High frequency of client errors
Anonymous
1 Mar 2005 19:12:42 UTC
Topic 12787
(moderation:
> I've currently processed eleven complete work units and of these all but the
> first one are shown as "client errors" with no credit awarded. I'm using BOINC
> 4.24 and I'm splitting time with all the other open BOINC projects. I'm
> beginning to wonder if there's some problem with my set up for Einstein@home.
> Am I just very unlucky or is there a deeper problem? What is likely to cause
> the "client errrors" and is there anything I cn do about it?
If you look at the results on the web, you'll see that stderr_out is showing repeated access violations at a particular address in memory. This is not an address within the Einstein@Home application. So it may be related to a graphics driver misbehaving, bad memory in your system, or some other similar system problem. Any idea what was different for the successful workunit?
Cheers,
Bruce
High frequency of client errors
)
> > Thanks for all that information. I suspect that BOINC 4.24 may be the
> guilty
> > party. I moved on to that version of BOINC on Feb 24th just after
> processing
> > my 1st successful work unit so the circumstantial evidence looks strong
> > against it. Probably I've jumped on to the Beta too early. For what it
> may be
> > worth to anyone by way of information I've downloaded Process Explorer
> as
> > suggested. This is a neat utility by the way, I wish I'd known about it a
> long
> > time ago!
> >
> > I used the suspend project facility of BOINC 4.24 to halt all my other
> BOINC
> > projects and force Einstein@Home to run.
> >
> > It appears that ntdll.dll is the DLL immediately below the exception
> (starts
> > at 0x77F50000)
> >
> > The following is extracted from the Process Explorer saved log with all
> DLL
> > column information selected :-
> >
> > ntdll.dll
> > NT Layer DLL
> > Microsoft Corporation
> > 5.01.2600.1217
> > 0x77F50000
> > 0xA7000
> > 01/05/2003 23:56
> > C:WINDOWSsystem32ntdll.dll
> > 0x77F50000
> >
> > Not sure where to go from here but maybe it should be BOINC 4.19 for a
> > while...
>
> Thanks for the info, it helps. Also, its the Einstein@home application that
> gets the error, not BOINC. BOINC just reports it.
>
> It would be great if the Einstein@home application ran the stackwalker
> automatically, as it would trace the calls from the application up thru the
> bad call to ntdll.dll (the app doesn't make any calls to ntdll.dll directly).
> It should be part of the handling exception, but for some reason its turned
> off. It does intercept the 0xc0000005 exception, as you see the error
> message in the result output.
>
> Maybe the developers can reply with a method for turning stackwalker back on.
Walt, we DO call stackwalker! And we distribute the .pdb file with the app as well. So I'm confused why we don't have a stack trace. Any ideas?
Cheers,
Bruce
> I'll leave the logging on,
)
> I'll leave the logging on, continue with no screen saver, and watch with
> interest to see what happens with future units.
David, thank you. If you are willing to sacrifice your credits 'for the general good of the project' I would be grateful if you continue to work with Walt to track down the source of these errors. In other words Walt might ask you to try turning on graphics again or something else, to try and reproduce the problem again. Understanding what causes the problem and how to fix it might increase the effective number of trouble free host machines by hundreds or thousands, so from the project perspective this would be beneficial.
In any case, please just carry on. Walt will let you know if he wants you to try something.
Bruce
> Yes, that's fine. I'm more
)
> Yes, that's fine. I'm more interested in making a contribution than racking up
> arbitrary credits, so if any of this helps elsewhere, then great. I just
> didn't want to be returning lots of computation errors for no purpose! I'll
> keep an eye on this thread, follow Walt's suggestions and post the results of
> any future trials to see if the problem can be diagnosed.
Thank you!
> I guess we need to wait for
)
> I guess we need to wait for the next release of Boinc or Einstein to have both
> a nice screensaver and to do some usefull work.
The Apps we put on the server work without these problems for 99,5% of the people including the developers who test them beforehand. I've set up quite some systems myself in the last two weeks without observing problems with the screensaver (apart from some experimental code I tested). If there is a problem, it is more likely to be bound to your specific system configuration than to the E@H or BOINC code.
> Is there a way to report bugs, or should we just hope.
We are reading the forums (though might be too busy sometimes for an immediate response).
In this case, however, I don't think that a new App would help. You seem to have a problem with your graphics hard- or software.
BM
BM