Actually what the problem is that when the WU gets to 10ish minutes to go it actually isn't doing any work it is just stuck. I've had a WU get to the 9 minute 56 second mark and just get stuck there. The longest I've had it get stuck is close to 18 hours and if BOINC get restarted it will go all the way back to 45 minutes and a then only take about 2 1/2 hours to complete after that. And there are alway a lot of lock file errors or watchdog reset errors. And not all have this problem. I've had several Rossetta Mini finish without getting stuck at just under 10 minutes and I can't remember see a beta have problems getting stuck.

The description by Sid Celery sounds very accurate to me. I've currently got a Mini work unit at 6 hours (3 hours default) and as he says it's still on the first model and ticking up nicely. No problem at all.

If you've got lock file errors I'd hazard a guess that it's not ticking up at all on the CPU Time side. That's the issue. Forget anything to do with the remaining time because that's only ever a complete guess - as likely to be wrong as right.

When you get other errors, the WU falls back to its last save position or the start of the current model within the WU. Maybe it sorts itself out by doing that and that's why it completes quickly after that.

Just my 2cents

Add me to the list of folks seeing WUs seemingly hang at around 9ish minutes to go to completion. I've seen WUs run as long as 11 hrs without completing before manually aborting them. Behavior seen on multiple hardware platforms (at least an AMD 9950BE and Opteron 180 and an Intel Core2 Duo dual core laptop with installed memory ranging from 1GB to 3+ GB), but all running WinXP.

Seems to be happening on 5-20% of my Rosetta Mini 1.40 WUs.

I really don't understand why people keep going on about this. It seems quite obvious to me that once the counter gets to around 10 minutes it stops counting altogether. Every WU does this, Mini or Beta. Always has, likely always will.

If the estimate is 3hours then even if the WU ends up running 3hours exactly the countdown still stops with 10 minutes to go. It ends when the model it's working on ends, then drops to zero as the WU finishes altogether. If the 1st model ends at 1h 31m then the WU ends because it'll assume the next model will take the same time and go over the 3hours. If 2 models complete at 2h 1m it'll do the same, assuming another 1h 0m 30s for the next model. And so on for 3 models at 2h 16m, 4 models at 2h 25m etc.

To see how many models have been done, click "Show Graphics" in the Boinc Manager. It's shown at the bottom right.

An estimate is an estimate. It's not a set time frame. Don't expect it to be cast in stone because it's not.

Same with all the long-running WUs. They don't end earlier because the first model hasn't even been completed. Don't look at the clock ticking down. As long as the CPU time is clicking up then it's running just fine. If you abort the WU while CPU time is running then it's your look-out. I think my record is about 14 hours.

Good day all. I've also noticed this problem with version 5.77 but it doesn't happen all of the time only occasionally. When I notice that this is happening I usually suspend that WU and let something else run. The one time I just let it run it was at 09:57 for about 3 hours and the count down timer was not moving. Also, I saw the amount of work being done drop to the 1000th of a percent/sec which was considerably slower than it was for the first 96% of the WU. I didn't want to abort the WU and just hoped that if it did something else for a while and worked it's way back to the WU it would finish it properly. Well, it did take about 1 hour to finish but it was much better than the rate it was moving at. I hope my experience is helpful.

Warren Rogers

Was it that WU which took your computer about 19K seconds? In that case there was nothing wrong, in my opinion. It is just one of those work units which need a large amount of time to generate even one model. And since at least one model needs to be generated, this one model may exceed your pre-set runtime. Maybe the low modelcount also causes the inaccuracies in the estimated time remaining.

Something goes wrong with 5.77 on my machine. It gets down to saying 00:09:57 and then stays there. So for the second time I am about to abort a task. Suspect this old PC just isn't capable or something. Been chugging along with RAH for over a year, I guess, but maybe it's time to quit.

Jerry, ever tried just letting them run? That are probably doing just fine. Rosetta has no way to know ahead of time exactly how long it will take to crunch a given model, so the estimate is... well... just an estimate. Once it gets down to within 10 minutes of your target runtime, it starts to exponentially reduce the time remaining less and less, with the idea being that the time to completion will generally still be going down.

The watch dog is always there looking over your tasks, and prepared to abort them if it deems necessary.