mersenneforum.org  

Go Back   mersenneforum.org > Prime Search Projects > No Prime Left Behind

Reply
 
Thread Tools
Old 2009-08-10, 21:40   #34
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

186916 Posts
Default

Quote:
Originally Posted by rogue View Post
I need the prpserver.log file that shows the communication back to the client at that time. The terminal output from the server would be even more helpful.

I actually think that I found a problem, maybe not the problem, but a problem related to this bug. I'll patch it tomorrow.
Okay, here's the prpserver.log file from 2009-08-09 00:00 GMT until now. I don't have this server on debug logging so I can't give you the terminal output, but hopefully this will do the trick.
Attached Files
File Type: bz2 prpserver-2000.log.bz2 (11.9 KB, 146 views)
mdettweiler is offline   Reply With Quote
Old 2009-08-13, 04:18   #35
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

22×7×359 Posts
Default

After discussion with Max and Ian, we have decided to delay the implementation of PRPnet at NPLB by 3-6 months or more due to stability issues as more clients are added to the server.

Testing at CRUS revealed that multi-threading is needed and that storing the file in memory and updating from there works OK for a limited # of clients but not as more clients are added. If you have technical questions related to that, Max and/or Ian can give more details.

PRPnet is still excellent for small efforts and the "BOINC-like" quality that it has that allows you to dictate what percentage of your machine(s) run what efforts on different servers is excellent!

If you understand the general setup, I recommend a personal PRPnet server for most people but I would not suggest running more than about 10 cores on each server that you use at any one time at the current approximate median level of n-ranges for our project (n=~450K-850K). The max # of cores that it can handle should vary with the length of time that each test takes. Therefore, the base 5 project may be able to run most of their project on PRPnet because their tests are so long. It PRPnet is still not fully ready by the time we're ready to start k=300-400 for n=1M-2M (I'd like to shoot for Nov. on that after we are hopefully complete to n=1M on that range), we may consider a somewhat limited-use public server for these long tests.

For the time being, we will not use port G5000 for the small tests as originally planned so I will/have removed it from the 1st post here. Such short tests could possibly cause problems with as few as 5 clients.

Over the next few days, I will have Max set up a server on one of my machines to start processing the non-top-5000 ranges of k=2000-3400. Shortly after that, I'll create a thread officially starting the effort. For those of you who don't like to mess with reporting primes either at top-5000 or here (admins take care of tracking all non-top-5000 primes) but would like to help us get some ranges filled in, I would suggest that effort for you.


Gary

Last fiddled with by gd_barnes on 2009-08-13 at 04:18
gd_barnes is offline   Reply With Quote
Old 2009-08-13, 04:22   #36
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

22×7×359 Posts
Default

Max and Mark,

The link http://nplb-gb1.no-ip.org:2000/server_status.html doesn't seem to be working. It comes up but nothing shows on the page.


Gary
gd_barnes is offline   Reply With Quote
Old 2009-08-13, 04:45   #37
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

3×2,083 Posts
Default

Quote:
Originally Posted by gd_barnes View Post
Max and Mark,

The link http://nplb-gb1.no-ip.org:2000/server_status.html doesn't seem to be working. It comes up but nothing shows on the page.


Gary
Hmm, that's weird. All the other pages work, but this one doesn't work even when I tried restarting the server. Mark, do you know what might be causing this?
mdettweiler is offline   Reply With Quote
Old 2009-08-13, 05:18   #38
IronBits
I ♥ BOINC!
 
IronBits's Avatar
 
Oct 2002
Glendale, AZ. (USA)

3×7×53 Posts
Default

chmod 666 *.html :)
IronBits is offline   Reply With Quote
Old 2009-08-13, 12:45   #39
rogue
 
rogue's Avatar
 
"Mark"
Apr 2003
Between here and the

11×509 Posts
Default

Quote:
Originally Posted by mdettweiler View Post
Hmm, that's weird. All the other pages work, but this one doesn't work even when I tried restarting the server. Mark, do you know what might be causing this?
Nope. Anything in the debug log?
rogue is offline   Reply With Quote
Old 2009-08-13, 14:17   #40
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

11000011010012 Posts
Default

Quote:
Originally Posted by rogue View Post
Nope. Anything in the debug log?
I hadn't checked that. I'll go do that now...

