Since I installed BOINC 5.8.8 there have been two ocassions, that I have seen, that Einstein units have reached 100% and the "to Completion" column is ---, but the Status is still "waiting to run".
It should have uploaded and the status column should be "waiting to Report".

Has anybody else noticed this. So far I have only noticed it on Einstein.
Part of the problem maybe that on this computer Einstein is on a very low resourse share, 4.71% (one hour/day) and usually when the Einstein unit has completed it switches to another project.

Andy

They checkpoint after processing is basically finished but before the clean up operations are done. Since the 5.8.8 version switches at checkpoints this can leave the app sitting at 100% but not finished. It is annoying but not really a bug. This was always possible, but switching at checkpoints makes it more likely.

---

BOINC WIKI

BOINCing since 2002/12/8

Since I installed BOINC 5.8.8 there have been two ocassions, that I have seen, that Einstein units have reached 100% and the "to Completion" column is ---, but the Status is still "waiting to run".
It should have uploaded and the status column should be "waiting to Report".

Has anybody else noticed this. So far I have only noticed it on Einstein.

Yes, Andy, it happens. In last few months I've observed and reported this for Rosetta (the devs fixed it for Ralph 5.41), WCG FAAH and recently Kathryn Marks (5.8.3) and me (5.8.4) reported it on Boinc alpha list for Windows einstein_S5RI version 424 (it was probably introduced in this version, it checkpoints 2-3 x in last few seconds at 100%).

Part of the problem maybe that on this computer Einstein is on a very low resourse share, 4.71% (one hour/day) and usually when the Einstein unit has completed it switches to another project.

This was also my problem. The mentioned apps (when with low share) build here large negative STD during crunching and when they are preempted at 100%, they often stay in memory for the next day or so.

Peter

Since I installed BOINC 5.8.8 there have been two ocassions, that I have seen, that Einstein units have reached 100% and the "to Completion" column is ---, but the Status is still "waiting to run".
It should have uploaded and the status column should be "waiting to Report".

Has anybody else noticed this. So far I have only noticed it on Einstein.

Yes, Andy, it happens. In last few months I've observed and reported this for Rosetta (the devs fixed it for Ralph 5.41), WCG FAAH and recently Kathryn Marks (5.8.3) and me (5.8.4) reported it on Boinc alpha list for Windows einstein_S5RI version 424 (it was probably introduced in this version, it checkpoints 2-3 x in last few seconds at 100%).

Part of the problem maybe that on this computer Einstein is on a very low resourse share, 4.71% (one hour/day) and usually when the Einstein unit has completed it switches to another project.

This was also my problem. The mentioned apps (when with low share) build here large negative STD during crunching and when they are preempted at 100%, they often stay in memory for the next day or so.

Peter

I think your last paragraph is the significant point, about staying in memory. The fact it hasn't uploaded is not a problem, just an observation on my part, that I think should be fixed.
I did post same msg on Einstein so maybe someone will see it and act accordingly.

Andy
edit] or maybe 5.8.9 will fix, I see it is now available and I have downloaded. [/edit

The fact it hasn't uploaded is not a problem, just an observation on my part,

It checkpoints multiple times last (tens of) seconds prior to being finished. In between the last checkpoints, it also compresses some (I assume, it is at least reported by the Linux version) ouptput file, so it possibly finally issues some 1-2 redundant checkpoints in last few seconds of runtime. And then it gets catched by the cogent core client.

[edit]
John Keck seems to say approximately the same:

"The latest version of the Einstein app checkpoints after processing is basically finished but before the clean up operations begin. The 5.8.8 client sees a checkpoint event as an chance to change which app is running. So you will see Einstein tasks stopped at 100% but not finished more often with the new client."

[/edit]

that I think should be fixed. [...] or maybe 5.8.9 will fix, I see it is now available and I have downloaded.

After hearing the symptoms, Boinc devs always agreed on it being an app issue.

Peter

[edit]
John Keck seems to say approximately the same:

"The latest version of the Einstein app checkpoints after processing is basically finished but before the clean up operations begin. The 5.8.8 client sees a checkpoint event as an chance to change which app is running. So you will see Einstein tasks stopped at 100% but not finished more often with the new client."

[/edit]

that I think should be fixed. [...] or maybe 5.8.9 will fix, I see it is now available and I have downloaded.

After hearing the symptoms, Boinc devs always agreed on it being an app issue.

Peter

Saw JK's post, I started that thread also.

And ref BOINC/Project. I get that all the time at work except there it is, software say Hardware problem see them. Go see hardware, Not us man, its those software geeks, go see them. and on and on it goes.

Andy

After hearing the symptoms, Boinc devs always agreed on it being an app issue.

And ref BOINC/Project. I get that all the time at work except there it is, software say Hardware problem see them. Go see hardware, Not us man, its those software geeks, go see them. and on and on it goes.

Sounds familiar :-)

(BTW the same happened when I reported it on WCG forum. Sekerob? ;-) no flame intended

Peter

One spicery - yesterday my Linux host started to crunch Predictor's dTasser_test 0.02 unit. It checkpointed for the first time after using 3:06:50 CPU time and was immediately preempted. Now it's stuck at 99.875% and 0:00:14 ETC :-)

(As the project was out of work for some months, I've set it's ressource share low, now the debts are accordingly deep in hell and it'll take some time.....)

Peter