Hi stupidname,

All jobs get an estimated number of iops/flops when send to a client. On basis of these estimates, the client computes the expected duration based on passed historic performance parameters. Problem is, non-deterministic calculations can end up doing many more than expected, extending the duration extensively. 4-6 hours for FA@H and HPF2 sound quite reasonable for the present work as they have more attempts and seeds packed since early this week. GC sizes remain unchanged, but fluctuate quite a bit in duration based on complexity.

Your preferences are not being changed and if they were, that could have been a program fluke, but given that there are several places to select choices - My Projects (for overall control) and Device Profile (for client level control), it's probable you looked at the other.

Even the long running jobs are fine 99,999%, it's the patience being tested very hard. Aborting a job does not help anyone as you loose the computing time and the WCG technicians will not know why it happened. We want you to post the BOINC message log that show any anomaly. Also we'd like you to look at the "Result Status" page and click on the suspect jobs Work Unit Name (first column) to see the quorum. If any other jobs were returned of the standard set of 3 (all projects except HPF2 which has standard 19 presently), and they were returned without error, it's more likely a random or local problem arose. Again WCG technicians like to see a description and a piece of the message log and client information like if the CPU is/was clocking up time for the job. What project running 40 hours and how far had the job progressed and were graphics viewed during any time?

And yes, FA@H and HPF2

Sekerob

PS Where normally HPF2 runs 6+ hours on my C2D, yesterday had a few which did 4.5 hours..... no duration is fixed.

.... forgot about that one , but only if HDC was an exclusive project!

Added: About 2,000 members had that profile condition.

Hello stupidname,
That will not be easy unless you use BOINC Manager. Otherwise, all you will see is a number of same-name application programs. The easy way to see what went5 wrong is to read the error messages stored in the 'Results Status' page with each result.

Do it the easy way. If it has been running too long, abort it and then look at the 'Results Status' page to see if anybody else in your quorum had that problem.

Lawrence

Hello twilyth,
Regardless of the official minimum requirements, I would only run the project with the shortest time (Genome Comparison). HCMD is working through really big proteins now, and will continue to do so until the next phase, which is months away. Even so, it sounds as though you will get your current work unit done in time, whatever project it is for.

Lawrence

I think so-far one (female) indicated to be perfectly okay with the extreme long run times...... don't you love the satisfaction to work thru a 100 hour unit, hibernate, resume, hibernate, resume, and that enormous flush on the 7th day leaving all your team mates in the exhaust fumes as you're speeding by riding that ol P3

BTW Lawrence, south paws first check the Result Status page of BOINC to see if others returned a WU without error and how long it took them. If I see others delivering with Pending Validation, I'm running on.

cheers

PS, my slowest/longest/toughest record stands at 110 hours on UD agent for a HCMD.... did an occasional peak to see in the graphics screen if percentages moved. There you can see even 0.1%. With BOINC one even sees 0.001% in the Tasks Tab section.