Message boards : Number crunching : Problems and Technical Issues with Rosetta@home
Previous · 1 . . . 8 · 9 · 10 · 11 · 12 · 13 · 14 . . . 55 · Next
Author | Message |
---|---|
![]() ![]() Send message Joined: 16 Oct 05 Posts: 711 Credit: 26,694,507 RAC: 0 |
Sorry about the empty queue. It certainly wasn't a planned outage. Fair enough. BUT, please post it on the front page. Not everyone goes digging within the message board to figure out what happened. ![]() |
HiFiTubeGuy![]() Send message Joined: 12 Jan 10 Posts: 22 Credit: 6,291,999 RAC: 0 |
Hopefully it won't be long before new WUs come through, but at times like these it's worth tweaking "Rosetta@home preferences" available from "Your preferences" in Boinc Manager & increasing "Target CPU run time" from 3 hours if you can, both to extend the remaining work you have & to reduce the demand of new WUs when they come back on stream, easing the workload on the servers and ensuring everyone gets back running a little quicker. If you click the 'Projects' tab, then click 'Your Preferences' which will take you to your account page, then click the 'Rosetta@home preferences' link. Then, on your preferences page, click the 'Edit Rosetta@home preferences' link. Select the 'Target CPU run time' you'd like, and click 'Update preferences'. Hope that helps, Kerry ![]() |
Michael G.R. Send message Joined: 11 Nov 05 Posts: 264 Credit: 11,247,510 RAC: 0 |
If you go to the "network preferences" tab, you can set it so that you have an extra buffer of a couple of days. This should mean that you have enough WUs to get through almost any outage without missing a beat. |
![]() ![]() Send message Joined: 30 May 06 Posts: 5690 Credit: 5,859,226 RAC: 381 ![]() |
Probably goes without saying but work came back on stream about 90 minutes after the last post. Desktop happily made it through but a cock-up by me meant I only lost 20 minutes on the laptop, but allowed a bunch of WCG WUs to get downloaded before grabbing from Rosetta again. My mistake but I built up some debt to WCG so best to clear it. No harm done. Rosie's credits are climbing up quite a bit now, but milkyway got swallowed by a black hole. So when it comes back on Rosie is going to have to go sit on the sidelines a bit until things catch up. |
SFCC Send message Joined: 3 Sep 09 Posts: 10 Credit: 227,659 RAC: 0 |
If you click the 'Projects' tab, then click 'Your Preferences' which will take you to your account page, then click the 'Rosetta@home preferences' link. Then, on your preferences page, click the 'Edit Rosetta@home preferences' link. Select the 'Target CPU run time' you'd like, and click 'Update preferences'. OK, you cleared that up for me - thanks. Now my question is: what, exactly, does "target run time" mean and what are the implications of changing it? Thanks. ![]() |
![]() Send message Joined: 3 Nov 05 Posts: 1829 Credit: 119,133,504 RAC: 4,704 ![]() |
If you click the 'Projects' tab, then click 'Your Preferences' which will take you to your account page, then click the 'Rosetta@home preferences' link. Then, on your preferences page, click the 'Edit Rosetta@home preferences' link. Select the 'Target CPU run time' you'd like, and click 'Update preferences'. That's how long Rosetta will aim to run each task for. Lots of models can be run within each task - and Rosetta will fit as many as it can into the target run time, to the nearest model. So, you can set your run-time to 24hrs and you'll get through one task per day, but it will contain lots of models, or you can run lots of tasks, each containing fewer models. Usually, larger tasks mean less bandwidth (although not always the case), but also means more work lost if there is a validation error. Either way, it shouldn't affect your RAC. Just don't go changing it wildly because BOINC uses it to calculate how much work to do and what projects need to run when on your machine - incrementally is the way to go if you want to change it. HTH Danny |
P . P . L . Send message Joined: 20 Aug 06 Posts: 581 Credit: 4,865,274 RAC: 0 |
Hi. Looks like the validator is on a go slow again, i've got tasks sitting for hours not going through. I see the flops estimate on the front page isn't going up either. ![]() |
![]() ![]() Send message Joined: 28 Nov 11 Posts: 3 Credit: 4,593,564 RAC: 0 |
I am new to Rosetta, but am a Computer Scientist of some 40 years experience. My query is as to why I have nine NEARLY complete Rosetta Mini 3.17 tasks that are labelled as "Waiting to run". (BOINC Manager V:6.12.34 x64, wxWidgets ver: 2.8.10) They are indicated, for example, as: 99.309%, 98.749%, etc complete. Yet there are 12 "high priority" tasks executing. (Perhaps I just answered my own question there, "high priority") But who or what gets to decide which tasks should run to completion? The blocked tasks have only minutes to completion, so it make sense to allow them to complete. Workstation: 6 core i7 extreme = 12 threads 24Gb RAM (yes, so don't conjure up the out-of-memory excuse!) 1/4 Tb SSD disk 3Tb mechanical Hard Drive 2 x nVidia GX580 with 3Gb memory each Win 7 Ultimate x64 Is there some way to instruct Rosetta to complete the task no matter what it's priority? Or at least some brief and understandable indication of *why* a task is in the state that it is? If it helps, at the same time as Rosetta, I am running 2 simultaneous GPUGrid tasks on my graphics cards, but these are screaming along flat-out with zero problems. If I could 'bump up' a task priority, that might suffice, but can find no way of achieving this. The absence of 'right-click' menus is a real stumbling block. Philosophy is Bunk - Richard P. Feynman |
![]() Send message Joined: 3 Nov 05 Posts: 1829 Credit: 119,133,504 RAC: 4,704 ![]() |
