Question

I’m designing a small program whose objective is to make sure one of our server is up and running. In some cases, the server won’t answer and a script must be launched to restart the server.

First I’m starting a new thread responsible of fetching the information, which is then joined by the main thread for a certain time span. I then abort the thread to disconnect and finally join it to leave enough time to execute the catch and finally blocks.

In theory it should work great : if the time span is short enough, it indeed indicates that the server is down (because no connection could be made in the short allocated timespan). But in certain cases, when the server is really down, the program will just keep executing, as if the ThreadAbortException had no effect. Problem is, these downtimes are quite sporadic so I couldn’t debug it myself to see what was not working properly.

Here’s how it goes :

This here is the main thread, calling the worker thread. Very straightforward.

public void LaunchCommand()
{
    Thread pingThread = new Thread(new ThreadStart(Ping));
    pingThread.Start();
    while (!pingThread.IsAlive);
    pingThread.Join(new TimeSpan(0, 0, _maxTime)); 
    pingThread.Abort(); //  Time's up.
    pingThread.Join();  //  Make sure we complete everything before moving on
}

And here’s the called thread :

private void Ping()
{
    try
    {
        Stopwatch stopwatch = new Stopwatch();
        stopwatch.Start();

        serviceType = Type.GetTypeFromProgID(serviceProgID, _server, true);
        service = Activator.CreateInstance(serviceType);
        _xmlResult = ApxServiceType.InvokeMember("ExecuteXML", BindingFlags.InvokeMethod, null, service, new string[] { _dataset, Command, string.Empty }) as string;

        stopwatch.Stop();
        _latency = stopwatch.Elapsed;

        // Trivial validations to make sure _status is true, such as _xmlResult.Contains(certainSubString); and such

        _status = true;     //  Everything seems to work fine if we could make up to here.
    }
    catch (ThreadAbortException)
    {
        Console.WriteLine("Server timeout :(");
        return;
    }
    catch (Exception e)
    {
        Console.WriteLine("Server exception: " + e.Message);
        return;
    }
    finally
    {
        if (!_status)
        {
            _latency = new TimeSpan(0, 0, _maxTime);
        }
    }
}

Variables such as Commands, serviceProgID, etc. have been declared elsewhere and are known to work well. I guess my problem spans from the three lines following the stopwatch declaration/initialization. First, I must say I copy pasted these lines from a similar application, but basically it should only fetch a result from the given Command. Unfortunately, because I couldn’t debug under the critical situation, I don’t which line is problematic, but anyway, it seems the ThreadAbortException has no effect. Is it because code has been switched off to unmanaged?

I’m lost here so any idea would be welcomed! Thanks!

Answer 1

I actually don’t see a need to use a separate thread for your scenario. You may check for server availability using synchronous operation. Does the function being used to check server availability offer timeout option? If yes then the timeout option should be enough. If the function returns before the timeout then it means server is online otherwise server is down.

If the function for checking server availability does not offer timeout option and may cause the thread to wait forever (or for a long time) then you may use a new thread. However if Thread.Join(Timeout) method returns true and _status variable is also true then you can be certain that the server is online. If Thread.Join(Timeout) returns false then it would mean that the server is down.

As a good practice, you should use Thread.ResetAbort in the catch block handling Abort exception. Otherwise runtime will rethrow the ThreadAbort exception once catch block finishes execution.