Several Assignments not automatically unreserved
 

I removed two machines from GIMPS in December, and normally when I do that, their exponents are unreserved. However for these two computers, the exponents are still there:

[url]http://i.imgur.com/DVffsQD.jpg[/url]

even after 60 days of inactivity. Should I manually unreserve them, or just keep waiting?

And possibly related George / James, or perhaps something completely different...

One of my R720s was working on 49139413. Out for only 3 days. I took a look at the [URL="http://www.mersenne.org/assignments/default.php?exp_lo=40000000&exp_hi=54000000&execm=1&exdchk=1&extf=1&B1=Get+Assignments"]Assignment report[/URL] for low LL candidates, and noticed a second entry in the list for this candidate -- picked up by "Spidy" this morning.

[CODE]49139413 LL LL, 10.50% 3 2 2014-03-06 2014-03-04 2014-03-04 2014-03-01 wabbit R720_4
49139413 LL 0 79 2014-05-22 2014-03-05 2014-03-04 2014-03-04 wabbit R720_5[/CODE]

As an experiment, I updated the worktodo.txt entry for this candidate with the new AID, and had the client report to Primenet. The second entry was then updated:

[CODE]49139413 LL LL, 10.50% 3 2 2014-03-06 2014-03-04 2014-03-04 2014-03-01 wabbit R720_4
49139413 LL LL, 10.90% 0 2 2014-03-06 2014-03-04 2014-03-04 2014-03-04 wabbit R720_4[/CODE]

So, this suggests to me that these are indeed two different LL assignments for this single candidate.

Is this a bug George? Or are you so sure it's a MP that you want it checked and double checked ASAP? :wink:

[QUOTE=sixblueboxes;368290]However for these two computers, the exponents are still there[/QUOTE]

When exponents are expired the reservation now stays in the database. This helps in the case where the computer eventually returns a result, we can better give proper CPU credit. Your assignments did expire on March 1st.

I've changed the workload report to not display these expired assignments. Someday, I might create a report to show expired and poached assignments and let you really delete them. Pretty low priority.

[QUOTE=Prime95;368318]When exponents are expired the reservation now stays in the database. This helps in the case where the computer eventually returns a result, we can better give proper CPU credit.[/QUOTE]

Does this explain the reassignment of 49139413?

No. I presume prime.log doesn't show anything interesting (like an unexplained unreserve of the exponent).

There are about 14 exponents that are either double-assigned or assigned-and-available. So, there is a leak somewhere that needs to be plugged.

[QUOTE=Prime95;368330]No. I presume prime.log doesn't show anything interesting (like an unexplained unreserve of the exponent).[/QUOTE]

Nope. Nothing interesting. Primenet was very happy to receive updates under the original AID, but signed it out again (it was only by chance that "Spidy" captured this assignment, which I was then able to experiment with).

[QUOTE=Prime95;368330]There are about 14 exponents that are either double-assigned or assigned-and-available. So, there is a leak somewhere that needs to be plugged.[/QUOTE]

AKA, a bug.

This is not unexpected, and really not a big deal. When changes happen unexpected results can occur. Thanks for working this.

[QUOTE=chalsall;368294]Is this a bug George? [/QUOTE]

Yes. Your Mar 1 assignment was due to expiring an old assignment.

The expired assignment reported a result on Mar 4. If the LL result had been clean, your assignment would have been demoted to a DC. However, the LL result was suspect. Instead of doing nothing, the server erroneously made the exponent available for a first LL test.

I fixed chashall's bug. However, the 8 exponents that were double-assigned remain double-assigned. I don't think this will cause any problems other than screwing up the counts on the work distribution web page.

[QUOTE=Prime95;368423]I fixed chashall's bug. However, the 8 exponents that were double-assigned remain double-assigned. I don't think this will cause any problems other than screwing up the counts on the work distribution web page.[/QUOTE]

It was actually your bug, I just very carefully documented it. :razz:

Thanks George. We're all having fun here. :smile: