The ODP.NET OracleCommand class has a CommandTimeout property that can be used to enforce a timeout for the execution of a command. This property seems to work in situations where the CommandText is a SQL statement. The example code is used to illustrate this property in action. In the initial version of the code, the CommandTimeout is set to zero - telling ODP.NET not to enforce a timeout.
using System;
using System.Collections.Generic;
using System.Data;
using System.Diagnostics;
using System.Linq;
using System.Text;
using Oracle.DataAccess.Client;
namespace ConsoleApplication3
{
class Program
{
static void Main(string[] args)
{
using (OracleConnection con = new OracleConnection("User ID=xxxx; Password=xxxx; Data Source=xxxx;"))
using (OracleCommand cmd = new OracleCommand())
{
con.Open();
cmd.Connection = con;
Console.WriteLine("Executing Query...");
try
{
cmd.CommandTimeout = 0;
// Data set SQL:
cmd.CommandText = "<some long running SQL statement>";
cmd.CommandType = System.Data.CommandType.Text;
Stopwatch watch1 = Stopwatch.StartNew();
OracleDataReader reader = cmd.ExecuteReader();
watch1.Stop();
Console.WriteLine("Query complete. Execution time: {0} ms", watch1.ElapsedMilliseconds);
int counter = 0;
Stopwatch watch2 = Stopwatch.StartNew();
if (reader.Read()) counter++;
watch2.Stop();
Console.WriteLine("First record read: {0} ms", watch2.ElapsedMilliseconds);
Stopwatch watch3 = Stopwatch.StartNew();
while (reader.Read())
{
counter++;
}
watch3.Stop();
Console.WriteLine("Records 2..n read: {0} ms", watch3.ElapsedMilliseconds);
Console.WriteLine("Records read: {0}", counter);
}
catch (OracleException ex)
{
Console.WriteLine("Exception was thrown: {0}", ex.Message);
}
Console.WriteLine("Press any key to continue...");
Console.Read();
}
}
}
}
Example output for the above code is shown below:
Executing Query...
Query complete. Execution time: 8372 ms
First record read: 3 ms
Records 2..n read: 1222 ms
Records read: 20564
Press any key to continue...
If I change the CommandTimeout to something like 3...
cmd.CommandTimeout = 3;
...then running the same code produces the following output:
Executing Query...
Exception was thrown: ORA-01013: user requested cancel of current operation
Press any key to continue...
Calling a stored procedure that returns a ref cursor is another matter though. Consider the test proc below (purely for test purposes):
PROCEDURE PROC_A(i_sql VARCHAR2, o_cur1 OUT SYS_REFCURSOR)
is
begin
open o_cur1
for
i_sql;
END PROC_A;
The example code below can be used to call the stored proc. Note that it sets the CommandTimeout to a value of 3.
using System;
using System.Collections.Generic;
using System.Data;
using System.Diagnostics;
using System.Linq;
using System.Text;
using Oracle.DataAccess.Client;
namespace ConsoleApplication3
{
class Program
{
static void Main(string[] args)
{
using (OracleConnection con = new OracleConnection("User ID=xxxx; Password=xxxx; Data Source=xxxx;"))
using (OracleCommand cmd = new OracleCommand())
{
con.Open();
cmd.Connection = con;
Console.WriteLine("Executing Query...");
try
{
cmd.CommandTimeout = 3;
string sql = "<some long running sql>";
cmd.CommandText = "PROC_A";
cmd.CommandType = System.Data.CommandType.StoredProcedure;
cmd.Parameters.Add(new OracleParameter("i_sql", OracleDbType.Varchar2) { Direction = ParameterDirection.Input, Value = sql });
cmd.Parameters.Add(new OracleParameter("o_cur1", OracleDbType.RefCursor) { Direction = ParameterDirection.Output });
Stopwatch watch1 = Stopwatch.StartNew();
OracleDataReader reader = cmd.ExecuteReader();
watch1.Stop();
Console.WriteLine("Query complete. Execution time: {0} ms", watch1.ElapsedMilliseconds);
int counter = 0;
Stopwatch watch2 = Stopwatch.StartNew();
if (reader.Read()) counter++;
watch2.Stop();
Console.WriteLine("First record read: {0} ms", watch2.ElapsedMilliseconds);
Stopwatch watch3 = Stopwatch.StartNew();
while (reader.Read())
{
counter++;
}
watch3.Stop();
Console.WriteLine("Records 2..n read: {0} ms", watch3.ElapsedMilliseconds);
Console.WriteLine("Records read: {0}", counter);
}
catch (OracleException ex)
{
Console.WriteLine("Exception was thrown: {0}", ex.Message);
}
Console.WriteLine("Press any key to continue...");
Console.Read();
}
}
}
}
Example output from the code above is shown below:
Executing Query...
Query complete. Execution time: 34 ms
First record read: 8521 ms
Records 2..n read: 1014 ms
Records read: 20564
Press any key to continue...
Note that the execution time is very quick (34 ms) and that a timeout exception was not thrown. The performance that we see here is because the SQL statement for a ref cursor is not executed until the first call to the OracleDataReader.Read method. When the first Read() call is made to read the first record from the refcursor, then the performance hit from the long running query is incurred.
