Assignment died of dysentery

[Dubslow]

Originally Posted by chalsall

PrimeNet assigned Spidy about 20 candidates that it already owned.

?

???????













????????????????????????????????????????????????????????????????????????

[chalsall]

Originally Posted by Dubslow

????????

Yeah. I know...

This happened for the first time two days ago. I noticed it last night when I was watching Spidy do it's thing at around midnight UTC, and observed that the "Assigned" count had dropped by a few. As the observation script had not run during that time (and no one had unreserved any), I knew that something was up and started drilling down...

The good news is I keep a backup of the database once a day (plus just before I'm doing any hand SQL work or testing new code which does updates), so I'll be able to determine exactly when the first assignment from PrimeNet was given, and when it was reissued.

In addition, I've "taught" "Spidy" to raise an alarm any time this happens in the future, and am going to write another "Sanity Check" spider which will run at a few times a day to make sure the system actually owns what it thinks it does.

[petrw1]

Originally Posted by chalsall

petrw1, if it's not too much of a pain, could you please remove these four P-1 lines from your worktodo.txt file(s) if you haven't already starting working on them?

51842653 is almost completed so I would like to let it finish so at least I get the prime95 credit.

One other had just started so I removed it.

The other two are gone, gone gone.

Thanks

[petrw1]

Originally Posted by chalsall

petrw1, if it's not too much of a pain, could you please remove these four P-1 lines from your worktodo.txt file(s) if you haven't already starting working on them?

Looks like there may have been a few more...and any that I finished were recorded in my completions but with 0 points.

51879281	72	P-1	2011-12-05 19:15:44	2011-12-30 00:34:04	0.000
51766133	72	P-1	2011-12-05 18:46:12	2011-12-30 00:34:04	0.000
51765079	72	P-1	2011-12-05 18:46:12	2011-12-30 00:34:04	0.000
51871483	72	P-1	2011-12-05 19:15:44	2011-12-30 00:34:04	0.000

[Dubslow]

Only the first of those show a completed P-1 on their Exponent Status pages. All have been factored to 72 bits.

http://www.mersenne.org/report_expon...&B1=Get+status
http://www.mersenne.org/report_expon...&B1=Get+status
http://www.mersenne.org/report_expon...&B1=Get+status
http://www.mersenne.org/report_expon...&B1=Get+status

[petrw1]

I could assume they are these 4; though as DUB pointed out the first one does have a P-1 and by me.

Or was it these 3 and another elsewhere?

If it helps I just did a reconciliation to Prime95 results.
They match exactly...EXCEPT for 51879281.
PrimeNet shows it completed by me; GPU does NOT.
Now if it was one Spidey accidentally gave to someone else; no skin lost; but if and when that person completes it PrimeNet may deem it unneeded.

[chalsall]

Originally Posted by petrw1

I could assume they are these 4; though as DUB pointed out the first one does have a P-1 and by me.

Or was it these 3 and another elsewhere?

Sorry. I had promised to report back as to what happened here. (It's been a *very* busy week...)

On 2011.12.29 and 2011.12.30, PrimeNet reassigned work to Spidy which it (read: Spidy) believed it owned.

Because of a "Stupid Programmer Error", the SQL statement "replace into GPU ..." was used instead of a more appropriate "insert into GPU ... on duplicate key update ..." statement.

The "replace" statement set to zero (the default) the "Status" value in the GPU table for those particular records. Because of a cascade error, when Spidy saw that the TF work had been completed for those P-1 work assignments, it erroneously marked the P-1 work as complete (but with 0 GHzDays credit given) and made the P-1 work available to others for those candidates.

The good news is I noticed what had happened on the 30th, and prevented erroneous assignments that evening. I figured out what was happening on the 31st, and wrote another spider to prevent it from happening again.

The bad news is there was a (fortunately) small amount (approximately 15) of duplicate P-1 work done.

My apologies.

So, why did PrimeNet reassign work already "owned" by Spidy? Because Spidy (until the 31st) was not reporting in on assignments it had, and PrimeNet (reasonably) will reassign work which has not been reported on for 60 (or perhaps it's 61) days. (I thought it was six months.)

Again, my apologies for this. I know how seriously many (myself included) take the avoidance of duplication of effort.

Please rest assured that Spidy and PrimeNet are now talking regularly about what Spidy beleives it owns, and a situation like this should never happen again.

[Dubslow]

Originally Posted by chalsall

Please rest assured that Spidy and PrimeNet are now talking regularly about what Spidy beleives it owns, and a situation like this should never happen again.

The use of third person has made this so much more entertaining to read

[Xyzzy]

Right thread?

Processing result: no factor for M50361253 from 2^71 to 2^72 [mfaktc 0.17-Win barrett79_mul32]
Error code: 40, error text: TF result for M50361253 was not needed

[chalsall]

Originally Posted by Xyzzy

Right thread?

Right enough...

Jerry... Is there any chance you are doing work you were not assigned?

According to the "records", you were assigned this to TF from 70 to 71 (on "2011-11-24 22:47:24"; completed "2011-12-12 03:33:38").

You were then assigned this candidate again to P-1 (on "2011-12-23 01:42:23"; completed "2012-01-09 18:54:11").

But, according to PrimeNet, you then submitted a result for TF between 71 to 72 on "2012-01-11";

Did Spidy (read: my code) make a mistake here?

[flashjh]

Originally Posted by chalsall

Right enough...

Jerry... Is there any chance you are doing work you were not assigned?

According to the "records", you were assigned this to TF from 70 to 71 (on "2011-11-24 22:47:24"; completed "2011-12-12 03:33:38").

You were then assigned this candidate again to P-1 (on "2011-12-23 01:42:23"; completed "2012-01-09 18:54:11").

But, according to PrimeNet, you then submitted a result for TF between 71 to 72 on "2012-01-11";

Did Spidy (read: my code) make a mistake here?

Everything I load into the worktodo files comes from the GPU-to-72 request pages (except a few LL tests in Prime95 that I'm finishing). I searched for M50361253 in my records and found my results files matched PrimeNet, though I can only find my P-1 assignment on GPU-to-72. The TF 70-71 is old, but I don't know where the 71-72 came from? Again, I appologize for the poached TF . (you want one of mine to make up for it?)

I just re-downloaded my entire TF assignments today so I could get them into date order, so there is nothing in those files that was not on my assignments page.