Folding Forum

I've never seen this before, and TBH looking at the logs to figure out what happened is making my head hurt. I think I now understand it and it was caused by an error on the finishing work unit as it completed.

Somehow I had a work unit download just as another work unit was starting to fold. The end result was the second unit being on hand and downloaded while the first chugged away. I considered dumping the second one, but decided to just let it sit and see if it would run. After the first one finished and sent, it picked up the second one fine, folded it and returned it properly.

I did set folding to "finish" after seeing two on hand, and didn't hit "fold" until after the first one returned, just out of concern the pattern might repeat.


It even confused HFM, and one of them slipped through the cracks and never showed up. The status bar was still showing 100% on the one that failed, and never picked up the one that followed. But both did return and give credit.

It sometimes happen if you are unlucky and get and error while computing the last checkpoint or finalizing the WU ... so you end end up with the client starting the new WU while one is still there almost finished. Both WU should be completed after the client finishes the new one ...

toTOW, that is what happened. As it was finalizing it tossed an error, downloaded another WU, finished it, downloaded another, finished the one that tossed the error, then went to the latest one downloaded.

Had it happened overnight I might not have even noticed. I just saw one on the advanced control showing as "ready" for quite some time and investigated. But to compliment the software, though it is a glitch, it is one that self corrected and allowed things to continue with no issue.

If I recall correctly, one of the updates to the v7 client included code for this particular corner case. Before that the double download would sometimes leave the client with on extra WU until the client was set to Finish. A new WU would be downloaded each time one finished, leaving one in the queue not being processed. Changing the way the finishing code worked in the folding core and the wrapper code would have required more work.

Joe,

That's interesting to know. It took three years of folding to show up on my machine, so I assume it's a fairly rare bug regardless, but glad to see they fixed what they could. I did use the "finish" command when the first of two on hand was folding, just in case. The other waiting unit on hand was lengthy and I didn't want to risk either having to dump it or waiting 6-7 hours for the work unit to finish.

It happened again. At least this time I know what causes it and what to expect.

Since the current WU only needs a couple hours, I'll just let it fold and pick up the second one. Better to not take the hit on the completion rate vs speed things up just a little tiny bit. I'll take the points hit and confirm that it can be cleared up.