First Invalid result... why now?


log in

Advanced search

Message boards : Number crunching : First Invalid result... why now?

Author Message
Profile Rae of Quakesville,CHCH NZ
Avatar
Send message
Joined: 30 Nov 01
Posts: 56
Credit: 195,657
RAC: 0
New Zealand
Message 1385944 - Posted: 30 Jun 2013, 0:27:01 UTC

Hello. Today I see my first invalid result that I have ever generated. yet my wingman comes up as inconclusive? Task details are in the link bellow.

http://setiathome.berkeley.edu/result.php?resultid=3045784574

The task has now been sent to a 3rd person. What if their result matches my "invalid" 1?

since finding this invalid result I have emptied my cache & reset the project.
I'm running Bonic on vista home basic with 2 gig of ram <thats the most the motherboard will support> & just under 5 gig virtual.
____________
A city destroyed by an earthquake is an opportunity to Rebuild, redeign & make it a better place to be. Better, stronger, faster like the 6 Million Dollar Man

Profile James Sotherden
Avatar
Send message
Joined: 16 May 99
Posts: 8904
Credit: 35,822,212
RAC: 42,790
United States
Message 1385968 - Posted: 30 Jun 2013, 2:26:14 UTC
Last modified: 30 Jun 2013, 2:27:25 UTC

http://setiathome.berkeley.edu/result.php?resultid=3045784574

Made your link active.

Kind of drastic to empty your cache for one invalid isnt it? I see it was an exit 0 and restarted at 45%. I dont have a clue as to what happend, But im sure someone will come along and say what it was.

Invailds and errors happen. As long as you dont do hundreds or thousands like some folks, I would not worry about one.
____________

Old James

Profile Donald L. JohnsonProject donor
Avatar
Send message
Joined: 5 Aug 02
Posts: 6257
Credit: 735,397
RAC: 1,139
United States
Message 1386001 - Posted: 30 Jun 2013, 3:54:31 UTC - in response to Message 1385968.
Last modified: 30 Jun 2013, 4:02:32 UTC

http://setiathome.berkeley.edu/result.php?resultid=3045784574

Made your link active.

Kind of drastic to empty your cache for one invalid isnt it? I see it was an exit 0 and restarted at 45%. I dont have a clue as to what happend, But im sure someone will come along and say what it was.

Invailds and errors happen. As long as you dont do hundreds or thousands like some folks, I would not worry about one.

I THINK exit state 0 just means it's successfully completed.

What jumped out at me is the lack of a results table on the stderr.txt
The "inconclusive" got a -9 overflow, but the "invalid" doesn't even show zeros. Maybe the result file got damaged in transit between NZ and Berkeley.
____________
Donald
Infernal Optimist / Submariner, retired

Profile James Sotherden
Avatar
Send message
Joined: 16 May 99
Posts: 8904
Credit: 35,822,212
RAC: 42,790
United States
Message 1386008 - Posted: 30 Jun 2013, 4:09:54 UTC

I see he has two errors also. to many exits.

____________

Old James

Profile Donald L. JohnsonProject donor
Avatar
Send message
Joined: 5 Aug 02
Posts: 6257
Credit: 735,397
RAC: 1,139
United States
Message 1386010 - Posted: 30 Jun 2013, 4:14:58 UTC - in response to Message 1386008.

I see he has two errors also. to many exits.

Rae, with -226 Too Many Exits errors, you probably need to look at this thread.
____________
Donald
Infernal Optimist / Submariner, retired

Profile Jeff Buck
Send message
Joined: 11 Feb 00
Posts: 345
Credit: 36,026,127
RAC: 95,646
United States
Message 1386012 - Posted: 30 Jun 2013, 4:21:01 UTC - in response to Message 1386008.

I see he has two errors also. to many exits.

That's because of this ongoing problem, not the host's fault:

http://setiathome.berkeley.edu/forum_thread.php?id=71953

For reasons I still can't fathom, they still insist on sending these v7 cuda22 and cuda23 tasks out en masse, with almost no hope of successful completion. I think they may be quietly trying to work on a fix (I see some intriguing v6 6.11 and 6.12 tasks showing up, and also dying, on some hosts), but in the meantime, they just keep wasting resources!

Profile Rae of Quakesville,CHCH NZ
Avatar
Send message
Joined: 30 Nov 01
Posts: 56
Credit: 195,657
RAC: 0
New Zealand
Message 1386071 - Posted: 30 Jun 2013, 8:43:46 UTC - in response to Message 1386001.

