Solving "MySQL server has gone away" errors

Question

I have written some code in PHP that returns the html content from .edu domains. A brief introduction is given here: Errors regarding Web Crawler in PHP

The crawler works fine when the number of links to crawl are small (something around 40 URLS) but I am getting "MySQL server has gone away" error after this number.

I am storing html content as longtext in MySQL tables and I am not getting why the error arrives after at least 40-50 insertions.

Any help in this regard is highly appreciated.

Please note that I have already altered the wait_timeout and max_allowed_packet to accomodate my queries and the php code and now I don't know what to do. Please help me in this regard.

Answer 1

You might be inclined to handle this problem by "pinging" the mysql server before a query. This is a bad idea. For more on why, check this SO post: Should I ping mysql server before each query?

The best way to handle the issue is by wrapping queries inside try/catch blocks and catching any database exceptions so that you can handle them appropriately. This is especially important in long running and/or daemon type scripts. So, here's a very basic example using a "connection manager" to control access to DB connections:

class DbPool {

    private $connections = array();

    function addConnection($id, $dsn) {
        $this->connections[$id] = array(
            'dsn' => $dsn,
            'conn' => null
        );
    }

    function getConnection($id) {
        if (!isset($this->connections[$id])) {
            throw new Exception('Invalid DB connection requested');
        } elseif (isset($this->connections[$id]['conn'])) {
            return $this->connections[$id]['conn'];
        } else {
            try {
                // for mysql you need to supply user/pass as well
                $conn = new PDO($dsn);

                // Tell PDO to throw an exception on error
                // (like "MySQL server has gone away")
                $conn->setAttribute(
                    PDO::ATTR_ERRMODE,
                    PDO::ERRMODE_EXCEPTION
                );
                $this->connections[$id]['conn'] = $conn;

                return $conn;
            } catch (PDOException $e) {
                return false;
            }
        }
    }

    function close($id) {
        if (!isset($this->connections[$id])) {
            throw new Exception('Invalid DB connection requested');
        }
        $this->connections[$id]['conn'] = null;
    }


}


class Crawler {

    private $dbPool;

    function __construct(DbPool $dbPool) {
        $this->dbPool = $dbPool;
    }

    function crawl() {
        // craw and store data in $crawledData variable
        $this->save($crawledData);
    }

    function saveData($crawledData) {
        if (!$conn = $this->dbPool->getConnection('write_conn') {
            // doh! couldn't retrieve DB connection ... handle it
        } else {
            try {
                // perform query on the $conn database connection
            } catch (Exception $e) {
                $msg = $e->getMessage();
                if (strstr($msg, 'MySQL server has gone away') {
                    $this->dbPool->close('write_conn');
                    $this->saveData($val);
                } else {
                    // some other error occurred
                }
            }
        }
    }
}

Answer 2

I have another answer that deals with what I think is a similar problem, and it would require a similar answer. Basically, you can use the mysql_ping() function to test the connection before your insert. Before MySQL 5.0.14, mysql_ping() would automatically reconnect the server, but now you have to build your own reconnect logic. Something similar to this should work for you:

function check_dbconn($connection) {
    if (!mysql_ping($connection)) {
        mysql_close($connection);
        $connection = mysql_connect('server', 'username', 'password');
        mysql_select_db('db',$connection);
    } 
    return $connection;
}

foreach($array as $value) {
    $dbconn = check_dbconn($dbconn);
    $sql="insert into collected values('".$value."')";
    $res=mysql_query($sql, $dbconn);
    //then some extra code.
}