BTW, I got a PM from Ian saying that in his experience, having that particular page blank out was a direct precursor to data being lost from the middle of the prpserver.candidates file. Sure enough, when I checked the server, it seems that there's a small hole in the file: it's supposed to be all of k=300-400 for n=300K-350K, but k=359 and k=361 are missing. If something did happen, though, it happened rather cleanly; no other k's, even the ones around them, seem to be affected (at least at a glance).

Okay, I just checked the server and it turns out I didn't have that one on debug logging. I've switched it on to debug logging now, though, so any future problems should be logged.

BTW, a brief look at the end of the non-debug log indicates that there's some barfing going on on G2000, too. I see entries saying "Rejected test on x due to no residue" and then immediately after that it accepts the result with the blank residue--just as I was seeing on G3000. If this problem could be fixed, then PRPnet would be at least stable enough to use for lower-volume sitatuations, even without being able to handle high-volume capacity without multithreading.

(It would probably be rather superfluous to provide the log for this barfing incident since it looks like the exact same thing that's happened a million times on G3000. )
mdettweiler is offline   Reply With Quote
Old 2009-08-13, 15:58   #41
rogue
 
rogue's Avatar
 
"Mark"
Apr 2003
Between here and the

11×509 Posts
Default

Quote:
Originally Posted by mdettweiler View Post
BTW, a brief look at the end of the non-debug log indicates that there's some barfing going on on G2000, too. I see entries saying "Rejected test on x due to no residue" and then immediately after that it accepts the result with the blank residue--just as I was seeing on G3000. If this problem could be fixed, then PRPnet would be at least stable enough to use for lower-volume sitatuations, even without being able to handle high-volume capacity without multithreading.

(It would probably be rather superfluous to provide the log for this barfing incident since it looks like the exact same thing that's happened a million times on G3000. )
If you can e-mail me the relevant pieces of the client and server logs, I would appreciate it. It has been difficult following the various posts.
rogue is offline   Reply With Quote
Old 2009-08-13, 16:17   #42
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

3·2,083 Posts
Default

Quote:
Originally Posted by rogue View Post
If you can e-mail me the relevant pieces of the client and server logs, I would appreciate it. It has been difficult following the various posts.
Which logs do you want me to send you? The ones sprinkled throughout the "PRPnet bugs" thread in the CRUS forum?
mdettweiler is offline   Reply With Quote
Old 2009-08-13, 22:29   #43
rogue
 
rogue's Avatar
 
"Mark"
Apr 2003
Between here and the

11·509 Posts
Default

Quote:
Originally Posted by mdettweiler View Post
Which logs do you want me to send you? The ones sprinkled throughout the "PRPnet bugs" thread in the CRUS forum?
The ones relevant to the latest issue that you mentioned
rogue is offline   Reply With Quote
Old 2009-08-13, 22:44   #44
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

186916 Posts
Default

Quote:
Originally Posted by rogue View Post
The ones relevant to the latest issue that you mentioned
Oh, I see. I'm afraid I don't have much data on the problem with the server dropping candidates from its queue; the server in question didn't have debug logging enabled at the time.

I'll try to reproduce the bug locally, this time with debug logging. Hopefully that will give us a clue as to what's going on here.
mdettweiler is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
PRPnet Servers for CRUS MyDogBuster Conjectures 'R Us 76 2018-03-09 19:05
LLRnet servers for NPLB kar_bon No Prime Left Behind 1343 2014-08-20 09:38
Public PRPNet Servers rogue Open Projects 26 2013-01-16 01:33
PRPNet servers down? opyrt Prime Sierpinski Project 13 2009-11-04 21:33
Servers for NPLB gd_barnes No Prime Left Behind 0 2009-08-10 19:21

All times are UTC. The time now is 14:06.

Fri Feb 21 14:06:09 UTC 2020 up 21 days, 8:38, 1 user, load averages: 1.98, 2.06, 2.05

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.