LLRnet servers for NPLB

[MyDogBuster]

We aren't going to add any more work to G4000 with this drive.

Let me know if and when I have to move again. I kinda like it over here though. I see no problem with 2 servers running. It gives everyone another choice of where to put their assets. 2 servers doing different slices of the pie (not the same drive) would be nice.

[gd_barnes]

Originally Posted by MyDogBuster

Let me know if and when I have to move again. I kinda like it over here though. I see no problem with 2 servers running. It gives everyone another choice of where to put their assets. 2 servers doing different slices of the pie (not the same drive) would be nice.

Cool. That works for us.

If you like different servers on different drives, you'll like our n>600K effort. We'll have 3 servers; one each on k=400-600, k=600-800, and 800-1001. Likely they'll be at different n-ranges at different times. Initially we'll push k=400-600 more in rallies and at other times since they are lower k's so they will generally be at a higher search range but people will still be free to search any of the k-ranges.

Also, we'll have k=1005-2000 for n=50K-100K to start with and n=350K-500K later on as well as our double-check drive for k<=1001 and n=100K-260K. I stated in an RPS thread that I'll take k=27-31 from the double-check drive after our 1st drive is complete, due to the missing primes found for k=31 for n>400K.

Completing all k=300-1001 to n=600K is just the beginning!


Gary

[em99010pepe]

I will stay on G4000 for three, four days, then I have to finish my manual range and help C443.

BTW, Glenn will only join us with more power if we do a rally.

[MyDogBuster]

If you like different servers on different drives, you'll like our n>600K effort. We'll have 3 servers; one each on k=400-600, k=600-800, and 800-1001. Likely they'll be at different n-ranges at different times. Initially we'll push k=400-600 more in rallies and at other times since they are lower k's so they will generally be at a higher search range but people will still be free to search any of the k-ranges.

Also, we'll have k=1005-2000 for n=50K-100K to start with and n=350K-500K later on as well as our double-check drive for k<=1001 and n=100K-260K. I stated in an RPS thread that I'll take k=27-31 from the double-check drive after our 1st drive is complete, due to the missing primes found for k=31 for n>400K.

Great ideas. And here I was thinking about resting a bit on our laurels.
Looks like enough work to choke a horse.

[mdettweiler]

Originally Posted by MyDogBuster

Okay, right out of an episode of the Twilight Zone, da da da da,
I tried again and this time they were accepted. So I guess I'm all caught up.

I haven't got a clue as to why they worked the second time and not the first.

Hmm...that's odd. I don't know why it would do that. Anyway, now that you've got your results in, I'll go and change jobMaxTime to 5 days again...

[mdettweiler]

Hi all,

In case you're all wondering what this new "port 8000/NPLB 7th Drive" thing that's just showed up on the http://nplb-gb1.no-ip.org/llrnet/ status page is, that's a new, empty server that we will eventually be loading work for k=800-1001, n>600K into after we're done with the 1st Drive, as Gary, David and I had discussed via PM. I just got all the behind-the-scenes stuff for it all taken care of ahead of time so that when the time comes, all I have to do is pop in the knpairs and let 'er rip.

Anyway, just wanted to let you guys know, because otherwise I'm sure there would have been somebody posting a message wondering what the heck this new thing was.

In the meantime, this has helped me spot a previously-unnoticed bug that causes the status page to display "-1 remaining knpairs" for a brand-new server.

Max

[mdettweiler]

Hmm...G4000 apparently crashed at least two times within the past hour or so. Here's what the console output said:

llrnet: net.cxx:138: static void* net_Server_t::connection_thread(void*): Assertion `thd->socket >= 0' failed.
net_signal called with code 6
shutting down listening socket
accept: errorno=22
waiting thread #0 exits
shutdown: errno 9
thread #2 exited

The console output is specifically different than last time I had problems with the server continually crashing; though it may very well be a corrupted executable again. I'll swap in a fresh binary and see if that fixes anything.

David, Carlos--any of you server gurus have any idea what's happening here?

[mdettweiler]

Originally Posted by mdettweiler

Hmm...G4000 apparently crashed at least two times within the past hour or so. Here's what the console output said:

llrnet: net.cxx:138: static void* net_Server_t::connection_thread(void*): Assertion `thd->socket >= 0' failed.
net_signal called with code 6
shutting down listening socket
accept: errorno=22
waiting thread #0 exits
shutdown: errno 9
thread #2 exited

The console output is specifically different than last time I had problems with the server continually crashing; though it may very well be a corrupted executable again. I'll swap in a fresh binary and see if that fixes anything.

David, Carlos--any of you server gurus have any idea what's happening here?

Okay, I've got the binary swapped out. Hopefully it won't crash any more...

[mdettweiler]

Originally Posted by mdettweiler

Okay, I've got the binary swapped out. Hopefully it won't crash any more...

Hmm...it's still crashing. This is weird; I have absolutely no idea why it's doing this. The only thing I do know is that it seems to crash whenever it prunes the joblist and knpairs files.

Anyone know why this is happening?

[gd_barnes]

Originally Posted by MyDogBuster

Great ideas. And here I was thinking about resting a bit on our laurels.
Looks like enough work to choke a horse.

The pressure will definitely be off. I'll likely drop to about 6-7 quads on NPLB, which is what I had on it prior to the last 1-2 months, from my current 8-9 after the 1st drive is done. There are several CRUS efforts that I'd like to put more CPU power into.

NPLB and CRUS like most prime-search efforts are infinite projects. (CRUS was finite to start with, although extremely huge, when we were only including bases <= 32 but others and now me have started processing much larger bases because they are fun.) It's all a matter of how big of a piece we choose to process at any one time. We prefer to process large #'s of k's at once because there are great efficiency gains in processing and it's far easier to set goals for entire swaths of k and n-ranges. The challenge is keeping it fun for everyone so we have to vary our efforts enough and have enough different ones going at a time to make things interesting. At this particular point in time, we likely have less different efforts going at once then we will have for the foreseeable future.


Gary

[gd_barnes]

Originally Posted by mdettweiler

Hmm...it's still crashing. This is weird; I have absolutely no idea why it's doing this. The only thing I do know is that it seems to crash whenever it prunes the joblist and knpairs files.

Anyone know why this is happening?

Is the server completely down?