MySQL Bugs: #24786: Command timeout does not raise exception, documentation incorrect

Bug #24786	Command timeout does not raise exception, documentation incorrect
Submitted:	4 Dec 2006 4:28	Modified:	11 Feb 2011 10:29
Reporter:	John Vanderburg	Email Updates:
Status:	Closed	Impact on me:	None
Category:	Connector / NET	Severity:	S2 (Serious)
Version:	5.0.2	OS:	Windows (Windows XP SP2)
Assigned to:	Tony Bedford	CPU Architecture:	Any
Tags:	CommandTimeout, Docs

Description:
Documentation 23.2.3.1.11. CommandTimeout states the default CommandTimeout is 0.  In command.cs, set constructor sets commandTimeout=30;

When the command does timeout, it appears that it is not raising the timeout exception.  Instead the connection just "goes away" and the client code hangs.

How to repeat:
Documentation issue: 
Read 23.2.3.1.11 of the documentation then line 68 of connect.cs

Timeout issue:
1. Create a query that will take longer than the CommandTimeout value. 
2. Execute query.
3. Monitor connection w/ MySql Administrator.  At exactly the timeout value, the connection should disappear.
4. Client program should continue to run.

Suggested fix:
Documentation issue:
Either change the default value in the documentaiton or in connect.cs

Timeout issue:
Haven't had a change to debug the connect.cs code, but 
'private void TimeoutExpired(object commandObject)' at line 318 checks the canCancel field.  What happens if the timeout occurs before 'statement.Execute(parameters);' at line 377 completes?  Should 'canCancel = true;' at line 379 be before this? or should TimeoutExpired check something else?

Hi John and thanks for your problem report. Can you please try to repeat with latest version available.

Tonci, thanks for the response.
I believe I already have the latest version.  By the way, here are some details I left out:

Client: Windows XP SP2, .NET 2.0
Server: Linux/Fedora Core 4, MySQL server 5.0.27-standard via TCP/IP

While researching the problem yesterday, I uninstalled the copy of 5.0.2 I had and downloaded and installed new copy from the mysql.com website.  I looked at the install file on both copies and the create date was November 3 2006 9:13 AM.

If there is a newer version, please let me know where I can find it.

I've compiled a debug version using the source and will do some more testing.   I want to make 100% that it isn't a problem with poor exception handling on my end.  I don't think it is but you know how that goes.

I'll let you know what I find out.  It may not be until tomorrow as things are getting busy here.

Thanks again.

John, you can use sources from subversion repository at http://svn.mysql.com/svnpublic/connector-net. You will need SVN tool like TortoiseSVN to get to it. In "branches" you'll find latest 1.0 code and in "trunk" 5.0 code. "Tags" contain released versions.
I'm waiting on your feedback.

Tonci,

Here is what I was able to determine during testing:

1. The TimeoutExpired delegate in command.cs (line 318) is called if the command takes longer than CommandTime to execute.
2. cmd.timedOut is set to true and cmd.Cancel called to kill the query.  At this time you can see the command disappear in the MySql Administrator.
3. After thatI wasn't able to follow the exact sequence of events due to the threading; however, I could tell that no exceptions were being raised after that point.
4. Since no exceptions are raised, the timeout exception at line 393 is never thrown.

Thanks for the information on the sources.  I'll check to see if this has been addressed in the latest development versions.

Hi John. We actually have 2 problems here:
 1. Docs not reflecting commandTimeout defaulting to 30
 2. TimeOut error not thrown

From what I see 
	catch (MySqlException ex)
	{
	// if we caught an exception because of a cancel, then just return null
		if (ex.Number == 1317)
		{
			if (timedOut)
				throw new MySqlException(Resources.Timeout);
			return null;
		}
		throw;
	}
ex.Number == 1317 is never satisfied because connection went dead... Test case is pretty simple, just break execution in debugger or with MessageBox and wait.

I will have to investigate this more.

Tonci:

I exported the sourcs from the trunk of the SVN repository and it appeared that this specific code had not changed.

For some reason I never see any exception raised, so I never got as far as if( ex.Number == 1317) check.

Thanks for your help.  I almost never use CommandTimeout, so I've set it to 0 and am able to use the connector without any issues.  I guess the documentation being wrong turned out to be a good thing in this case, as I wouldn't have found the bug in the first place.

Good luck on finding the bug.  Please let me know if you need any more information.  If I happen to do more debugging and find out anything new I'll let you know.

Thanks agian.

Verified on two points:
   1. Docs not reflecting commandTimeout defaulting to 30
   2. TimeOut error *never* thrown. (ex.Number == 1317) condition is never satisfied because connection went dead, as from what I see in debugger. Test case
is pretty simple, just break execution in debugger or with MessageBox and wait.

Just to add another confirmation on this.  As of 5.0.6 this problem has not been resolved.  I am running Connector/NET on a Windows 2003 Server along with SharePoint 2007.  From what I read, does setting CommandTimeout to 0 bypass the issue?

Hi Randall. It works for me, at least in 5.0.7.

Thanks for your interest in MySQL.

Do you know where OR when I can get 5.0.7?  It hasn't been released yet.

Not able to verify as a bug.  Please be aware that the timeout only occurs when using a non debug build of the library  and that it only occurs when you are connecting to a mysql host of version 5.0 or later.

So this is NOT a bug?  I am connecting to a MySQL 5.0.18 database, so what are my options?  My ASP.NET application is failing because I cannot trap the timeout exception coming from Connector/NET.  I need to be able to handle the timeout exception and display a user-friendly error message.

Randall, I'm still following this report. We'll see how your problem will be resolved.

Randall

I'm saying that I can't reproduce it using 5.0.7 (which should be released in the next day or two).

I got an advanced copy of 5.0.7 and tested my code against it.

According to the thread above, there appears to be a timeout problem.  When I execute a query that exceeds the default 30 sec period, my code traps a MySqlException with number 1028 and a message "Sort aborted".  Not sure that this is the correct error, but I can use it for my purposes.  At least now it is handling an exception.

Randall, this is a rare error and usually caused by no space in TMPDIR or by bad disk or RAM. Can you check on that and get back to us?

Documentation still does not list a default value for default command timeout in the table at http://dev.mysql.com/doc/refman/5.0/en/connector-net-examples-mysqlconnection.html#connect...

command.cs sets 30 as the default

Tony, please document this according to last remark from Scott.

The documentation on this was reorganized and updated some time ago. Default Command Timeout is clearly documented (with default value) on this page:

http://dev.mysql.com/doc/refman/5.1/en/connector-net-connection-options.html