I just got a new unit and it came with the new 5.51 rosetta and i am having problems. The first unit i had got to 1% and then it stopped goingup in percent and the time to completion went up, like second by second. So i reset the project. The next unit i got got to about 1.25%, and now it is going up, but very slowly and my to completion time goes up second by second instead of going down.

Please post any advice.

I have the same problem on various workunits, on Ubuntu linux running BOINC v.5.4.11 (the stock BOINC client that comes in the Ubuntu repositories).

Do you have a similar setup by any chance?

- trib'

EDIT: This should probably be in the number crunshing forum, not in here... Anyways...

---

I have the same problem on various workunits, on Ubuntu linux running BOINC v.5.4.11 (the stock BOINC client that comes in the Ubuntu repositories).

Do you have a similar setup by any chance?

- trib'

I'm on a windows xp unit using BOINC 5.8.15, I havn't had a problem but when rosetta 5.51 released, i have this problem

If you're working on a large WU, it can take a long time to update the status. 1% or 1.024% are not actual percentage measurements of the work done. The .024 means that it's at the 24th step in the creation of your first decoy. Once you finish your first decoy, Rosetta gives you an actual estimate based on the time taken to create your first decoy and your run time setting. If your run time setting is 4 hours, and the first decoy took 2 hours, it will state that it's 50% done.

If nothing happens for a sufficiently long time period, the watchdog process will terminate that WU, and start the next WU.

It sounds like you weren't patient enough.

vp10__BOINC_ABRELAX_cterm_hom002__1581_17304 is this the WU you terminated (on the 11th), or is there another one that you aborted today?

---

Rosetta Moderator: Mod.Tymbrimi
ROSETTA@home FAQ
Moderator Contact

It sounds like you weren't patient enough.

vp10__BOINC_ABRELAX_cterm_hom002__1581_17304 is this the WU you terminated (on the 11th), or is there another one that you aborted today?

Yeah, that is the only one that i terminated, i let the second one go and it is going up slowly, but it is working.

Any idea why my time remaining keeps going up instead of down?


Edited to add: It just jumped from like 1.57 to 31.7 or so, so i guess i did jump the gun a bit.

Any idea why my time remaining keeps going up instead of down?

Same reasons it always has. I'm working on building a new FAQ and an attempt to explain this will certainly be a part of it. As you can see, it was the first and second question that feet1st put in to their QA.

Things are running normally. No need to fiddle with it, just observe for a day or so before concluding there is any problem.

---

Rosetta Moderator: Mod.Sense

I just got a new unit and it came with the new 5.51 rosetta and i am having problems. The first unit i had got to 1% and then it stopped goingup in percent and the time to completion went up, like second by second. So i reset the project. The next unit i got got to about 1.25%, and now it is going up, but very slowly and my to completion time goes up second by second instead of going down.

Please post any advice.

The one you aborted was completely successfully by the next guy that got it.

People need to stop micro-managing BOINC and just let it do its job -_-

---

My second wu (1st crashed after opening the graphics) runs since more than 3 hours and sits at 1% too, and the time to completion is going up constantly.

It reminds me SZTAKI a bit, in the sense that it is clear that if your runtime goes up and the percentage stays at 1%, your time to completion has to go up. This is mathematically given, since your boinc manager thinks that the 1% represents a bigger and bigger runtime, thus the 100% has to be bigger and bigger as well. What is likely to happen is that the 1% will jump to 100% at one point and successfully finish the wu. Thus I agree, let it run until it finishes itself or the Rosetta people actually tell us that there is something wrong with 5.51.

P.S: How many steps does a model usually have? I am with the 1st model and 1.41 million steps.

---

1.41 million steps sounds higher then normal. But have no fear, if there is indeed a problem the watchdog is on the scene to take care of things.

B-roy, your computers are hidden, what is the name of the WU you are crunching?? Different types of tasks have different expected number of steps.

Yes the 1% gets recomputed once you complete model number 1. At that point, Rosetta has a good idea how long a model takes your computer to complete, and estimates how that fits in to your runtime preference. It then adjusts the percent complete, and either ends the task or begins model two. If the first model took long enough that you won't be able to complete a second one prior to your preferred runtime, then the task will complete at that point, and yes, the result of that can be a jump from 1% to 100%.

---

Rosetta Moderator: Mod.Sense

the wu finished shortly after my posting, so I guess it reached something like 1.5 mio steps. I also switched my settings to show my computers if you'd like to have a look.

Concerning the credits, I am wondering whether the fixed credit has been changed, as it seems it has been going down. May last wus are (time, claimed, received):

12,440.99 24.61 14.77 (new version)
10,141.14 20.10 16.64
10,549.60 20.91 13.89
10,199.85 20.21 20.86
11,249.42 22.02 18.15

---

Rosetta does not use a fixed credit system. Credit it based on number of models completed and how long it is taking others to compute models on the same set of tasks.

I agree that you seem to have a trend going there where you are granted roughly what you claim or slightly less, and then your latest is a more extreme drop. But one task doesn't make a trend. Credits are still computed the same way as prior task types and Rosetta releases. There is always some variation between different task types and different machine types.

But then again, your third one in the list was not far off in percentage terms either.

---

Rosetta Moderator: Mod.Sense