Gravitation Wave Computation Errors

Eric
Eric
Joined: 14 Jan 06
Posts: 1
Credit: 1223881
RAC: 0
Topic 196036

I recently started participating in Einstein again. Since then I have processed Binary Radio Pulsar Search, Gravitation Wave S6 GC Search, and the Gamma-ray Pulsar Search. I have yet to complete a Gravitation Wave work unit as every unit I get has computation errors. These units mostly have no processing time but sometimes have a few seconds. I have no problems with either the Binary Pulsar Search, the Gamma-ray Pulsar Search, or anything I get from SETI. I don't know what's wrong or how to fix it and I can't stop accepting Gravitation Wave WU's as the box is grayed out on the preference page. I tried emailing some of the help support people and the second guy told me to post on here so the application developers can have a look at the issue. If anyone know how to fix this I'd like to know so that I can stop inhibiting timely results.

Here's a link to the most recent WU with a computation error. http://einsteinathome.org/task/254732768

I'm running Windows 7 Premium 64 bit with an i7-2630QM 2.0GHz proessor, 8gb of ram, a NVIDIA 560M GTX video card with 3gb of ram. I have BOINC 6.12.34 64 bit.

Thank you.

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4265
Credit: 244921393
RAC: 16900

Gravitation Wave Computation Errors

Sorry, this seems to be an error that occurs more and more frequent recently on Windows 7. The stacktrace points to a problem in the BOINC library. BOINC developers have been contacted, but I didn't get an answer yet.

BM

BM

mxdog
mxdog
Joined: 24 May 06
Posts: 3
Credit: 1552687
RAC: 0

I seem to have the exact same

I seem to have the exact same problem

win7 SP1 X64 , nvidia GT 520 with 4GB

here is a sample unit also

http://einsteinathome.org/task/257515801

I do wonder if it has something to do with the Video card memory or memory size
even though it stress tests fine

Moonwrist
Moonwrist
Joined: 11 Sep 11
Posts: 2
Credit: 8394396
RAC: 0

It seems that I have the same

It seems that I have the same problem.
All 1.01 Gravitation Wave S6 GC Search (SSE2) WUs have status "reported: Computation error (5,)" and have worked less than 10 seconds.

22
22
Joined: 6 Nov 11
Posts: 14
Credit: 757787137
RAC: 3915

what I tried: boinc

what I tried: boinc 6.10.60, boinc 6.12.33, boinc 6.12.34, boinc 6.12.43 some in x64, some in 32bit flavour. none worked.
what I didn't try: boinc 6.13.versions, they seem to be a pain to rollback to 6.12
what worked: installing boinc (6.12.43) to run what boinc calls "protected application execution". drawback: you can't employ your graphics processors for boinc
a guess: one of Microsoft's recent security updates is incapable of handling permissions correctly on GW datafiles because they have 8 letters/figures after the "dot" (*.85_S6GC1) ?

22
22
Joined: 6 Nov 11
Posts: 14
Credit: 757787137
RAC: 3915

Hello mxdog ! You seem to

Hello mxdog !

You seem to have found a way to work around the problem. Can you tell us what you did?

lots of valid GW task since November 20th
http://einsteinathome.org/host/4237982

tia

Stephen Balch 2
Stephen Balch 2
Joined: 16 Jul 08
Posts: 7
Credit: 155909
RAC: 0

I seem to be having the same

I seem to be having the same type of WU failures. I'm running a slightly back level Win Vista (32-bit), on an AMD dual core Turion64 X2 laptop (http://einsteinathome.org/host/1482287) and BOINC 6.12.34 for Windows (the latest version).

Per a suggestion from BOINC, when the errors continued I reset the project, losing several WU in the process, but to no avail. I'm still seeing the "Error while computing" failures on "Gravitational Wave S6 GC search v1.01 (SSE2)" WUs. I do not run any screen savers, BOINC projects or otherwise.

The most recent WUs to fail are Task IDs http://einsteinathome.org/task/257557050 (after 63,397.49 CPU Secs) and http://einsteinathome.org/task/257557049 (after 86,231.68 CPU Secs).

