Hello WCG.

Attention: uplinger -- WCG Tech
cross-reference: my [Dec 28, 2009 6:34:05 PM] post

From the latest batch of WCG WUs I have uploaded (about 12.5hrs earlier) to WCG, there was only one WU (with 129 mixed-project WCG WUs as trouble-free) with an error -- an HPF2 project WU, "nb947_00022_4" with details as follows:

Boinc_v6.2.28 display
----------------------
Name: nb947_00022_4
CPU time: 03:15:41
Progress: 100%
Report deadline: 2010.02.01.Mon 23:27:37
Status: Computation error

Snippets from "stdoutdae.txt":
------------------------------
28-Jan-2010 23:38:56 [World Community Grid] Starting nb947_00022_4
28-Jan-2010 23:38:56 [World Community Grid] Starting task nb947_00022_4 using hpf2 version 603
29-Jan-2010 03:02:23 [World Community Grid] Computation for task nb947_00022_4 finished
29-Jan-2010 03:02:23 [World Community Grid] Output file nb947_00022_4_0 for task nb947_00022_4 absent

Others:
--------
OS: 32-bit Vista Ultimate; SP2

The earlier post I did (Dec28,2009) was about some HPF2 WUs with stuck/frozen progress, else consumes unusually long crunchTimes (~30hrs, which, at that time, I opted to abort). Would the codebase of those HPF2 WUs be the same as those HPF2 WUs that also (sometimes) exhibit the above-mentioned error? If so, would there possibly be some connection between a suspect HPF2 WU 'getting lost' in a non-convergence zone on one extreme (exhibiting stuck/frozen progress), and on the other extreme, the WU somehow 'trapped' (which triggers an error) in the non-convergence?

Good day.

I'm also running HPF2 on my older machine with an AMD Athlon XP 2400+ processor running at 2Ghz with 2 GB of RAM and 32-bit XP Home, SP-3. HPF2 has been running exclusively on both machines for quite some time now with absolutely No Errors returned. I also have BOINC rel 6.10.18 on the older computer.

I'm no expert on such matters, but can't help but wonder if it's somehow a combination of OS (and perhaps the version thereof), processor make and model, version of BOINC running the jobs and possibly other software that may be running on the various machines at the same time HPF2 is running.

Hello WCG.

The 'error phenomenon' surrounding WCG's HPF2 WUs seem to have, as I see it, affected a number of WCG crunchers. I thought I would provide some feedback to the WCG community regarding the crunching of HPF2 WUs in my machine in my hope that some useful data may be extracted therefrom that may serve as sign-posts in the search for a solution. Thus..

I just had one HPF2 WU (nc519_00084_13) whose progress was stuck for some time (since when, I don't know), and when I restarted my BOINC_v6.2.28, the said WU resumed incrementing its progress. My image-capture of the said WU in BOINC shows:
-- 04:54:42 and counting up (CPU time)
-- 46.125% and stuck (Progress)
-- 04:42:02 and counting up unevenly/irregularly (To competion)

After waiting for about 30minutes, with the progress still stuck at 46.125%, I opted to restart BOINC. Some minutes after that BOINC restart, results are as follows:
-- 02:01:49 and counting up (CPU time)
-- 51.67% and counting up (Progress)
-- 02:39:43 and counting down (To completion)

Finally, the said HPF2 WU completed error-free with a BOINC-indicated CPU time of -- 04:07:56.

Good day.
;

ZoSo, we've been battling with a numb ax on this for longer... add the number of concurrent HPF2 jobs to the variations v.v. multicores :-|

Quite a few would like to see an option Never send me this science in combo with the Send me something else if you don't have my fav project. BUT, the techs are working on a process that will remember if a client has continuous problems with a specific science and then will only send 1 periodically to check if the issue was fixed, if asked for of course.

edit: inserted continuous

Hello WCG.

Reference:
-- uplinger's [Feb 1, 2010 9:04:35 PM] post
-- Sekerob's [Feb 1, 2010 10:59:09 AM] post

Gentlemen:

I commend your efforts in dealing with the HPF2 issue. Sekerob's idea proposes to address those concerns that emphasize getting points off crunching WUs while uplinger's idea addresses the nuts-and-bolts of HPF2 WU processing itself with a view of hunting down the source of the issue and with that, proposes to address concerns that emphasize the importance of HPF2 WUs; that is, for crunchers who decided to stick with HPF2 WUs (because of the importance of the underlying science) despite the relatively few errors that may arise crunching them.

P.S. To this hour, I have finshed crunching 14 WCG HPF2 WUs, averaging 243_minutes-per-HPF2-WU. No HPF2 problems thus far (since my last report).

Good day.
;