We are developing a website using
Tomcat 7
JDBC
PostgreSQL 9.2
We've had some connection leaks and think we corrected them (the database no longer stops responding), but the behaviour of the connection pool still seems leaky, as we have a number of idle connections greater than the maxIdle set in context.xml. I'd like to be sure the problem is fixed.
For testing purposes, I'm using the following context.xml :
<Resource
auth="Container"
name="jdbc/postgres"
factory="org.apache.tomcat.jdbc.pool.DataSourceFactory"
type="javax.sql.DataSource"
username="admin"
password="..."
driverClassName="org.postgresql.Driver"
url="jdbc:postgresql://127.0.0.1:5432/..."
initialSize="1"
maxActive="50"
minIdle="0"
maxIdle="3"
maxWait="-1"
minEvictableIdleTimeMillis="1000"
timeBetweenEvictionRunsMillis="1000"
/>
If I understand correctly, we should have 1 idle connection on startup and from 0 to 3 depending on the load, right ?
What is happening is : 1 connection on startup, up to 3 idle connections if the load is low, and more than 3 idle connections after a high load. Then these connections are not closed immediatly, and we don't know when/if they will be closed (sometime some of them are closed).
So the question is : is this behaviour normal, or not ?
Thanks for your help
EDIT : added factory attribute, didn't change the problem
EDIT 2 : using removeAbandoned & removeAbandonedTimeout make the idle connexions being closed every removeAbandonedTimeout. So we probably still have some connection leaks. Here are some pieces of code we are using to connect to the database and execute requests :
PostgreSQLConnectionProvider, just a static class to provide a connection :
public class PostgreSQLConnectionProvider {
public static Connection getConnection() throws NamingException, SQLException {
String dsString = "java:/comp/env/jdbc/postgres";
Context context = new InitialContext();
DataSource ds = (DataSource) context.lookup(dsString);
Connection connection = ds.getConnection();
return connection;
}
}
DAO abstract class :
public abstract class DAO implements java.lang.AutoCloseable {
// Private attributes :
private Connection _connection;
// Constructors :
public DAO() {
try { _connection = PostgreSQLConnectionProvider.getConnection(); }
catch (NamingException | SQLException ex) {
Logger.getLogger(DAO.class.getName()).log(Level.SEVERE, null, ex);
}
}
// Getters :
public Connection getConnection() { return _connection; }
// Closeable :
#Override
public void close() throws SQLException {
if(!_connection.getAutoCommit()) {
_connection.rollback();
_connection.setAutoCommit(true);
}
_connection.close();
}
}
UserDAO, a small DAO subclass (we have several DAO sublasses to request the database) :
public class UserDAO extends DAO {
public User getUserWithId(int id) throws SQLException {
PreparedStatement ps = null;
ResultSet rs = null;
User user = null;
try {
String sql = "select * from \"USER\" where id_user = ?;";
ps = getConnection().prepareStatement(sql);
ps.setInt(1, id);
rs = ps.executeQuery();
rs.next();
String login = rs.getString("login");
String password = rs.getString("password");
String firstName = rs.getString("first_name");
String lastName = rs.getString("last_name");
String email = rs.getString("email");
user = new User(id, login, password, firstName, lastName, email);
}
finally {
if(rs != null) rs.close();
if(ps != null) ps.close();
}
return user;
}
}
An example of a DAO subclass use :
try(UserDAO dao = new UserDAO()) {
try {
User user = dao.getUserWithId(52);
}
catch (SQLException ex) {
// Handle exeption during getUserWithId
}
}
catch (SQLException ex) {
// Handle exeption during dao.close()
}
Looking at the code it appears the connection is grabbed for the lifetime of the DAO, not the lifetime of the statement, which is the usual expectation. Normally, you would grab a connection from the pool just as you're about to execute the statement, and call close() on it when you're done in order to return it to the pool.
Additionally, in your finally clause, both rs.close() and ps.close() can throw exceptions resulting in missing the last call against the prepared statement.
In Java 7 you can also use a try with resources statement that will close both the prepared statement and the connection for you. According to the spec, the driver is supposed to close the result for you when the statement is closed.
Related
I'm currently working on a college project, and I'm creating a very simple e-commerce style website.
I'm using JDBC driver manager and connection pool for the connection to the db, while using Tomcat 9.0 as the container.
