Got my own WU back for a Pending recheck .... looks on the face of it an error somewhere in the Validation code (it will come up in my queue for crunching just before midnight UK time 13 July).

http://moowrap.net/workunit.php?wuid=963690

EDIT: The two task numbers are: 1345827 & 1365975 - the former is the original WU, the latter is the tasked re-check, computer ID is 476

Regards
Zy

Got my own WU back for a Pending recheck .... looks on the face of it an error somewhere in the Validation code (it will come up in my queue for crunching just before midnight UK time 13 July).

http://moowrap.net/workunit.php?wuid=963690

EDIT: The two task numbers are: 1345827 & 1365975 - the former is the original WU, the latter is the tasked re-check, computer ID is 476

Regards
Zy

I have always wondered if that happened? I am guessing this is NOT on the same pc?

Same PC - ID 476

Regards
Zy

Just to complete the loop ..... it went through just now and validated. Still no idea why it was sent back to the same machine. Let me know if there is anything I can check for you that will shed light on it.

http://moowrap.net/workunit.php?wuid=963690

Regards
Zy

Just to complete the loop ..... it went through just now and validated. Still no idea why it was sent back to the same machine. Let me know if there is anything I can check for you that will shed light on it.

http://moowrap.net/workunit.php?wuid=963690

Regards
Zy

Something is wrong then, according to Dr. A, many years ago now, Boinc is set up to specifically avoid that!

This one gets stranger as time goes on. It was moved to invalid in spite of the fact that it validated. I suspect the old hassle of "two invalids = invalid" no matter that another actually validated the WU, has returned.

If this is the base BOINC Validator code, it needs flagging up on the Alpha list if the server version currently used post dates the last time this was tackled by BOINC devs.

Regards
Zy

Hi,

Hmm, just missed the workunit/tasks, looks like they already got purged. Which is odd, considering I thought I set it to hold them up to 21 days after last change.. Oh well, guess I can instead verify what happened to them and when. :)

Anyway, I did do some code changes in the stock validator code (in addition to normal project specific validation parts) to detect these partial packets and especially to issue partial credit after they complete correctly. Scheduler code should be same, though.

For us it doesn't matter that much if the same host gets a second crack on the work unit as long as one task gets done correctly in the end. If my memory works correctly, I believe there are some config settings regarding issuing tasks to same hosts and we have them disabled (affects scheduler). I know there are some to request different platforms to do tasks (HR) and to get more than one result (initial replication) but those are definitely disabled for us.

-w