I am writing a multi-threaded application in C++ using Boost threads (pthread). The application spawns 100 threads and each thread does the following task (I am writing a code snippet that will be running in each thread):
try {
driver = get_driver_instance();
con = driver->connect(SettingsClass.HostName, \
SettingsClass.UserName,SettingsClass.Password);
// SettingsClass is a global static class whose members
// (HostName, UserName, Password, etc) are initialized once
// before *any* thread is created.
con->setSchema("MyDatabase");
driver->threadInit();
string dbQuery = "select A, B, C from XYZTable where D=?";
prepStmt = con->prepareStatement(dbQuery);
prepStmt->setInt(1, 1);
rSet = prepStmt->executeQuery();
/* Do Something With rSet, the result set */
delete rSet;
delete prepStmt;
if (con != NULL && !con->isClosed()) {
con -> close();
driver->threadEnd();
delete con;
}
catch (SQLException &e)
{
/* Log Exception */
}
On running the process (the app, as earlier mentioned, i.e. with 100 such threads), I attach gdb midway and observe that more than 40% of the threads have hanged in the read() call. All the backtraces have mysql library functions (vio_read(), etc) and none are from my code as my code does not perform any I/O.
Could anyone point out why is this issue arising. Should I check my code / network or MySQL server configuration? Have I used the C++ connector library properly?