24 hour rally

[gd_barnes]

OK, guys. I have the LLRnet high-range server loaded on 10 total cores in preparation for the rally, both home and work...Currently running 5 total.

For the rally, assuming that we agree that we'll do the lower range, I want to quickly change them all to the lower-range server.

Is all that I need to do is change port 300 to 100 in the llr-clientconfig file?

Any other files to change? Do I need to reload the entire software again, etc.?


Thanks,
Gary

[mdettweiler]

Originally Posted by gd_barnes

OK, guys. I have the LLRnet high-range server loaded on 10 total cores in preparation for the rally, both home and work...Currently running 5 total.

For the rally, assuming that we agree that we'll do the lower range, I want to quickly change them all to the lower-range server.

Is all that I need to do is change port 300 to 100 in the llr-clientconfig file?

Any other files to change? Do I need to reload the entire software again, etc.?


Thanks,
Gary

Here's what you need to do:

-Shut down LLRnet.
-Open llr-clientconfig.txt, and change WUCacheSize to 1. Save the file.
-Start LLRnet.
-Wait until the queue clears out. You can check on how it's doing by looking at the workfile.txt file.
-When the workfile.txt file only has one k/n pair left in it, shut down LLRnet. Open llr-clientconfig.txt, and remove the "--" from the line that says "--once=1". Save the file.
-Start LLRnet. It will finish up the current k/n pair, then exit.
-Open llr-clientconfig.txt. Change the "port = 300" line to say "port = 100". Then, change the line that you edited earlier, which now says "once=1", and change it back to "--once=1". After you've done that, go down to the WUCacheSize variable, and change it to whatever you'd normally like your queue size to be.
-Start LLRnet. You're all set now!

The same applies for switching between the two CRUS servers. The reason why you need to go through this whole rigmarole is so that LLRnet cleans out its queue of k/n pairs that it's gotten from the server you originally had the client on, so when you start using it with the new server, you aren't trying to send it k/n pairs that are supposed to be for the first server.

[mdettweiler]

Originally Posted by gd_barnes

I thought for a long time while I had my kids over tonight before logging back on late here. I've done an "about face". You know it's been you that have brought all the Free-DC people here and you're running the servers on your machines. Also, having as much power as a rally will bring to a range that is not top-5000 is hard to come by. All you have to do is look at k>1000 to realize that a large majority of people are only interested in top-5000 primes instead of accurate and complete lists of primes. Even if we don't have massive firepower, we can probably get at least a reasonable number of people to search top-5000 ranges. What is not easy is getting people to search the lower ranges.

As long as we're moving forward toward the goals established by the project, having the rally on the lower range works for me. Filling in this range is a big step in that direction. I feel that the original #1 priority is now in good shape. That is I think we're past the 'danger zone' of having any primes n>333.3K not being top-5000 in a few months since we've already searched most k/n pairs up to ~n=360K.

The next most important step will be getting 300<k<400 searched quickly up to n=350K-360K. This k-range will have more of a 'double-check' flavor to it with probably about 20-25% of primes already found but becoming less as the n-range gets higher. That is why I initially avoided sieving it. I think having a double-check flavor to it makes it good for a rally. Some people won't like double-checking as much but are happy to participate in a fun rally with still a good chance of finding new top-5000 primes.

So...let's do the lower range for this rally. Then I'd like to do the next rally on 300<k<400 for n>333.3K 2-3 weeks after that if everyone can continue the break-neck pace. Sieving to P=6T will be complete Feb. 28th-29th. That will still allow over a week to get the new range loaded into the server, port # changed, and all that stuff.

Then in 5-6 weeks, I'd like to do a rally at CRUS. After that, we can decide where to go from there. FYI for those who don't know: CRUS = the "Conjectures 'R Us" project.

Thoughts, opinions, objections anyone?


Gary

Sounds good to me.

[gd_barnes]

Originally Posted by Anonymous

Here's what you need to do:

-Shut down LLRnet.
-Open llr-clientconfig.txt, and change WUCacheSize to 1. Save the file.
-Start LLRnet.
-Wait until the queue clears out. You can check on how it's doing by looking at the workfile.txt file.
-When the workfile.txt file only has one k/n pair left in it, shut down LLRnet. Open llr-clientconfig.txt, and remove the "--" from the line that says "--once=1". Save the file.
-Start LLRnet. It will finish up the current k/n pair, then exit.
-Open llr-clientconfig.txt. Change the "port = 300" line to say "port = 100". Then, change the line that you edited earlier, which now says "once=1", and change it back to "--once=1". After you've done that, go down to the WUCacheSize variable, and change it to whatever you'd normally like your queue size to be.
-Start LLRnet. You're all set now!