The problem is: when I modify some product through the website (let's say the amount available for example), the website doesn't always reflect the changes, while I can always see the data correctly in MySql Workbench.
It actually works one time out of two on the same query:
I run the query for the first time after the changes -> it shows the old value
I run the query for the second time after the changes -> it shows the new value
I run the query for the third time after the changes -> it shows the old value
And so on.
I've already tried to set caching off (from the query, using the SQL_NO_CACHE), but it didn't seem to solve the problem, I've tried to use Datasource instead, but it causes other problems that most likely I won't have the time to solve.
This is the connection pool file, which I think might be problem, I'm not that sure tho:
public class DriverManagerConnectionPool {
private static List<Connection> freeDbConnections;
static {
freeDbConnections = new LinkedList<Connection>();
try {
Class.forName("com.mysql.cj.jdbc.Driver");
} catch (ClassNotFoundException e) {
System.out.println("DB driver not found:"+ e.getMessage());
}
}
private static synchronized Connection createDBConnection() throws SQLException {
Connection newConnection = null;
String ip = "localhost";
String port = "3306";
String db = "storage";
String username = "root";
String password = "1234";
newConnection = DriverManager.getConnection("jdbc:mysql://"+ ip+":"+ port+"/"+db+"?useUnicode=true&useJDBCCompliantTimezoneShift=true&useLegacyDatetimeCode=false&serverTimezone=UTC", username, password);
newConnection.setAutoCommit(false);
return newConnection;
}
public static synchronized Connection getConnection() throws SQLException {
Connection connection;
if (!freeDbConnections.isEmpty()) {
connection = (Connection) freeDbConnections.get(0);
freeDbConnections.remove(0);
try {
if (connection.isClosed())
connection = getConnection();
} catch (SQLException e) {
connection.close();
connection = getConnection();
}
} else {
connection = createDBConnection();
}
return connection;
}
public static synchronized void releaseConnection(Connection connection) throws SQLException {
if(connection != null) freeDbConnections.add(connection);
}
}
I really hope you can help me, I haven't found any solution online!
I guess it is because of auto-commit is disabled. Please try using #Transactional or set auto-commit to true. You can also try to use db.commit after each statement.
As per your connection pool implementation, all connection in your pool seems to be auto committed false.
Please check you have properly committed the connection after executing the query or not.
So it might be the case that, when executing the query after changes with same connection it reflects those changes, done earlier and on other connections, old values are might get returned.
I need to make a special treatment when a connection problem to the database is occurring like database server down and not an sql problem.
In the source code we can get various exceptions but which ones are belonging to the connection ones ?
We would like if this kind of problem occurs to make less logs.
EDITED
I have many methods that perform connection to the database but all get the session from the same method (initSession):
Here an example:
private Session initSession(HibernateUtil hibernateUtil) {
Session oSession = null;
try {
oSession = hibernateUtil.getSession();
} catch (Exception e) {
log.error("unable to log, Please check the details of your database");
}
return oSession;
}
public List findAlerts(int pFirstLine, int pNbElement) throws AnalyzerException {
List oAlerts = new ArrayList();
Session oSession = initSession(lHibernateUtil);
try {
oAlerts = AlertFinders.instance().findAlertByStatus(oSession, false, pFirstLine, pNbElement);
Iterator iterAlerts = oAlerts.iterator();
while (iterAlerts.hasNext()) {
...
}
} catch (UnableToLocateObjectException eU) {
throw new AnalyzerException(eU.getMessageSource(), eU.getClassNameSource(), eU.getMethodSource(), eU);
} finally {
oSession.close();
}
return oAlerts;
}
Multiple possible ways.
Use Java Connection isValid method.
Use connection pool - All major connection
pools support this functionality (including c3p0 and dbcp).They can
throw SQLException has getErrorCode() and getSQLState() methods
Write Java code & poll frequently - sample code below
Run arguments sample: jdbc:oracle:thin:#localhost:1521:XE system mypassword123 oracle.jdbc.driver.OracleDriver
public class DbConnCheck {
public static void main(String[] args) throws Exception {
String url = args[0];
String username = args[1];
String password = args[2];
String driver = args[3];
Class.forName(driver);
Connection conn = DriverManager.getConnection(url, username, password);
try {
Statement statement = conn.createStatement();
ResultSet rs = statement.executeQuery("SELECT SYSDATE FROM DUAL");
while(rs.next()) {
System.out.println(rs.getObject(1));
}
} finally {
conn.close();
}
}
}
Edit : Adding details on hibernate part
Not done in Hibernate but to be precise you can check in connection pool configuration.