The BOINC log shows:

11/22/2011 21:31:01 | Einstein@Home | Computation for task h1_0393.45_S6GC1__1209_S6BucketA_0 finished
11/22/2011 21:31:01 | Einstein@Home | Output file h1_0393.45_S6GC1__1209_S6BucketA_0_0 for task h1_0393.45_S6GC1__1209_S6BucketA_0 absent

for the most recently failed WU (Task ID 257557050). You can see the STDERR text if you look at the above Task IDs, It's far too long to paste here.

I am not having problems with any other projects, only Einstein Gravity Wave WUs. I currently have one "Gravitational Wave S6 GC search v1.01 (SSE2)" and one "Gamma-ray pulsar search #1 v0.23" in progress, but I've set NNT for Einstein. These things run too long before they fail. I could more productively be doing work for other projects, like LHC, Cosmology, and Rosetta. NNT will remain set for Einstein until some resolution to this issue is found. 60K seconds here, 90K seconds there, sooner or later it adds up to some real hours.

Cheers,
Stephen

(EDIT)P.S. I just checked and while I can opt out of the "Binary Radio" and "Gamma-ray " Pulsar searches, I cannot opt out of the "Gravitational Wave S6 GC" search either, so NNT stays set for Einstein.(/EDIT)

I The perversity of the universe tends to a maximum.
II If something can go wrong, it will.
-- Finagle's First and Second Laws
Join Team Richard Feynman and crunch in memory of the great Physicist and Teacher (and Bongo Player) !!!

Gundolf Jahn
Gundolf Jahn
Joined: 1 Mar 05
Posts: 1079
Credit: 341280
RAC: 0

RE: for the most recently

Quote:
for the most recently failed WU (Task ID 257557050). You can see the STDERR text if you look at the above Task IDs, It's far too long to paste here.


That's right, but I think the interesting part is:

upload failure: 
  h1_0393.45_S6GC1__1209_S6BucketA_0_0
  -161


Gruß,
Gundolf

Computer sind nicht alles im Leben. (Kleiner Scherz)

Stephen Balch 2
Stephen Balch 2
Joined: 16 Jul 08
Posts: 7
Credit: 155909
RAC: 0

RE: That's right, but I

Quote:

That's right, but I think the interesting part is:
upload failure: 
  h1_0393.45_S6GC1__1209_S6BucketA_0_0
  -161


Gruß,
Gundolf

@Gundolf,

Thanks for the response. I had noticed that message too, but I don't know if that refers to my connection to the Internet or to a server connection issue. Quite some time before the Einstein task finished I had suspended BOINCs network activity:

11/22/2011 10:55:20 | LHC@home 1.0 | Scheduler request failed: Couldn't resolve host name
11/22/2011 10:55:50 | | Suspending network activity - user request
...
11/22/2011 21:31:01 | Einstein@Home | Computation for task h1_0393.45_S6GC1__1209_S6BucketA_0 finished
11/22/2011 21:31:01 | Einstein@Home | Output file h1_0393.45_S6GC1__1209_S6BucketA_0_0 for task h1_0393.45_S6GC1__1209_S6BucketA_0 absent
11/22/2011 21:31:02 | Einstein@Home | Restarting task LATeah0053S_736.0_418200_0.0_0 using hsgamma_FGRP1 version 23

Normally, for a communications problem, the transfer waits in a pending state until networking is resumed and then uploads the results, rather than failing the task as indicated in STDERR. At least that's how other projects seem to behave. I think the "file_xfer_error" above may be because of the missing output file. BOINC would certainly have problems transferring a file which wasn't there, or had a zero length.

BTW, I shouldn't be having problems with disk space, there is 8.9 GB free and available to BOINC (BOINC is using 1.1 GB total). Free and not available to BOINC is almost 45 GB. Einstein _is_ using more disk space than any other project with 649.01 MB in use. Also, my computing preferences are set at:

Disk: use at most 100 GB
Disk: leave free at least
Values smaller than 0.001 are ignored 0.001 GB
Disk: use at most 50% of total

My computer is a laptop (which runs almost 24/7) with an air card (a UTStarcom USB Modem) for accessing the Internet. (For those in the US, my carrier is Cricket Witless, er... Wireless. They have much better signal strength _in my apartment_ than my smartphone carrier, T-Mobile, does.) More often than _I_ like, my carrier drops my connection. Although the software automatically tries to reestablish the connection, it doesn't always work as designed. If I know I'm going away for a while, I will set BOINC to "Network activity suspended", otherwise the log gets flooded with "date time | project_name | Scheduler request failed: Couldn't resolve host name" messages. If I'm just around the apartment, I leave it set to "Network activity always available".

