Manual result submission could fail

sonjohan · 2012-04-23, 17:39

I just uploaded a file with results, and the server took a long time responding.
In the end, the load screen ended (IE/FF stops loading the page), but I didn't get the summary of my submission.

Thanks to the fact that gpu272 informed me that I had 21 exponents left and not the 10 I saw, I could see that only half of my submitted results were loaded.

I re-uploaded the file and now it's ok.
Maybe a warning on the manual results submission (files) should be displayed, so people know to resend their files upon failed/partial check-in. I was happy with the 353% CPU for 24h, but the 797% was more accurate :).

chalsall · 2012-04-23, 17:50

Quote:

Originally Posted by sonjohan

Maybe a warning on the manual results submission (files) should be displayed, so people know to resend their files upon failed/partial check-in. I was happy with the 353% CPU for 24h, but the 797% was more accurate :).

Yeah...

Unfortunately PrimeNet has been having problems for the last few weeks (and the last few years).

Scott is no where to be found.

George listens, but doesn't always react.

Guys, it's time to reboot the Mersenne server....

Dubslow · 2012-04-23, 19:10

Quote:

Originally Posted by sonjohan

Maybe a warning on the manual results submission (files) should be displayed, so people know to resend their files upon failed/partial check-in.

A warning is just as hard to get out as the success message is, so if the latter doesn't work, nor would the former. There's not much we can do about it, except perhaps do uploads in smaller batches (or use the submission spider, which does them one at a time and waits specifically for the success message before moving on to the next one).
(Edit) I take it back. I cannot even submit one result at a time with my spider. The server definitely needs a reboot. (Edit2: I got the "CGI timeout/deleted process" error. Could a mod move these posts to the server problems thread?)

LaurV · 2012-04-24, 00:35

As I said before, that's part of the why I report "by hand"

. I can afford 5 minutes in the evening when I arrive home, to report whatever mfaktc harvested during the day, and sometimes (like now) 5 minutes in the morning to report the overnight results. Don't trust the spiders

Well... sometimes I have to try many times. But that's the part of the fun... And you say that a reboot solves the problem? Hmmm... I am thinking now to some virus to reboot every 24 hours, the question is how to cheat George to run it on the server...

Dubslow · 2012-04-24, 00:44

Quote:

Originally Posted by LaurV

As I said before, that's part of the why I report "by hand"

. I can afford 5 minutes in the evening when I arrive home, to report whatever mfaktc harvested during the day, and sometimes (like now) 5 minutes in the morning to report the overnight results. Don't trust the spiders

...I fail to see what's wrong with spiders. My spider (based on chalsall's) observed that it wasn't working, and aborted gracefully. Btw, it did work fine 20 minutes ago.

LaurV · 2012-04-24, 01:26

Just teasing you man... (but I do report by hand, I am a skeptical sob, hehe).
Edit: and yes, it is worse and worse and worse day by day when I report my mfaktc stuff. Yesterday, after trying 3-4 times to send the whole file without success, I had to split the file in 5 or 6 parts. There was no way to report all-at-once, it always ended with timeout error. The most of the reported lines were recorded. When I reported the "splits" afterthat I got error 40, result not needed, for most of the lines, this means they were recorded, only the communication timeouted. But some of them were not recorded, in quite a random fashion (when I reported the "splits", random lines gave me the right credit, meaning they were not recorded first/second/third time when I tried to report the whole file). These "trials" also generate server traffic, and not only a little! If the things would go well at the first trial, then the traffic would also be less, we won't need to try and try. But here we run in circles... or in cascade, if it starts to go wrong, we try and try (by hand or by spiders, I think that the "more and more" spiders in the last time contribute to process, with their number and their "auto-retry" functions, more then the pure manual-reporting workers, like me

) and make the things worse and worse.

Xyzzy · 2012-04-24, 03:34

Such tedious work, reporting completed work.

Alas, we have resorted to only reporting factors.

Dubslow · 2012-04-24, 05:07

Quote:

Originally Posted by Xyzzy

Such tedious work, reporting completed work.

Alas, we have resorted to only reporting factors.

...but that screws up both GPU272 and PrimeNet. PrimeNet has no idea that there isn't a factor if you don't report it... meaning the work will be duplicated anyways, so you may as well have not done it.

Xyzzy · 2012-04-24, 12:36

We will dump the no factor results someday once the server is in a better mood.

