OFFICIAL "SERVER PROBLEMS" THREAD

[Mark Rose]

Originally Posted by James Heinrich

Sounds plausible, the current report shows exactly 3000 rows.

I'm submitting over 1000 results an hour right now, so that's likely it.

[Madpoo]

Originally Posted by chalsall

Perhaps there is a "limit" clause in the SQL statement.

There is a limit. I suppose it should say "the 3000 most recent" results.

If someone needs everything for the day, there's the daily XML of all the results that came in. Much better than trying to crawl once an hour and hope it's all there, filtering out duplicates when less than 3000 came in that hour, etc. And XML is probably so much easier to parse (and it's super fast to generate too, FYI).

[GP2]

Originally Posted by GP2

Reportedly, the server is giving bad worktodo lines for Fermat number F12, omitting most of the already-known factors. See this thread.

This might be the cause of the bogus already-known factors for F12 (exponent 4096) that constantly appear in the Recent Cleared list.

(note: there are actually 6 known factors of F12):

ECM2=1,2,4096,1,800000000,80000000000,3,"114689,26017793,63766529,190274191361,1256132134125569,568630647535356955169033410940867804839360742060818433"

Well, Recent Cleared is still showing a whole bunch of new rediscoveries of known factors for F12 (2⁴⁰⁹⁶+1), so either those folks are copy-pasting their own worktodo lines, or they are still working off a backlog of old assignments, or the server is still sending bad worktodo lines.

Unfortunately I can't test it, because even if I set Memory=8 in local.txt, the server won't send any assignment smaller than F13. Is anyone else able to get automatic F12 assignments to test this?

[thyw]

No, in the client or /cpus/ I can only change the worktype with no range. In manual assigments page, where you can request it by range, the Fermat worktype is not present.
Also when I select it from the client menu (Advanced/ECM), it registers, but doesn't fill in the factors in the worktodo (is it intended?)

About mersenne.org: the workload page is really nice with separated cpus.
For me, the recently cleared (and other pages) are "in the middle of the screen"/squished, with grey bars in the sides and the main table in recently cleared is scrollable "inside the page". Like an iframe, but without scrollbars. <-only in Firefox, not in Chrome

[GP2]

Originally Posted by thyw

Also when I select it from the client menu (Advanced/ECM), it registers, but doesn't fill in the factors in the worktodo (is it intended?)

Ah, that's an interesting point.

There are only 18 Fermat numbers even remotely feasible for mprime to work on and only 34 known factors of those Fermat numbers, and new discoveries are rare to nonexistent after an initial burst a few years ago. So the mprime executable could perhaps check k*b^n+c for the case of k=1, b=2, n=power of 2, c=1 and then just automatically incorporate and apply those hardcoded values.

It might seem a bit weird to special-case this directly in the executable, but there is no conceivable practical reason that anyone would need to run a Fermat ECM without the known factors, and for many months if not years a few people have been doing so continuously and pretty much wasting all those cycles. Each ECM curve simply terminates in the rediscovery of a small factor or composite product of small factors, and then a new ECM curve starts up only to arrive at the same futile end. Although on 2017-05-02 05:13, one user returned a 75-digit composite factor consisting of the three smallest plus the one largest known prime factors, so I guess there's still hope.

It's the same handful of usernames who keep showing up in Recent Cleared, so I doubt that they are filling in all the values in Advanced/ECM by hand over and over again. But possibly some of them started by filling in the values for F12 in Advanced/ECM and then used the resulting worktodo.txt for copy-pasting over and over ever since.

Anyways, it's possible they're still working off a backlog of old assignments, although with each passing day since April 21 that might seem less likely.

[LaurV]

For some reasons, manual reports page does not work. I get all pages right, but on https://www.mersenne.org/manual_result/ I get a blank page. I think you got upset on me reporting too many primes and banned me, hehe...

[James Heinrich]

Fixed. Sorry about that.

[LaurV]

working now, thanks a lot!

[error]

There is another problem with manual submissions -

For some exponents I get a message like this:
processing: TF factor 15750471591088124777 for M17086021 (263-264)
Error code: 40, error text: Factor 15750471591088124777 reported for completely factored M17086021


and it does not accept the result.

[axn]

Apparently, for some reason the server thinks that that particular exponent is completely factored?!

[Prime95]

Originally Posted by error

There is another problem with manual submissions -

For some exponents I get a message like this:
processing: TF factor 15750471591088124777 for M17086021 (263-264)
Error code: 40, error text: Factor 15750471591088124777 reported for completely factored M17086021


and it does not accept the result.

My bad, please try again.