStatement(sql);
			// Send SQL to get results
			stmt.executeUpdate();

			// b+100
			// Get the object to send the SQL
			String sql2 = "update account set money=money+100 where name='b'";
			stmt = conn.prepareStatement(sql2);
			// Send SQL to get results
			stmt.executeUpdate();

			int j = 1 / 0;
			System.out.println("All ok, commit transaction");
			conn.commit();
		} catch (Exception e) {
			e.printStackTrace();
			System.out.println("Exception occurred, rollback transaction");
			try {
				if(conn ! =null) { conn.rollback(); }}catch(Exception e1) { e1.printStackTrace(); }}finally{ JDBCUtils.release(conn, stmt); }}Copy the code

DBUtils transaction operation

When creating QueryRunner objects using DBUtils, you must select the no-parameter constructor. If you use the constructor passed in to the data source and the database link is managed by DBUtils, it is possible that different SQL instructions and different connection objects will not be able to use transactions normally

The method name of the Connection object describe
conn.setAutoCommit(false) Open the transaction
new QueryRunner() Create the core class without setting up the data source (manually managing connections)
Query (conn, SQL, handler, params) or update(conn, SQL, params) Manually pass the connection and execute THE SQL statement CRUD
DbUtils.commitAndCloseQuietly(conn) Submit and close the connection without throwing exceptions
DbUtils.rollbackAndCloseQuietly(conn) Roll back and close the connection without throwing an exception

Code demo

/* DBUtils transaction */
public class TransactionDemo2 {

    /* 1. Get connection 2. Start transaction 3. Commit transaction to release resource 5. If an exception occurs, roll back transaction to release resource */
    @Test
    public void test1(a) throws SQLException {
        Connection conn = null;
        try {
            //1. Get the connection
            conn =  DruidUtils.getConnetion();
            // 2. Start transaction
            conn.setAutoCommit(false);
            // 3. SQL = add money, subtract money
            QueryRunner qr = new QueryRunner();
            / / lose money
            String sql = "update account set money=money-? where name=?";
            qr.update(conn, sql, 1000."jack");
            // Simulation error
            int n =1/0;
            / / add money
            sql = "update account set money=money+? where name=?";
            qr.update(conn, sql, 1000."rose");
            //4. Commit transaction to release resources
            DbUtils.commitAndCloseQuietly(conn);
        } catch (Exception e) {
            e.printStackTrace();
            //5. If an exception occurs, roll back the transaction to release resourcesDbUtils.rollbackAndCloseQuietly(conn); }}}Copy the code