LLRnet and PRPnet servers for automated LLR

[pinhodecarlos]

Originally Posted by The Carnivore

Agreed, this may be preventing a lot of people from contributing. For now, the only solution is to use an older version of PRPNet and an older version of cLLR.

I think that you need to downgrade to at least PRPNet version 5.3.0 and cLLR version 3.8.14, but I could be wrong.

You’re right.

[MooMoo2]

It looks like the server will dry out soon. We're at k=~398K now.

Gary, could you load k=400K-440K into port 12000?

[gd_barnes]

done

[MooMoo2]

Originally Posted by gd_barnes

done

Thanks!

[pinhodecarlos]

How many candidates are between k=440K-1M?

[pinhodecarlos]

Can I request if k=440K-600K can be load onto the server because I have someone who will start hitting hard the server with 100 cores or more please.

[MooMoo2]

Originally Posted by pinhodecarlos

How many candidates are between k=440K-1M?

Approximately 3.5-3.6 million.

Originally Posted by pinhodecarlos

Can I request if k=440K-600K can be load onto the server because I have someone who will start hitting hard the server with 100 cores or more please.

That would be great! I'm getting around 1300 tests/day/core, so 100 cores would be ~130,000 tests/day. At that rate, we should find the twin within 6-7 months.

[gd_barnes]

Originally Posted by pinhodecarlos

Can I request if k=440K-600K can be load onto the server because I have someone who will start hitting hard the server with 100 cores or more please.

As soon as I see the cores hit the server, I'll load it up to k=480K. Since I just loaded it up a few days ago up to k=440K, there are currently ~230k tests remaining so that would keep him busy for > 1.5 days.

After loading it up to k=480K (3-4 days total work starting right now with 100+ cores hitting it), if the cores continue for at least a couple of days, then I'll go ahead and load it up to k=600K.

I will check the server 2-3 times a day over the next few days.

A huge number of cores hitting the server will cause a possible lag in response time so I'd like to know for sure that the person is serious before loading so many tests into it.

Perhaps in the next couple of days I'll remove all tests from the server for k<200K. That should help the response time if a significant number of cores are hitting it. The stats and primes will stay but just the tests/results would be removed, which is the lions share of data.

[MooMoo2]

There is something strange going on. The tests that are being assigned all have low k values, such as 675*2^498497-1 and 765*2^498483-1. Weren't those already tested a few years ago?

[gd_barnes]

I have deleted all tests and results from the server for k<200K. In doing so, I saved off the results/residues so that nothing is lost. This will help response time from the server, especially if we load an unusually large amount of tests in there.

In doing this, I found an anomoly: There were 1514 tests for k<100K that had "NOTDONE" in their residue. I have attached them here. Since the k-values were smallish they should have been handed out long ago. I don't know what could have caused it. No others like that were found for k<400K.

I felt like the best course of action was to save them off, delete them like all of the other k<200K, and reload them into the server.

MooMoo since you are the only one hitting the server at the moment it looks like you will get to run these stray tests since the server hands out smallest k-values first. After I reloaded them I see that they've already started being handed out.

After they have been run, I will save off their residues and delete them again like all other k<200K.

If our big searcher begins to hit the server hard, I'll go ahead and delete all k<300K also.

[MooMoo2]

Originally Posted by gd_barnes

I found an anomoly: There were 1514 tests for k<100K that had "NOTDONE" in their residue. I have attached them here. Since the k-values were smallish they should have been handed out long ago. I don't know what could have caused it. No others like that were found for k<400K.

I felt like the best course of action was to save them off, delete them like all of the other k<200K, and reload them into the server.

MooMoo since you are the only one hitting the server at the moment it looks like you will get to run these stray tests since the server hands out smallest k-values first. After I reloaded them I see that they've already started being handed out.

After they have been run, I will save off their residues and delete them again like all other k<200K.

I see. Thanks for the explanation!