The behavior that I've illustrated means that the OracleCommand.CommandTimeout property can't be used to cancel a long running query associated with a ref cursor. I'm not aware of any property in ODP.NET that can be used to limit the execution time of a ref cursor SQL in this situation. Anyone have any suggestions on how the execution of a long running ref cursor SQL statement could be short circuited after a certain amount of time?
Here's the solution that I ultimately went with. It's just an extension method for the OracleDataReader class. This method has a timeout value and a callback function as parameters. The callback function would typically (if not always) be OracleCommand.Cancel.
namespace ConsoleApplication1
{
public static class OracleDataReaderExtensions
{
public static bool Read(this OracleDataReader reader, int timeout, Action cancellationAction)
{
Task<bool> task = Task<bool>.Factory.StartNew(() =>
{
try
{
return reader.Read();
}
catch (OracleException ex)
{
// When cancellationAction is called below, it will trigger
// an ORA-01013 error in the Read call that is still executing.
// This exception can be ignored as we're handling the situation
// by throwing a TimeoutException.
if (ex.Number == 1013)
{
return false;
}
else
{
throw;
}
}
});
try
{
if (!task.Wait(timeout))
{
// call the cancellation callback function (i.e. OracleCommand.Cancel())
cancellationAction();
// throw an exception to notify calling code that a timeout has occurred
throw new TimeoutException("The OracleDataReader.Read operation has timed-out.");
}
return task.Result;
}
catch (AggregateException ae)
{
throw ae.Flatten();
}
}
}
}
Here's an example of how it can be used.
namespace ConsoleApplication1
{
class Program
{
static string constring = "User ID=xxxx; Password=xxxx; Data Source=xxxx;";
static void Main(string[] args)
{
using (OracleConnection con = new OracleConnection(constring))
using (OracleCommand cmd = new OracleCommand())
{
cmd.Connection = con;
con.Open();
Console.WriteLine("Executing Query...");
string sql = "<some long running sql>";
cmd.CommandText = "PROC_A";
cmd.CommandType = System.Data.CommandType.StoredProcedure;
cmd.Parameters.Add(new OracleParameter("i_sql", OracleDbType.Varchar2) { Direction = ParameterDirection.Input, Value = sql });
cmd.Parameters.Add(new OracleParameter("o_cur1", OracleDbType.RefCursor) { Direction = ParameterDirection.Output });
try
{
// execute command and get reader for ref cursor
OracleDataReader reader = cmd.ExecuteReader(CommandBehavior.CloseConnection);
// read first record; this is where the ref cursor SQL gets evaluated
Console.WriteLine("Reading first record...");
if (reader.Read(3000, cmd.Cancel)) { }
// read remaining records
Console.WriteLine("Reading records 2 to N...");
while (reader.Read(3000, cmd.Cancel)) { }
}
catch (TimeoutException ex)
{
Console.WriteLine("Exception: {0}", ex.Message);
}
Console.WriteLine("Press any key to continue...");
Console.Read();
}
}
}
}
And here's an example of the output.
Executing Query...
Reading first record...
Exception: The OracleDataReader.Read operation has timed-out.
Press any key to continue...
It would appear that you are not the first to ask: https://forums.oracle.com/forums/thread.jspa?threadID=2125208
You could monitor within the loop over reader.Read() the elapsed time and exit the loop. That's nice and simple but of course it's only going to be able to exit after a potentially long running call to Read finishes.
Your best bet would probably be to do the loop within a Task on a separate thread, monitor it, and then call cmd.Cancel on the original thread:
[Test]
public void TimeBasicSql()
{
using (OracleConnection con = new OracleConnection("User ID=id; Password=pass; Data Source=db;"))
using (OracleCommand cmd = new OracleCommand())
{
con.Open();
cmd.Connection = con;
Console.WriteLine("Executing Query...");
try
{
cmd.CommandTimeout = 1;
String sql = "begin open :o_cur1 for select count(*) from all_objects, all_objects; end;";
cmd.CommandText = sql;
cmd.Parameters.Add(new OracleParameter("o_cur1", OracleDbType.RefCursor) { Direction = ParameterDirection.Output });
var task = System.Threading.Tasks.Task.Factory.StartNew(() =>
{
try
{
Stopwatch watch1 = Stopwatch.StartNew();
OracleDataReader reader = cmd.ExecuteReader();
watch1.Stop();
Console.WriteLine("Query complete. Execution time: {0} ms", watch1.ElapsedMilliseconds);
int counter = 0;
Stopwatch watch2 = Stopwatch.StartNew();
if (reader.Read()) counter++;
watch2.Stop();
Console.WriteLine("First record read: {0} ms", watch2.ElapsedMilliseconds);
Stopwatch watch3 = Stopwatch.StartNew();
while (reader.Read())
{
counter++;
}
watch3.Stop();
Console.WriteLine("Records 2..n read: {0} ms", watch3.ElapsedMilliseconds);
Console.WriteLine("Records read: {0}", counter);
}
catch (OracleException ex)
{
Console.WriteLine("Exception was thrown: {0}", ex);
}
});
if (!task.Wait(cmd.CommandTimeout * 1000))
{
Console.WriteLine("Timeout exceeded. Cancelling...");
cmd.Cancel();
}
}
catch (OracleException ex)
{
Console.WriteLine("Exception was thrown: {0}", ex);
}
}
It's worth noting that the ORA-01013 exception is thrown on the worker thread and not on the thread calling OracleCommand.Cancel.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With