Hi every body,

Sorry if the question has been ready answered but i didn't find it in this board.
Usually, a WU take me 3 hours to complet and i gain beetween 40 and 60 points.
Since last week, i get a lot of WU that take 7 hours, sometime more and i gain just 20.00 points ?
Is there a problem with Rosetta WU ?
Is there something that has been changed ?

---

Rejoignez nous sur http://www.grid-france.fr

No reply. Superb.
And the problem continue.
Ok, no problem for me. I stop crunching for Rosetta until this problem is solve.

---

Rejoignez nous sur http://www.grid-france.fr

I can only see one task on any of your computers that has been granted 20 credits at the moment.

If a work unit is very complex then it might take a long time (e.g. 7hrs) to complete one model. Rosetta packages as many models into a Work Unit as it can fit, but obviously 1 is the minimum, so sometimes it will run for longer than your preference, but you will, on average, get appropriate credit for the time taken. On average... The RAC for my two Q6600 PCs is generally very similar, although the credit per work-unit can vary.

---

Same issue reported here. No checkpointing, watchdog cuts in at runtime+4 hours, only 20 credits awarded

---

Same issue reported here. No checkpointing, watchdog cuts in at runtime+4 hours, only 20 credits awarded

So the solution/workaround is to increase your run time so the watchdog is less likely to kick in, especially on slower computers as they're less likely to complete one model prior to the watchdog's action.

---

I have 2 tasks that ran for 7 hours getting 20 points and had some more in the past few days. I cancelled some tasks that were going to do the same. I finish the tasks that i have downloaded and will see it the problem is still there. I will then decide after if i stop this project for the moment or not.

---

Rejoignez nous sur http://www.grid-france.fr

Same issue reported here. No checkpointing, watchdog cuts in at runtime+4 hours, only 20 credits awarded

So the solution/workaround is to increase your run time so the watchdog is less likely to kick in, especially on slower computers as they're less likely to complete one model prior to the watchdog's action.

I don't believe so. That's a comparatively slow machine, but the lack of any checkpointing kind of indicates it would just take longer to hit the wall and still only be awarded 20 credits

---

All tasks checkpoint when a model is completed. And no task is complete until at least one model is completed. Some tasks are able to checkpoint more often than only at the end of a model. And some tasks complete models very quickly and others may take a couple of hours per model. Some models within a task vary and take an exceptionally long time.

If a task goes for more than 4 hours beyond the runtime preference, it is ended for you by the watchdog and credit is granted to you for your help in better understanding what is running long.

If a machine is slow, it takes more hours to complete a model or reach a checkpoint. Increasing your runtime preference (in the Rosetta-specific preferences area here on the website) would allow more time for normal completion of any given model and avoid the watchdog being what causes a task to be returned as completed.

---

Rosetta Moderator: Mod.Sense

All tasks checkpoint when a model is completed. And no task is complete until at least one model is completed. Some tasks are able to checkpoint more often than only at the end of a model. And some tasks complete models very quickly and others may take a couple of hours per model. Some models within a task vary and take an exceptionally long time.

If a task goes for more than 4 hours beyond the runtime preference, it is ended for you by the watchdog and credit is granted to you for your help in better understanding what is running long.

If a machine is slow, it takes more hours to complete a model or reach a checkpoint. Increasing your runtime preference (in the Rosetta-specific preferences area here on the website) would allow more time for normal completion of any given model and avoid the watchdog being what causes a task to be returned as completed.

I do get all that, but the machine was running numerous other tasks and checkpointing frequently on a 4-hour runtime, except for specific tasks which hadn't checkpointed at all over 8 hours before the watchdog cut in. The 20.00 credits seemed to just be a default amount for that eventuality.

---