I don't think the shortened due date is necessarily going to have the desired effect since this project can have very long running jobs. I recently completed a 2-day resend. It was crunching for 38.7 hours. It timed out before my rig finished it and now it is out to another wingman. If it had a 3-day due date, it would be validated now and the 4th computer could be crunching new work rather than wasting time on a WU that could be validated right now.

Cheers

We have been dealing with a large build up of data for mcm1 batches 'in-progress'. This project has larger than average results and is currently using 1.4TB of storage for the in-process batches and growing. As a temporary method to limit the impact of this we did drop the length of time for resends from 30% of original deadline to 20%. However, we have found another way of dealing with the issue so I have just not now restored the deadline back to 30%.

We also found a bug that allowed a resend to be assigned 100% of the original deadline in limited cases. We have fixed that bug. Now if the code determines that a user needs more time than 3 days it will simply not send it to them.

Yup. Corrected above.

Instead of just a fixed percentage, couldn't the deadline be the same as the other outstanding work or 30%, whichever is greater. In the example I have crunched (MCM1_0000216_8191), the other original copy is still out and has 5 days left to go. The resend that I got was for a wingman that errored-out in 2-1/2 days.

Cheers

I think I have to tell why I opened a new thread for this 2days dead line issue.
This is not the first time I have got the 2-days dead line WUs for other projects. However, MCM1 WUs are special. As you answerd to the other question, the runtime of the MCM1s are unpredictable. Some may needs more than 2 days to complete.
In my case, I have never experienced timeout for MCM1 WUs though, I set 1 day buffer on BOINC (This amount of buffer is rather smaller than the others running 24/7.) so very easy to exceed the timeout. As you say, if my rig timeouts, sombody picks it up and finally, some results will be returned though, there could be multiple timeout may take place. It my impact to the project it self. It dosen't make sense.
I have been an adoministrator of a datacenter about 10 years . So, I can understand sahorter timeout saves the disk space thouth, I beleave this is not the right way to solve the disk shortage issue.

Thanks.