If using c3p0 then check how you can best use setting like idle_test_period, preferredTestQuery and testConnectionOnCheckout;
If using dbcp then validationQuery can do the job.
If you want to use c3p0 with Hibernate and Spring check this link
I have many threads accessing MYSQL database, at first I didn't use connection pool so I had this error "You can't operate on a closed ResultSet"
I searched on Google and found out that I should used connection pool so I tried c3p0 API for implementation, but I still have the same problem and nothing changed.
so should I Synchronize getAllcountries method or there's another better solution.
public class DataSource {
private static DataSource datasource;
private ComboPooledDataSource cpds ;
private DataSource() throws IOException, SQLException, PropertyVetoException {
cpds = new ComboPooledDataSource();
cpds.setDriverClass("com.mysql.jdbc.Driver"); //loads the jdbc driver
cpds.setJdbcUrl("jdbc:mysql://localhost/question_game");
cpds.setUser("root");
cpds.setPassword("");
cpds.setMaxPoolSize(500);
}
public static DataSource getInstance() throws IOException, SQLException, PropertyVetoException {
if (datasource == null) {
datasource = new DataSource();
return datasource;
} else {
return datasource;
}
}
public Connection getConnection() throws SQLException {
return this.cpds.getConnection();
}
public List<Country> getAllCountries() {
String query = "SELECT * FROM country order by name ";
List<Country> list = new ArrayList<Country>();
Country country = null;
ResultSet rs = null;
try {
try {
connection = DataSource.getInstance().getConnection();
} catch (IOException e) {
e.printStackTrace();
} catch (PropertyVetoException e) {
e.printStackTrace();
}
statement = connection.createStatement();
rs = statement.executeQuery(query);
while (rs.next()) {
//getting countries
}
} catch (SQLException ex) {
ex.printStackTrace();
} finally {
//DbUtil used to check if null
DbUtil.close(rs);
DbUtil.close(statement);
DbUtil.close(connection);
}
return list;
}
In addition to #stephen-c 's observation,
you basically have two options: either synchronize getAllCountries method, or make the database connection local to that method instead of having it as a class member.
As you have it now, 'connection' is a class member (available to all invocations of getAllCountries(), on all threads), so the connection is probably being overwritten by a second thread. Move it to a variable in the method, and then each invocation of the method will have its own connection.
I see a method called getConnection, I don't see where you are calling it. Instead, I see your getAllCountries method using a static DataSource object that is created like this:
new DataSource();
I searched on Google and found out that I should used connection pool so I tried c3p0 API for implementation, but I still have the same problem and nothing change
Yea ...
It looks like you just cut-and-pasted the code you found into your application without really thinking about it. Obviously ... you need to understand what the code is doing AND figure out how to use it in your application.
I deployed my first Java web application a couple of days ago and realized a strange thing was happening. After a period of time all the dynamic content and functionality that relied on a connection to my database (testimonial submission, admin login) stopped working. It seems like this is happening every 24 hours or so. Every morning I realize it isn't working again.
I solve the issue by going in to the Tomcat web application manager and clicking "reload" on the web app in question. Immediately the dynamic features of the website work again.
My server is running Tomcat 7 and MySQL and the web app uses the JDBC driver to establish the connection to the database. I've made no alterations to Apache or Tomcat settings.
I have other web apps written in PHP that work persistently without fault it just seems to be this Java web app that has this problem.
What would cause this to happen and how can I make it so the web app doesn't need to be reloaded before it can establish a database connection again?
