LLRnet servers for NPLB

[mdettweiler]

Originally Posted by gd_barnes

The danger about having Sherri go by and turn it on is that is how I fried a motherboard before myself. That is...the fact that it shut itself down was a 'warning' sign that something was amiss. I turned it back on, started crunching again and a few days later it went off again. I did it again and it went off again in about a day. That was it...it had fried itself at that point.

I'm not going to turn it on and start crunching on it until I verify temps and stuff. Well, I suppose I could have her turn it on but not start crunching on it (assuming it will even come on; which I suspect there is < 50% chance of). Since the server actually does no crunching, it shouldn't heat up the machine. I'll see if she can do it. I hate to burden her with messing with stuff again though. She's already been by my house twice to make sure everything is OK and I told her that should be enough. Oh well, I'll see what I can do.

If the machine won't turn on, yes, I will swap hard drives with another machine after I get back to the coolest running machine so that we can make sure the server is on likely the most stable machine that I have. Actually, I've done that twice already based on the priority of stuff that was running on a machine that went down, even after you got the remote access set up. You just didn't know it. lol

Stupid machines!


Gary

Ah, I see...in that case, don't worry about getting it restarted just yet. Since all the clients working on G4000 have been moved to other servers by now, it shouldn't hurt if it's down just a few more days--rather than taking the risk of frying yet another motherboard. The main significant thing that is waiting on the server coming back online is about 500 results that I've got sitting on my computer that I had pulled down from G4000 and crunched with manual LLR, but that can wait.

[em99010pepe]

Looks like Bliss is back. Hey Flatlander, only 10 tests per hour? Those 6 cores run too slow...lol I can't see either Henryzz and Max on the stats...

[em99010pepe]

NPLB LLRnet server #2 (updated 2008-12-02 08:00 GMT):
maintained by em99010pepe
Short identification: C443
server = "nplb.dynip.telepac.pt"
port = 443
k-range: 401 <= k <= 1001
n-range: 577K-578K, 587.4K-588K, 598K-600K
currently processing at n= ~598.23K

[MyDogBuster]

The main significant thing that is waiting on the server coming back online is about 500 results that I've got sitting on my computer that I had pulled down from G4000 and crunched with manual LLR, but that can wait.

I also have about 300 results that were in my caches and complete. I saved the tosend files before I switched to IB400. Will their reservations expire when you bring GB4000 back up?

[henryzz]

Originally Posted by em99010pepe

Looks like Bliss is back. Hey Flatlander, only 10 tests per hour? Those 6 cores run too slow...lol I can't see either Henryzz and Max on the stats...

i was finishing a gnfs factorization this morning in safe mode so i had access to more memory so i didnt have access to the internet then

out of interest i started the postprocessing using ggnfs which worked straight away once i had gone into safe mode to free up memory
once i finally got to starting the linear algebra its estimate was about 10.5 hours with one core
if i had done that it would be still going
i thought that was way too long to bother with as msieve is faster
i ran msieve and it wouldnt solve the matrix as the weight was way too low about 15 per cycle
i fiddled for ages and eventurely got it to do it by removing some of the relations so it was less oversieved and doing the linear algebra with msieve took 1 hour and 40 mins with four cores instead of one
somewhat an improvement over ggnfs although not being able to rediculously oversieve with msieve drives me crazy sometimes

[IronBits]

We set a record yesterday with 10,583 knpairs completed.

[gd_barnes]

Originally Posted by MyDogBuster

I also have about 300 results that were in my caches and complete. I saved the tosend files before I switched to IB400. Will their reservations expire when you bring GB4000 back up?

Yes, technically, they would expire. BUT...no, as long as they are not handed out to someone else, you should be OK. What I'll do is tell you exactly when I expect to bring it back up and coordinate it with you being online. When I bring it back up, you can then connect to the server and it will immediately send those results and not hand them back out to someone else or even "funnier"; try to hand them back out to YOU again. (lol)

Max, is that your understanding about what will happen if Ian connects to port 4000 after I bring it up and no one else is connected to it?


Gary

[MyDogBuster]

When I bring it back up, you can then connect to the server and it will immediately send those results and not hand them back out to someone else or even "funnier"; try to hand them back out to YOU again. (lol)

What I'll do is gather all of them up into 1 tosend file so that I only have to connect once when it comes back up. That way it should go lots faster and not have to wait till I get to all the cores.

[mdettweiler]

Originally Posted by MyDogBuster

I also have about 300 results that were in my caches and complete. I saved the tosend files before I switched to IB400. Will their reservations expire when you bring GB4000 back up?

Originally Posted by gd_barnes

Yes, technically, they would expire. BUT...no, as long as they are not handed out to someone else, you should be OK. What I'll do is tell you exactly when I expect to bring it back up and coordinate it with you being online. When I bring it back up, you can then connect to the server and it will immediately send those results and not hand them back out to someone else or even "funnier"; try to hand them back out to YOU again. (lol)

Max, is that your understanding about what will happen if Ian connects to port 4000 after I bring it up and no one else is connected to it?


Gary

Originally Posted by MyDogBuster

What I'll do is gather all of them up into 1 tosend file so that I only have to connect once when it comes back up. That way it should go lots faster and not have to wait till I get to all the cores.

As Gary said--yes, they probably would be expired by the time I get the server started up, but I'm planning to, before restarting the server, set jobMaxTime to 20 days or so to give everyone a chance to return their old results. Then, after Ian and I can both confirm that we've tied up any loose ends, I can change it back to the normal settings of 5 days.

[gd_barnes]

Originally Posted by mdettweiler

As Gary said--yes, they probably would be expired by the time I get the server started up, but I'm planning to, before restarting the server, set jobMaxTime to 20 days or so to give everyone a chance to return their old results. Then, after Ian and I can both confirm that we've tied up any loose ends, I can change it back to the normal settings of 5 days.

How about setting it to 3 days like port 400 after we get Ian's (and other's) already processed results returned to the server?

[mdettweiler]

Originally Posted by gd_barnes

How about setting it to 3 days like port 400 after we get Ian's (and other's) already processed results returned to the server?

Well, I usually like to have it set to 5 days so that I can manually cache k/n pairs from the server and crunch it manually. So far I haven't had any problems with bottlenecked k/n pairs; though, yes, I will keep an eye on things and be sure to decrease the time a bit if it seems necessary.