LLRnet and PRPnet servers for automated LLR

[mdettweiler]

Originally Posted by Lennart

I have emailed Max that I will put some more on the server. I see now that it will soon dry out so I may have to move them for a while.

Lennart

Okay, let 'er rip. I've loaded an n=500 range this time (480501-480999) so it should last longer--though I'll watch it closely. (I'll need more sieve file to add anything more, though, as I mentioned in the LLR reservation thread.)

[mdettweiler]

Apologies for the recent ~3 hour downtime on PRPnet--I had forgotten to set maxclients=1000 in the prpserver.ini file, so Lennart's hordes overloaded the server on the default of 10. It appears that we may have also run into the dreaded "too many connections" bug (exacerbated by the anemic maxclients= figure), but I'm not entirely positive on that. Either way, though, the server was down for 3 hours and I probably could have avoided it had I paid more attention when initially setting up the server.

I've now fixed the maxclients= setting and restarted the server. Hopefully we won't run into any more problems now.

[mdettweiler]

There's another power outage at Gary's house right now, so the whole noprimeleftbehind.net site and all the LLRnet/PRPnet servers on it are down.

Man, we're getting them outages right and left now...usually Gary will go for months without any power outages, and even then they're mostly just <1 minute flickers, which the UPS can handle fine. Actually, the UPS should be able to handle the big ones as well--it can keep the computer and router going for up to ~1.25 hours. The problem seems to be on the ISP end: the internet goes kapooey every time the power cuts out.

[mdettweiler]

Originally Posted by mdettweiler

There's another power outage at Gary's house right now, so the whole noprimeleftbehind.net site and all the LLRnet/PRPnet servers on it are down.

Man, we're getting them outages right and left now...usually Gary will go for months without any power outages, and even then they're mostly just <1 minute flickers, which the UPS can handle fine. Actually, the UPS should be able to handle the big ones as well--it can keep the computer and router going for up to ~1.25 hours. The problem seems to be on the ISP end: the internet goes kapooey every time the power cuts out.

...and we're back! That was fast, fortunately. And thanks to the UPS, even though the server was inaccessible for a little while in there, it picked up right where it left off when the internet came back online.

[Lennart]

Originally Posted by mdettweiler

No problem--that's the nice thing about PRPnet, you can dry out one server and nobody goes idle.

Do you by chance know how long you'll be staying on the TPS server with this number of cores? I can then calculate an approximate figure for how much work will need to be loaded in to cover that time frame.

I will not have many left now before next weekend

Lennart

[mdettweiler]

Oddball, I see the LLRnet server is going to run out of work in a few minutes. Do you want me to load in a little more work (say, n=481501-481600)?

[Oddball]

Originally Posted by mdettweiler

Oddball, I see the LLRnet server is going to run out of work in a few minutes.

I just finished drying it up

Do you want me to load in a little more work (say, n=481501-481600)?

It looks like LLRnet is less popular than PRPnet, so you can load n=484900-484999 instead of n=481501-481600.

[mdettweiler]

Originally Posted by Oddball

It looks like LLRnet is less popular than PRPnet, so you can load n=484900-484999 instead of n=481501-481600.

Done.

[mdettweiler]

The servers are down right now--this time it's actually not a power outage, but rather a strange double outage of phone and internet. (Gary's internet is cable-based, so one wouldn't expect them to necessarily go out together.) Hopefully it will be back up soon.

[Oddball]

Max, could you replace n=481400-481500 on the PRPnet server with these candidates?

http://www.sendspace.com/file/4asdih

gribozavr just posted that file a short time ago, and it has a greater sieve depth.

Also, the LLRNet server has dried out, so it'll be nice if you load n=484000-484899 into port 12050.

[mdettweiler]

Originally Posted by Oddball

Max, could you replace n=481400-481500 on the PRPnet server with these candidates?

http://www.sendspace.com/file/4asdih

gribozavr just posted that file a short time ago, and it has a greater sieve depth.

Also, the LLRNet server has dried out, so it'll be nice if you load n=484000-484899 into port 12050.

Done. The current load of each server is now:

Port 12000 (PRPnet)
n=481000-481399 from the 5 September sieve file (in progress)
n=481400-481500 from the 1 October sieve file

Port 12050 (LLRnet)
n=484900-484999 from the 5 September sieve file (in progress--1 pair left)
n=484000-484899 from the 1 October sieve file

Note that for LLRnet I listed the ranges seemingly "out of order". This reflects the exact order in which the server "sees" them and is due to a slight difference in how PRPnet and LLRnet handle these. PRPnet is set to sort by n, so it looks in the database for the lowest available n (and the lowest available k within that). LLRnet, on the other hand, reads pairs in the exact order they're listed in knpairs.txt. In this case, when adding the new range I concatenated the new pairs onto the existing knpairs.txt, so the one last pair from the old range actually comes first (even though it's higher n-wise). This will be reflected on the web page's first/last pair display (at least until that one last pair from the old range is completed).

Confused yet? (It's 3:30 AM here so please pardon me if this wasn't explained as well as it could be.)