EDIT: attached some code for database connection
Database connection
public class DBConnection {
private static Connection conn;
private static final Configuration conf = new Configuration();
private static final String dbDriver = conf.getDbDriver();
private static final String dbHostName = conf.getDbHostname();
private static final String dbDatabaseName = conf.getDbDatabaseName();
private static final String dbUsername = conf.getDbUsername();
private static final String dbPassword = conf.getDbPassword();
public Connection getConnection(){
try{
Class.forName(dbDriver);
Connection conn = (Connection) DriverManager.getConnection(dbHostName + dbDatabaseName, dbUsername, dbPassword);
return conn;
} catch(Exception e){
e.printStackTrace();
}
return conn;
}
public void disconnect(){
try{
conn.close();
} catch (Exception e){}
}
}
Controller for login form:
protected void doPost(HttpServletRequest request, HttpServletResponse response) throws ServletException, IOException {
String form = request.getParameter("form");
// check login details
if(form.equals("loginForm")){
String username = request.getParameter("username").trim();
String password = request.getParameter("password").trim();
password = loginService.hashPassword(password);
boolean isValidUser = loginService.checkUser(username, password);
if(isValidUser){
Cookie loggedIn = new Cookie("loggedIn", "true");
loggedIn.setMaxAge(60*60*24);
response.addCookie(loggedIn);
out.print("success");
}else{
out.print("nope");
}
}
}
Login service checks login details are correct:
public boolean checkUser(String username, String password){
boolean isValid = false;
try{
sql = "SELECT username, password FROM morleys_user WHERE username=? AND password=? AND isActive=1 LIMIT 1";
prep = conn.prepareStatement(sql);
prep.setString(1, username);
prep.setString(2, password);
rs = prep.executeQuery();
if(rs.next()){
return true;
}
}catch(Exception e){
e.printStackTrace();
}finally{
connection.disconnect();
}
return isValid;
}
UPDATE
If I understand correctly I should not be handling a direct connection to a database and instead be using a service that will manage connections for me.
This is my example of establishing a DataSource connection to a MysQL database.
Establish a new DataSource instance of this class:
package uk.co.morleys;
import java.io.FileInputStream;
import java.io.IOException;
import java.util.Properties;
import javax.sql.DataSource;
import com.mysql.jdbc.jdbc2.optional.MysqlDataSource;
public class DataSourceFactory {
public static DataSource getMySQLDataSource() {
Properties props = new Properties();
FileInputStream fis = null;
MysqlDataSource mysqlDS = null;
try {
fis = new FileInputStream("db.properties");
props.load(fis);
mysqlDS = new MysqlDataSource();
mysqlDS.setURL(props.getProperty("MYSQL_DB_URL"));
mysqlDS.setUser(props.getProperty("MYSQL_DB_USERNAME"));
mysqlDS.setPassword(props.getProperty("MYSQL_DB_PASSWORD"));
} catch (IOException e) {
e.printStackTrace();
}
return mysqlDS;
}
}
Instantiating a new DataSource for checking user login details
public boolean checkUser(String username, String password){
boolean isValid = false;
DataSource ds = DataSourceFactory.getMySQLDataSource();
Connection con = null;
ResultSet rs = null;
PreparedStatement ps = null;
try{
con = ds.getConnection();
sql = "SELECT username, password FROM morleys_user WHERE username=? AND password=? AND isActive=1 LIMIT ";
ps = con.prepareStatement(sql);
ps.setString(1, username);
ps.setString(2, password);
rs = ps.executeQuery();
if(rs.next()){
return true;
}
}catch(SQLException e){
e.printStackTrace();
}finally{
try {
if(rs != null) rs.close();
if(ps != null) ps.close();
if(con != null) con.close();
} catch (SQLException e) {
e.printStackTrace();
}
}
return isValid;
}
Given that you've never heard of a connection pool before I'm assuming that you not are not very effectively managing database resources.
The most basic way to access the database is to obtain a connection, execute some statements & close the connection.
In the code you provided I don't see you obtaining or closing a connection, so I assume that you create a single connection when you start your application and keep the connection open "forever". After a certain amount of time your MySql server decides to kill the connection as it's been open for too long.
When you create and close a connection each time you need one, you normally won't encounter any connection timeouts, but you might experience a lot overhead from creating a connection each time your application needs one.
This is where a connection pool comes in; a connection pool manages a number of database connections and your application borrows one each time it needs one. By properly configuring your connection pool the pool will normally transparently take care of broken connections (you might for example configure the pool to renew a connection once it's x minutes/hours old).
You also need to pay attention to resource management; e.g. close a statement as soon as you no longer need it.