The same applies for switching between the two CRUS servers. The reason why you need to go through this whole rigmarole is so that LLRnet cleans out its queue of k/n pairs that it's gotten from the server you originally had the client on, so when you start using it with the new server, you aren't trying to send it k/n pairs that are supposed to be for the first server.

Thanks for the detail. I already knew all that about clearing out the queue. But it was tricky to understand the first time I did it after the last rally. On machines that have not been running LLRnet, all I need to do is change port 300 to 100 on the one file. And on ones that have, I have to clear the queue and then change the port.

This is excellent documentation to have for clearing out the queue since it takes several steps. I'm assuming it is in the instructions somewhere for the LLRnet server.


Gary

[mdettweiler]

Originally Posted by gd_barnes

Thanks for the detail. I already knew all that about clearing out the queue. But it was tricky to understand the first time I did it after the last rally. On machines that have not been running LLRnet, all I need to do is change port 300 to 100 on the one file. And on ones that have, I have to clear the queue and then change the port.

This is excellent documentation to have for clearing out the queue since it takes several steps. I'm assuming it is in the instructions somewhere for the LLRnet server.


Gary

Actually, it's in the instructions for the LLRnet server over at CRUS, but not here. When we got the second LLRnet server here, I figured I wouldn't bother with the instructions for switching, since it originally seemed to be set up mainly so Carlos's Free-DC buddies could clean it up in a couple weeks and be done with it, though of course non-Free-DC help was welcomed.

I guess I should probably edit the post announcing the port 100 server to include a link to the instructions in my previous post here.

[henryzz]

people seem to have got an overdose of this smilie since the rest became accessable only from the more page

[mdettweiler]

Originally Posted by henryzz

people seem to have got an overdose of this smilie since the rest became accessable only from the more page :-) :-) :-) :-) :-) :-) :-)

Actually, for the most part, I just make my smilies by typing things like :smile: and :wink:, etc. (That way, I don't have to take my hands off the keyboard and move them over to my mouse just to insert a smiley, even when all the smilies were available on the main post-composition page.)

But yes, it would be good to have the rest of the smilies--or at least the most commonly used ones, such as , , , , , , , etc.--back on the main post-composition page. By now I've memorized the codes for most of the ones I use frequently, but probably not everyone wants to go to that bother.

(Note: I had to replace most of the smilies in henryzz's quoted original post with :-), since otherwise I run into the 10-smiley limit and can't submit my post.)

[em99010pepe]

I'm seeing an increase of activity on server port 300 when it should be on 100 .
I'll clean the lower ranges by myself, I won't share the glory with you guys. Bye.

Carlos

[Mini-Geek]

Originally Posted by em99010pepe

I'm seeing an increase of activity on server port 300 when it should be on 100 .
I'll clean the lower ranges by myself, I won't share the glory with you guys. Bye.

Carlos

So...what server are we using for this rally anyway? (And which port corresponds with which ranges?) Your post suggests 100 only, but other discussion suggests both. I'm cornfused.
I think the new 24 hour rally should be split off into another thread...it's such a long thread. In the new thread (or in this one) the first post should be updated to show the updated info about the rally.
It's the same time (7 PM GMT this Sat) as the last one, right?

[mdettweiler]

Originally Posted by Mini-Geek

So...what server are we using for this rally anyway? (And which port corresponds with which ranges?) Your post suggests 100 only, but other discussion suggests both. I'm cornfused.
I think the new 24 hour rally should be split off into another thread...it's such a long thread. In the new thread (or in this one) the first post should be updated to show the updated info about the rally.
It's the same time (7 PM GMT this Sat) as the last one, right?

We're doing port 100 (lower range) only. We were discussing doing it on both servers, but Carlos wanted to get the port 100 server cleaned out and closed up (all those servers use a bit too much RAM on his machine), so we decided to do it on port 100.

As for splitting it off into another thread, I suppose we could do that. What does everyone else think?

[mdettweiler]

Originally Posted by em99010pepe

I'm seeing an increase of activity on server port 300 when it should be on 100 .
I'll clean the lower ranges by myself, I won't share the glory with you guys. Bye.

Carlos

Maybe it's just some people who want to do a "last blast" on the port 300 server before they switch to the port 100 server for the rally.

Don't worry, I'm sure we'll see everyone on the port 100 server by tomorrow afternoon. (I can't imagine anyone minding the fact that it's working on the lower range; remember, most of the first rally was done with the port 300 server still on the lower range, and nobody seemed to mind that.)