I had already run the cxache dry before the reset, so I didn't ditch any other workunits prior to the reset. I was kinda hoping that the reset might clear away any cuda 2.3 files that might have been lurking on my desktop, since they seam to spit the dummy within a few seconds of starting.
____________
A city destroyed by an earthquake is an opportunity to Rebuild, redeign & make it a better place to be. Better, stronger, faster like the 6 Million Dollar Man

rob smithProject donor
Volunteer tester
Send message
Joined: 7 Mar 03
Posts: 8526
Credit: 59,115,299
RAC: 78,786
United Kingdom
Message 1386076 - Posted: 30 Jun 2013, 8:52:18 UTC

Jeff, they are not sending out "Cuda22" or "Cuda23" tasks - they are assigning tasks to a particular processor, or way of processing (version of Cuda). It was explained some time ago that there were trials on-going with the various Cuda versions to work out exactly which ones were correct for particular processor combinations and that this would run for a few weeks (or months) depending on how clear the results were. Just now there is an advantage in running optimised applications - you effectively say "Only send me tasks assigned to Cuda version x".
____________
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Josef W. SegurProject donor
Volunteer developer
Volunteer tester
Send message
Joined: 30 Oct 99
Posts: 4299
Credit: 1,068,158
RAC: 1,012
United States
Message 1386131 - Posted: 30 Jun 2013, 15:34:38 UTC - in response to Message 1385944.

Hello. Today I see my first invalid result that I have ever generated. yet my wingman comes up as inconclusive? Task details are in the link bellow.

http://setiathome.berkeley.edu/result.php?resultid=3045784574
...

The reason it was marked invalid is an unfortunate conjunction of two factors.

The Validator code had to be revised to protect against users trying to process S@H v7 work with older applications which don't perform the Autocorr search. The revised code marks v7 results invalid if there's no Autocorr information in the uploaded result file, except for a result_overflow task.

The BOINC client has some flaw which makes it sometimes fail to send back the stderr produced by science applications, or truncate it. It did that to your task report, so the stderr didn't contain the result_overflow keyword.
Joe

Juha
Volunteer tester
Send message
Joined: 7 Mar 04
Posts: 184
Credit: 143,295
RAC: 42
Finland
Message 1386240 - Posted: 30 Jun 2013, 23:18:51 UTC - in response to Message 1386012.
Last modified: 30 Jun 2013, 23:36:04 UTC

For reasons I still can't fathom, they still insist on sending these v7 cuda22 and cuda23 tasks out en masse, with almost no hope of successful completion.

There are hosts that have not received any tasks for higher CUDA levels. (Too old driver I guess.) Not sending v7 cuda22/23 work would cut those hosts from GPU work (they might get AP work if anything is available and the host meets whatever requirements AP has.)

edit: I should clarify before anyone mis-interprets anything. That was just my personal observation.

Profile Jeff Buck
Send message
Joined: 11 Feb 00
Posts: 345
Credit: 36,026,127
RAC: 95,646
United States
Message 1386256 - Posted: 1 Jul 2013, 0:11:38 UTC - in response to Message 1386240.


There are hosts that have not received any tasks for higher CUDA levels. (Too old driver I guess.) Not sending v7 cuda22/23 work would cut those hosts from GPU work

But if almost all those hosts are simply trashing those tasks (my observation), they're not actually doing GPU work anyway, they're just piling up errors!

Profile Rae of Quakesville,CHCH NZ
Avatar
Send message
Joined: 30 Nov 01
Posts: 56
Credit: 195,657
RAC: 0
New Zealand
Message 1386348 - Posted: 1 Jul 2013, 7:28:39 UTC - in response to Message 1386131.

since the reset, I have had no more errprs or invalids, 3 pendings & 3 valids, so with that I am happy. all the current tasks that have been assigned to desktop at the moment are cuda 5.0 <runs sweet as> & ord seti v 7.0 for the cpu to nibble on.
____________
A city destroyed by an earthquake is an opportunity to Rebuild, redeign & make it a better place to be. Better, stronger, faster like the 6 Million Dollar Man

Juha
Volunteer tester
Send message
Joined: 7 Mar 04
Posts: 184
Credit: 143,295
RAC: 42
Finland
Message 1386460 - Posted: 1 Jul 2013, 14:47:37 UTC - in response to Message 1386256.


There are hosts that have not received any tasks for higher CUDA levels. (Too old driver I guess.) Not sending v7 cuda22/23 work would cut those hosts from GPU work

But if almost all those hosts are simply trashing those tasks (my observation), they're not actually doing GPU work anyway, they're just piling up errors!

I suppose that's true. When I went through the hosts I was concentrating on finding any that had received v6.11 instead of looking how the hosts worked with v7 cuda22/23.

Message boards : Number crunching : First Invalid result... why now?

Copyright © 2014 University of California