The following code demonstrates how your "check user" method can be improved:
public boolean checkUser(String username, String password) throws SQLException {
//acquire a java.sql.DataSource; the DataSource is typically a connection pool that's set-up in the application of obtained via jndi
DataSource dataSource = acquireDataSource();
//java 7 try-with-resources statement is used to make sure that resources are properly closed
//obtain a connection from the pool. Upon closing the connection we return it to the pool
try (Connection connection = dataSource.getConnection()) {
//release resources associated with the PreparedStatement as soon as we no longer need it.
try(PreparedStatement ps = connection.prepareStatement("SELECT username, password FROM morleys_user WHERE username=? AND password=? AND isActive=1 LIMIT 1");){
ps.setString(1, username);
ps.setString(2, password);
ResultSet resultSet = ps.executeQuery();
return resultSet.next();
}
}
}
Common connections pools are Apache Commons-DBCP and C3P0.
As managing sql resources can be quite repetitive and cumbersome you might want to consider using a template: for example Spring's JdbcTemplate
Example C3p0 configuration:
public ComboPooledDataSource dataSource(String driver, String url, String username,String password) throws PropertyVetoException {
ComboPooledDataSource dataSource = new ComboPooledDataSource();
dataSource.setDriverClass(driver);
dataSource.setJdbcUrl(url);
dataSource.setUser(username);
dataSource.setPassword(password);
dataSource.setAcquireIncrement(1);
dataSource.setMaxPoolSize(100);
dataSource.setMinPoolSize(1);
dataSource.setInitialPoolSize(1);
dataSource.setMaxIdleTime(300);
dataSource.setMaxConnectionAge(36000);
dataSource.setAcquireRetryAttempts(5);
dataSource.setAcquireRetryDelay(2000);
dataSource.setBreakAfterAcquireFailure(false);
dataSource.setCheckoutTimeout(30000);
dataSource.setPreferredTestQuery("SELECT 1");
dataSource.setIdleConnectionTestPeriod(60);
return dataSource;
}//in order to do a "clean" shutdown you should call datasource.close() when shutting down your web app.
MySQL times out the connection after some period of time. The standard way to deal with this is to use a properly configured connection pool (with a configured DataSource) instead of using DriverManager directly.
The connection pool will check for and discard "stale" connections.
Recently I was looking at the frequent hangs of the tomcat server and came across some exceptions thrown frequently in some part of the code.
When I examined the code, this what it looked like
public static String doSomething() {
String returnVakue = "default value";
try {
ResultSet rs = getMyResultSet();
rs.first();
returnValue = rs.getString("my_field"); // Note that Exception happens at times when the ResultSet is empty
} catch (Exception e) {
throw new Exception(e);
} finally {
return returnValue;
}
}
While I am aware that it is perfectly OK to have both throws exception and return, wondering if this can cause any kind of leaks in tomcat. And Is there any potential risk with performance. ? Hoowever my caller function stops the execution at this point. Any views on this? Does it affect the GC?
EDIT : NOTE : I Know how to correct this code. Please share your views whether this can potentially cause tomcat hanging.
First check if returned ResultSet is empty.
while( rs.next() ) {
// ResultSet processing here
// Result set is not empty
}
In my opinion throwing exception is your decision, but in finally you should be doing clean up e.g. closing Connections.
Open Connections if not closed will cause tomcat to hang because new requests coming to server will be waiting for connections to become available.
Any object in Java which is referenced is not garabage collected, in your case if Connections are not closing then these objects will not be garbage collected.
Cheers !!
If a query takes a long time, not a question of JDBC. The database is responsible. Of course, if JDBC is used properly. In another hand, if you use simple JDBC, it is best that you add a layer DAO in your application.
public class UserDAO extends DAO {
private static final String FIND_BY_PK_SQL =
"SELECT mail, password " +
" FROM user WHERE mail = ?";
public User findByPk(final String mail) throws DAOException {
Connection conn = null;
PreparedStatement ps = null;
ResultSet rs = null;
try {
conn = getConnection();
ps = conn.prepareStatement(FIND_BY_PK_SQL);
ps.setString(1, mail);
rs = ps.executeQuery();
if (rs.next()) {
return fill(rs);
}
return null;
} catch (final SQLException e) {
throw new DAOException(e);
} finally {
DbUtils.closeQuietly(conn, ps, rs);
}
}
private User fill(final ResultSet rs) throws SQLException {
final User user = new User();
user.setMail(rs.getString("mail"));
user.setPassword(rs.getString("password"));
return user;
}
}