I am new to Rosetta, but am a Computer Scientist of some 40 years experience. Hi It's BOINC which decides which jobs are high priority and which aren't, and is based on the deadlines. I think the main problem with the scheduling is that there are quite a few variables involved, such as deadlines, multiple projects, credit and debt to the different projects etc. It seems sensible that they should add some logic to get tasks that are near completion finished, but the % complete isn't linear on Rosetta tasks which makes that more difficult. The best rule is to stay clear and let BOINC figure things out - it's not perfect but in my experience it's usually reasonably good at getting things in on time. Nice PC by the way! |
![]() ![]() Send message Joined: 14 Sep 10 Posts: 11 Credit: 2,348,002 RAC: 0 |
Hi all, at this moment all WU pending had 0 granted credit ! I think will be an error on server ! Thanks in advance for your help. I'm a old italian programmer (do you know cards ?). Now, i recycle/repair old pc of my friends, and they revive for research. A long trip begin with a little step ... |
[PST]Howard Send message Joined: 17 Sep 05 Posts: 1 Credit: 201,085 RAC: 0 |
Same here, all validating with zero credits |
Otto Send message Joined: 6 Apr 07 Posts: 27 Credit: 3,567,665 RAC: 0 |
Yep, atm getting no credit for completed WUs. |
![]() ![]() Send message Joined: 28 Nov 11 Posts: 3 Credit: 4,593,564 RAC: 0 |
It's BOINC which decides which jobs are high priority and which aren't, and is based on the deadlines. I think the main problem with the scheduling is that there are quite a few variables involved, such as deadlines, multiple projects, credit and debt to the different projects etc. It seems sensible that they should add some logic to get tasks that are near completion finished, but the % complete isn't linear on Rosetta tasks which makes that more difficult. The best rule is to stay clear and let BOINC figure things out - it's not perfect but in my experience it's usually reasonably good at getting things in on time. Ta. I specialize in design of *true* real-time bespoke embedded operating systems for safety-critical applications, such as defence, medical controllers and fuel & oil refineries, etc. Anything where a program scheduling error can maim or kill one or more humans. From what you have suggested, the issue seems to be a very basic one known as reverse priority inversion, where a task's priority is not 'aged' properly in relation to its resource allocation vs impact ratio. But, as you say, there appears to be nothing about it which I can actively solve, so I must just 'put up with it', unless I become a programmer on the Rosetta team, and that ain't about to happen anytime soon! Philosophy is Bunk - Richard P. Feynman |
![]() Send message Joined: 3 Nov 05 Posts: 1829 Credit: 119,133,504 RAC: 4,704 ![]() |
But, as you say, there appears to be nothing about it which I can actively solve, so I must just 'put up with it', unless I become a programmer on the Rosetta team, and that ain't about to happen anytime soon! You'd have to become a BOINC programmer ;) - BOINC just runs the rosetta app when it deems fit (as far as I'm aware anyway!)... The manual work-around is to hit 'no new tasks' for the project and then suspend all but the tasks you want to run. Then re-enable new tasks once they've completed. I wouldn't recommend it though - you'll still get credited for almost all work submitted after its deadline up to a point with Rosetta. And it's very easy to forget to re-enable new work! |
edikl Send message Joined: 16 Jun 10 Posts: 10 Credit: 186,187 RAC: 0 |
Come on Administrators, give us some explanation on what's going on with Rosetta. I've just got 0 credits for my tasks.. Is it really so difficult for you to write one or two sentences of explanation on Rosetta's main page? I don't believe that any of you doesn't have a minute to do that.. I won't stop crunching, because I do know that in future it will help humanity to fight many deadly diseases. But it's not fair how you treat us.. I am almost sure that my message will not change your attitude towards us at all. At least it's helped me to calm down a bit ;) |
cwood1 Send message Joined: 13 Oct 11 Posts: 1 Credit: 37,198 RAC: 0 |
Same here, and I also notice that "CPU time" is all over the place. Typical should be around 10,800 sec (3hrs), but there are suddenly a lot that are sub-1000 and not listed as errors. Granted credit is often zero, regardless of CPU time. Unrelated issue: many times threads that are supposedly "running" are getting zero CPU time in the Windows task manager. This computer has 4 logical processors and commonly 2-3 tasks get 24-25% and the rest get zero (all take similar amounts of memory). It seems to be totally random, as sometimes 4 processes will take the ~100% processor time. |
ITgreybeard Send message Joined: 19 Sep 07 Posts: 3 Credit: 932,275 RAC: 0 |
My computer is waiting for new tasks but none seem to be coming down the pipe. Is there a communications issue at the server? |
http Send message Joined: 20 Oct 11 Posts: 1 Credit: 7,586 RAC: 0 |
My computer is waiting for new tasks but none seem to be coming down the pipe. Is there a communications issue at the server? It's happening the same to me. |
HW&JC Send message Joined: 2 May 08 Posts: 21 Credit: 7,870,114 RAC: 782 ![]() |
New tasks came down about 30 minutes ago. Hopefully the first of many. See front page & server status pages |
googloo![]() Send message Joined: 15 Sep 06 Posts: 133 Credit: 22,441,278 RAC: 1,505 ![]() |
Work units seem to be flowing again. However, I just got zero credits for a task: 467264044. |
Message boards :
Number crunching :
Problems and Technical Issues with Rosetta@home
©2024 University of Washington
https://www.bakerlab.org