I have begun to have a lot of units that complete too early and always are labeled 'output file absent'.

I just installed Boinctasks to see if any WU's are finished properly anymore.

I'm changing my cores to process another project although this is likely to be a processor problem. Perhaps 3 years of running 24/7 has taken it's toll.

---

You have several that failed as they were starting that start with a "T" and then a 4 digit number. These are the ones that were built incorrectly as described here.

But those probably didn't display the message about the output file being absent.

I see a couple of others you have that begin with "blind_rhoda_boinc_nmr_" fail:
ERROR: ct == final_atoms
ERROR:: Exit from: ....srccorescoringrms_util.cc line: 475
BOINC:: Error reading and gzipping output datafile: default.out
called boinc_finish

I'll see what I can find out about these.

---

Rosetta Moderator: Mod.Sense

Others are having similar problem with the absent output file error on tasks starting with "blind_". The Project Team has removed the tasks from the queue and is working on correcting them. See James' post here

---

Rosetta Moderator: Mod.Sense

Will there be a news item when everything has been fixed?

---

It doesn't take them long to remove tasks, so by the time we see their post indicating they are removing them, they're usually already gone.

---

Rosetta Moderator: Mod.Sense

So is it my processor now? 2 or 3 tasks already uploaded today, which is not right.

---

The last failures I see from your machine are from April 26. Do you have a machine under another profile as well?

The three from this morning completed normally, with no error about the output file. They did not use the full 24hr runtime your preference would like, because they completed 100 models sooner then that and those tasks were apparently limited to 100 models.

---

Rosetta Moderator: Mod.Sense

I don't know this lingo about "models".

Ploop 15.00 16.47
T0569 15.44 18.42
ProteinG 17.02 2.56

Here are the start and finish times. You are saying these are correct and it's just a coincidence they are so short?

---

I'm just saying that a completion earlier (or later) than the runtime preference is not an error. The tasks are not marked as having any problems. They just finished earlier than you might have expected. So while I would agree that it is "not common", I would not agree that it is "not right" in this case.

The bottom line is that some types of tasks will not run for a full 24hour runtime preference ...at least not on a faster machine.

You can display the task details and down towards the bottom it shows how many models your machine completed on the protein being studied. Unfortunately, on that display they are called "decoys" rather then models. The terms are synonymous for discussion here.

---

Rosetta Moderator: Mod.Sense

Not that it matters to me, but when I clicked Task details, I did not have a "decoys" number there.

Now I found out the place where I can (easily) check if the WU's computed right.

Thanks for your assistance.

---

Seems you were wrong. 2 rejected WU's again.

---

I don't recall saying all tasks run perfectly, nor that all tasks that end early are ending without error, so I'm not certain how you feel I was wrong. I'm simply trying to help you understand how the system works, and what to expect.

Your initial report was of tasks that were failing for everyone. These last two were then sent to another machine to run and completed normally.

I see a large number of restarts in your completed tasks by looking at the task details on the project website and seeing many lines such as this:
[2011- 4-30 4:55:19:] :: BOINC:: Initializing ... ok.

Perhaps your machine is periodically busy with other work? The CPU threshold and keeping tasks in memory were recently discussed here.

---

Rosetta Moderator: Mod.Sense