All this musing over runtimes has led me to another, possibly related question for Mod.Sense. If the preferred runtimes are shortened either by the cruncher or the project what happens if the current running task is mid-model?


Suppose a task reaches a checkpoint mid-model and then is stopped for task switching or because the computer is turned off. Upon restarting it learns that it has overshot it's new preferred runtime. If it was still on the first model it would continue on but what if it was in the middle of a later model? Does it have instructions to go ahead and finish that extra model or does it quit mid-model? Not only would quitting mid-model look like a premature exit it would make the credit claim larger (and the grant lower) than otherwise expected, an issue addressed in another thread. Of course this would only explain a one time event unless the cruncher or the project is repeatedly making changes to the preferred runtime.

Snags

All this musing over runtimes has led me to another, possibly related question for Mod.Sense. If the preferred runtimes are shortened either by the cruncher or the project what happens if the current running task is mid-model?


Suppose a task reaches a checkpoint mid-model and then is stopped for task switching or because the computer is turned off. Upon restarting it learns that it has overshot it's new preferred runtime. If it was still on the first model it would continue on but what if it was in the middle of a later model? Does it have instructions to go ahead and finish that extra model or does it quit mid-model? Not only would quitting mid-model look like a premature exit it would make the credit claim larger (and the grant lower) than otherwise expected, an issue addressed in another thread. Of course this would only explain a one time event unless the cruncher or the project is repeatedly making changes to the preferred runtime.

Snags

It would quit mid-model and report the complete decoys as you suggest...

---

I'll get a new batch of Rosetta tasks and watch things more closely.

Task 187847619 ended after 1:50 hours after computer and task restarted. Predicted run time was about 2:50 hours (from memory) which is normal for me.

From watching the graphic before shut down, the task was at least on its tenth model (maybe eleventh). In the results it says the task completed one decoy (which I guess equals models?).

Rosetta didn't used to do this; I think its a bug.