1) Message boards : Number Crunching : undistributed (Message 5742)
Posted 25 Jul 2015 by Profile Henk Haneveld
I'm seeing a similar issue as Conan but not that far back - only to 16 Jul 2015, 2:07:07 UTC.
Task by other user was aborted and new task remains unsent.


I think the problem is because we generate so many workunits at once in large batches -- maybe these are being put in the back of the queue and not getting sent out?


I am not sure but based on what I have seen at others projects there is a option in the server settings to give resends priority when work is send to users. May be you can have a look.
2) Message boards : Number Crunching : Badges (Message 5128)
Posted 15 Apr 2015 by Profile Henk Haneveld
DoctorNow

Very nice design. It gets my vote.
3) Message boards : News : [dna,wildlife] status update (Message 5036)
Posted 20 Mar 2015 by Profile Henk Haneveld
Travis,

Have you considered the possibility that residual values in internal computer memory may have a effect.

I noticed that results that have a restart from checkpoint because of a Boinc heartbeat loss will mostly be valid.

However a start from checkpoint after a shutdown and restart of the computer will to the best of my knowledge always end in invalid.
4) Message boards : News : [wildlife] New Badges (Message 5008)
Posted 28 Feb 2015 by Profile Henk Haneveld
In the 2 months that I am a member of CSG there has not been a single workunit for Wildlife.

What is the point of creating badges when it is impossible to earn them?
5) Message boards : Number Crunching : Error with new longer tasks. (Message 4939)
Posted 4 Jan 2015 by Profile Henk Haneveld
Travis,

There is a bit of additional information.

I have found out that some of the results that have a unequal seed value will validate.

There may be some kind of margin in the validation process that is allowed.
6) Message boards : Number Crunching : Error with new longer tasks. (Message 4926)
Posted 27 Dec 2014 by Profile Henk Haneveld
Ananas

Leaving suspended tasks in memory is already on, so that is no help.

I have checkpoint writing set at a 5 minute interval instead of the 1 minute default.
I will increase this to 15 minutes but I run several projects and a higher value is unpractical

However I did locate the source of the validation error.
In the stderr output file there is a value "argument seed (number)" at the start and a value "seeding (number)" at the end.

If these are equal then the result is valid, unequal is invalid.

It looks to me that restarting from a checkpoint can cause this seed value to become corrupt and that points to somekind of error in the way the application loads checkpoint data.
7) Message boards : Number Crunching : Error with new longer tasks. (Message 4924)
Posted 27 Dec 2014 by Profile Henk Haneveld
Correction to my first posting.

I was to fast thinking that restarting from a checkpoint is a problem.

I have just returned another result with a checkpoint restart and this one validated just fine.

I still don't understand why the other 2 had problems.

I will be gratefull for any help dealing with this.
8) Message boards : Number Crunching : Error with new longer tasks. (Message 4923)
Posted 27 Dec 2014 by Profile Henk Haneveld
In joined the project a couple of days ago and I have a problem with results getting called invalid.

It looks like the cause is that I don't run my systeem 24/7 and that these results where in progress when I shut down for the night and then had to start up the next day from the last checkpoint.

http://volunteer.cs.und.edu/csg/result.php?resultid=2123977

http://volunteer.cs.und.edu/csg/result.php?resultid=2127647

Please fix the problem or advise a way to avoid these longer running results