S34960zz · 2012-04-24, 15:25

Per Scott K. (email to me, Mon. Apr 23, 2012 5:17 PM [New York time(?)]): "I paused the web service a minute to let the database catch up and resumed the web. We are upgrading the server soon -- we've outgrown our data center host's server -- probably the next few months." (also posted in 'OFFICIAL "SERVER PROBLEMS" THREAD')

kladner · 2012-04-25, 02:21

Quote:

Originally Posted by Xyzzy

We will dump the no factor results someday once the server is in a better mood.

But it seems that even submitting just factors won't catch you up with Craig on the "Work Saved" charts.

Maybe you need a couple more GTX 580s?

2012-04-23, 17:39	#1
sonjohan May 2003 Belgium 2·139 Posts	Manual result submission could fail I just uploaded a file with results, and the server took a long time responding. In the end, the load screen ended (IE/FF stops loading the page), but I didn't get the summary of my submission. Thanks to the fact that gpu272 informed me that I had 21 exponents left and not the 10 I saw, I could see that only half of my submitted results were loaded. I re-uploaded the file and now it's ok. Maybe a warning on the manual results submission (files) should be displayed, so people know to resend their files upon failed/partial check-in. I was happy with the 353% CPU for 24h, but the 797% was more accurate :).

2012-04-24, 01:26	#6
LaurV Romulan Interpreter Jun 2011 Thailand 9,377 Posts	Just teasing you man... (but I do report by hand, I am a skeptical sob, hehe). Edit: and yes, it is worse and worse and worse day by day when I report my mfaktc stuff. Yesterday, after trying 3-4 times to send the whole file without success, I had to split the file in 5 or 6 parts. There was no way to report all-at-once, it always ended with timeout error. The most of the reported lines were recorded. When I reported the "splits" afterthat I got error 40, result not needed, for most of the lines, this means they were recorded, only the communication timeouted. But some of them were not recorded, in quite a random fashion (when I reported the "splits", random lines gave me the right credit, meaning they were not recorded first/second/third time when I tried to report the whole file). These "trials" also generate server traffic, and not only a little! If the things would go well at the first trial, then the traffic would also be less, we won't need to try and try. But here we run in circles... or in cascade, if it starts to go wrong, we try and try (by hand or by spiders, I think that the "more and more" spiders in the last time contribute to process, with their number and their "auto-retry" functions, more then the pure manual-reporting workers, like me) and make the things worse and worse. Last fiddled with by LaurV on 2012-04-24 at 01:42

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Help! My manual submission of trial factoring didn't associate with my account	daxmick	Information & Answers	4	2018-02-12 12:12
Data required for manual submission of results	Wetitpig	Data	10	2018-01-29 20:34
New Manual Results Submission Page	TheMawn	PrimeNet	12	2014-04-09 14:30
Manual submission of automatic assignment result	tichy	PrimeNet	4	2010-12-17 09:57
Manual Testing - Results Submission	rogue	Sierpinski/Riesel Base 5	5	2008-04-05 02:52

2012-04-24, 00:35	#4
LaurV Romulan Interpreter Jun 2011 Thailand 9,377 Posts	As I said before, that's part of the why I report "by hand" . I can afford 5 minutes in the evening when I arrive home, to report whatever mfaktc harvested during the day, and sometimes (like now) 5 minutes in the morning to report the overnight results. Don't trust the spiders Well... sometimes I have to try many times. But that's the part of the fun... And you say that a reboot solves the problem? Hmmm... I am thinking now to some virus to reboot every 24 hours, the question is how to cheat George to run it on the server...

2012-04-24, 03:34	#7
Xyzzy "Mike" Aug 2002 5·1,609 Posts	Such tedious work, reporting completed work. Alas, we have resorted to only reporting factors.

2012-04-24, 12:36	#9
Xyzzy "Mike" Aug 2002 5·1,609 Posts	We will dump the no factor results someday once the server is in a better mood.

2012-04-24, 15:25	#10
S34960zz Feb 2011 110100₂ Posts	Per Scott K. (email to me, Mon. Apr 23, 2012 5:17 PM [New York time(?)]): "I paused the web service a minute to let the database catch up and resumed the web. We are upgrading the server soon -- we've outgrown our data center host's server -- probably the next few months." (also posted in 'OFFICIAL "SERVER PROBLEMS" THREAD')