Questions?

Cheers,
Stephen

I The perversity of the universe tends to a maximum.
II If something can go wrong, it will.
-- Finagle's First and Second Laws
Join Team Richard Feynman and crunch in memory of the great Physicist and Teacher (and Bongo Player) !!!

22
22
Joined: 6 Nov 11
Posts: 14
Credit: 757787137
RAC: 3915

Hi Stephen, i'm sorry to

Hi Stephen,

i'm sorry to say, you probably have a different problem altogether. Eric's , mxdog's , Moonwrist's and my tasks all error out within seconds. Your tasks seem to have run for hours before they failed. Maybe to do with powersaving or hibernation or ventilation.

On the "instant" errors I have done further experimentation. They went away when I physically removed my Geforce. I ran one task to completion and then reinserted the graphics card for further testing.
http://einsteinathome.org/task/258440731
Changing the driver from 285.79 ß to 270.61 whql didn't help either. Note that I didn't test whether the presence of the graphics device in the PCIe slot/memory space was related to the errors or whether it was a piece of (Nvidia) software that was only loaded when the card was present.

All crunchers above have in common:
Win 7 64 bit + Nvidia graphics + more than 4 GB system RAM

mxdog's 2600k now simultaneously pushes out GW WUs that fail and GW WUs that validate, maybe he has got news for us.
Also read mxdog's post in another thread, obviously on the same issue:
http://einsteinathome.org/node/196074
(I didn't read it properly at first and mistook it as reporting errors with BRP GPU tasks)

Sometimes we have to crunch bugs before we can crunch WUs again :-)

Stephen Balch 2
Stephen Balch 2
Joined: 16 Jul 08
Posts: 7
Credit: 155909
RAC: 0

RE: (SNIP) i'm sorry to

Quote:

(SNIP)

i'm sorry to say, you probably have a different problem altogether. Eric's , mxdog's , Moonwrist's and my tasks all error out within seconds. Your tasks seem to have run for hours before they failed. Maybe to do with powersaving or hibernation or ventilation.

(SNIP)

Also read mxdog's post in another thread, obviously on the same issue:
http://einsteinathome.org/node/196074
(I didn't read it properly at first and mistook it as reporting errors with BRP GPU tasks)

Sometimes we have to crunch bugs before we can crunch WUs again :-)

@Kohler

Hello, and thanks for the response. I don't allow this lap top to sleep/hibernate and it's running at the "High Performance" power setting, rather than "Balanced" or "Power saver", since it's almost always on AC (Mains) power. It doesn't travel much any more since I retired (former IBM OS/MVS/ESA Systems Programmer, specializing in SNA data networks). There shouldn't be any power problems. It will very occasionally shut down to prevent overheating, but it hasn't done that recently since I removed the battery and put the external fan on it.

I was aware of the difference in OS's, and I don't have a GPU to use to run the GPU version of BOINC, or GPU-based project tasks. I missed the "instant" failures part, you're not the only one who misreads messages .

Thanks for the pointer to mxdog's post. I will read it after I post this.

I started with SETI, so I understand "crunching bugs". I've been with Einstein and other BOINC projects since at lest 2005 (originally "2267 Stephen Balch" on Einstein, and this account since 2008 when I lost access to the original accounts, thanks to having my computers stolen, and had to set up this account) and have worked through A LOT of issues with BOINC and the various projects.

*** To everyone who celebrates "Thanks Giving" (an American holiday) today, I wish you a happy and safe holiday with friends and family near, and may your favorite (American) football team(s) win their game(s). ***

Cheers,
Stephen

I The perversity of the universe tends to a maximum.
II If something can go wrong, it will.
-- Finagle's First and Second Laws
Join Team Richard Feynman and crunch in memory of the great Physicist and Teacher (and Bongo Player) !!!

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.