Wow, thanks for the responses guys!

There are plenty of hints here, about increasing result file limits... hope World Community Grid finds the configuration problem and fixes it, so our resources aren't continuing to be wasted.

Hi all.

The problem here is simply that there are a small number of jobs towards the end of what was first sent out that have a large number of electrons. This pretty much equates to having a large number of basis functions, which is what determines the size of the large binary file that BOINC is saying is too big.

In the old style of work unit, it was a calculation in a basically minimal basis set that was getting exposed to this limit - and these calculations have far fewer basis functions per atom than the one we have changed to.

For the vast majority of work units, this file falls under the size limit of BOINC and so was not causing a problem, and frankly I did not expect this to be a problem at all, based on the in-house testing I had done. However, this is research and the unexpected frequently occurs (it is what makes it interesting after all!). Rather than continue to send new jobs, we decided to profile the jobs which BOINC was refusing to send to see if we can allow larger files to be transferred to *minimise* the wasted time on the grid. Because of this, and due to the fact that failed jobs get sent multiple times, the failure of these jobs is more visible since no smaller jobs (for which this would not be a problem) are being sent until we change the settings (which makes sense for diagnosing errors).

Siedentopf: whilst that is a viable theory based upon what has been happening, that is not what the issue is here. As reported, the validation problem was due to two different 'styles' of work unit being on the grid at the same time (something which has not happened at all in the CEP2 until now) and was quickly and completely resolved. With regard to the calculations themselves, I gave a pretty complete summary of why jobs might fail here : https://secure.worldcommunitygrid.org/forums/wcg/viewpostinthread?post=465685 and you can add the current BOINC file size limit to that. It is the complete nature of our type of high throughput screening that some jobs will fail for various reasons, and there is pretty much nothing that can be done to completely remove failures - it is the nature of the quantum calculations. Rest assured we do our very best not to send things to the grid which are flawed or rushed.

Your Harvard CEP Team

We have increased the file size limit for CEP2 and when it starts again this issue should be solved.

Thanks,
armstrdj