LLRnet servers for NPLB

[gd_barnes]

Max,

Since there are no primes found yet on port 6000, the link to "All primes found so far on this server" here gives an HTTP 404 page not found error.

Can you instead code something that will show a message about no primes found yet?


Thanks,
Gary

[gd_barnes]

At the current rate, David's port 2000 will dry out while I'm asleep. So I'm going to start all of my machines running 4 instances each of his port 8000 while leaving 4 instances of port 2000 running so that my machines don't sit idle. This means that each instance will run half as fast.

So instead of port 2000 drying out at 2-4 PM GMT, it will likely dry out at 8-10 PM GMT Thursday (3-5 PM EST).

Once dry, I'll put at least 4 cores on new port 6000 to keep the 11th drive moving while the bulk of my machines work towards drying out David's final port 8000 for the 10th drive.

[vaughan]

Will IB port 2000 get work in the future or is it retired? I read the post above about changing to Gary's server on Port 6000. Is this permanent?

[mdettweiler]

Originally Posted by vaughan

Will IB port 2000 get work in the future or is it retired? I read the post above about changing to Gary's server on Port 6000. Is this permanent?

Yes, it's permanent. In preparation for the imminent move to Gary's new server box, we're drying out all the IB servers and replacing them with GB ones. That way all we have to do is pick up and move them from humpford, the current interim server box, when we're ready to do the big move.

BTW, for those interested, the new server (christened "jeepford", in deference to some facetious remarks from Lennart on that a while back) is all ready to roll. While we haven't begun transferring any of the stuff to it, the OS is installed and it's been crunching happily for the last week or so. Once all the IB servers have finished drying, we'll begin moving the DB and main website over.

[gd_barnes]

Originally Posted by vaughan

Will IB port 2000 get work in the future or is it retired? I read the post above about changing to Gary's server on Port 6000. Is this permanent?

Yes, that is permanent. We are in the process of retiring all of Ironbits servers. He will no longer be doing DC projects or at most, won't be hosting servers for them. Port 8000 is the last one and it will be retired in favor of a new port on one of my machines for the 10th drive. Max and I still need to decide whether we want it to be PRPnet or LLRnet.

Here are some things I've done so far to make my servers more stable:

Swapped out the bad hard drive from before.
Swapped out a bad ethernet switch (splitter box, I call it).

Stuff done previously:
Upgraded to a commercial internet acct. to add more bandwith and get a fixed IP address.
Have put server machine running on a heavy-duty UPS backup.

I still need to follow up on getting the fixed IP address in place but we have not had problems with that in many months.

Also in the works 1-2 weeks after all of the servers on my machine is to move them all over to a brand new machine with high-end parts; including cooling, memory, and hard drive.

Although I can't make any 100% guarantees, I feel confident in saying that my servers are much more robust than before with more checks and balances in place.

Once the final 2 things are done, I'll feel confident running a heavy-load LLRnet rally on my servers.

I'd like to have people give the gb servers another chance. For those that aren't quite comfortable with PRPnet yet, we are still running LLRnet servers. They are port 6000 on the 11th drive and port 4000 on the 7th drive.


Gary

[mdettweiler]

We now have LLRnet G8000 online to replace IB8000, which will dry out sometime within the next 24 hours. As it nears completion, those of you who wish to continue crunching this range can move to the new server (except Gary, of course, who's supposed to at least one of his cores on the old server until it's fully dried to avoid stragglers ).

[gd_barnes]

Max,

Thanks for taking care of that. One thing on the web page: In the link where it shows all primes found for this server, there is a bunch of k=300-400 primes listed. Can you remove all of those?

One more thing: Since there will be no primes, please make it display a blank page or something that says "no primes found yet". On port 6000, it would just give a "404 - no page found" error or something like that before the first prime is found. I'm thinking it's just a matter of leaving the primes.txt file there but leaving it blank (vs. having no primes.txt file).


Thanks,
Gary

[mdettweiler]

Originally Posted by gd_barnes

Max,

Thanks for taking care of that. One thing on the web page: In the link where it shows all primes found for this server, there is a bunch of k=300-400 primes listed. Can you remove all of those?

Ah, I'd forgotten about that. Those are from port 8000's previous tenure as a server for the 2nd 6-k mini-drive on k=300-400. I'm a little reluctant to remove them since, after all, it is supposed to be a complete list, and keep in mind that in some ways (like in the DB, for example), this is the same server as before, just with a new workload. Having considered that, if you still want me to remove them, it's your decision.

[gd_barnes]

OK. Who's the wise guy who took the last ~600-700 pairs of Ironbits port 8000 and caused my machines to sit idle for 2 hours? lol Karsten, I've got my eye on you.

Anyway, I'll leave 1-2 cores connected to it to clean out any stragglers. GB port 6000 is about to get a serious workout. Here we go... :-) Anyone care to join me? If folks would like more k=1400-2000 for n>670K work, gb port 8000 is ready to rumble.

There is officially no more new work in Ironbits servers. David, if you happen to read this, thank you very much for your huge contribution to NPLB. Both your machines and technical knowledge were greatly appreciated!


Gary

[gd_barnes]

Since the port 8000 server is intended solely for use on the 10th drive, I'd prefer that they were removed. Can you copy them off somewhere? Once we get the individual-k drive going again (that is k=300-400) for n>1M, we'll probably runs some mini drives. What I'd like to have is a server dedicated to such mini drives. When we set that up, I think it'd be good to have those primes in it so that we have a one-stop place for all k=300-400 mini-drive primes.

On another note, Email is not working for port 6000. Is it working for any of the gb servers? Regardless, that should be top priority to get that working now that NPLB is fully dependent on gb servers.

A second priority: We need to have the hourly stats page updated hourly with info. from gb servers. As it is, it is only updated once/day. Although it shows hourly totals, you have to wait a day to see them. It would be good to be able to check on the status of one's machines at a glance to see if anything went astray without having to wait as much as a day.


Thanks,
Gary

[mdettweiler]

Originally Posted by gd_barnes

Since the port 8000 server is intended solely for use on the 10th drive, I'd prefer that they were removed. Can you copy them off somewhere? Once we get the individual-k drive going again (that is k=300-400) for n>1M, we'll probably runs some mini drives. What I'd like to have is a server dedicated to such mini drives. When we set that up, I think it'd be good to have those primes in it so that we have a one-stop place for all k=300-400 mini-drive primes.

Okay, I'll figure something out within the next day or so so that we can hang on to the old primes without having them in the same file.

On another note, Email is not working for port 6000. Is it working for any of the gb servers? Regardless, that should be top priority to get that working now that NPLB is fully dependent on gb servers.

Ah, whoops, looks like the DB didn't have email notification turned on for that port. I'd forgotten that it sets it to "off" by default for newly detected ports, so in fact I had to fix it for all the respective new GB ports. Should be all set now, though primes found before I flipped the switch will not have emails sent retroactively.

A second priority: We need to have the hourly stats page updated hourly with info. from gb servers. As it is, it is only updated once/day. Although it shows hourly totals, you have to wait a day to see them. It would be good to be able to check on the status of one's machines at a glance to see if anything went astray without having to wait as much as a day.

That's only for the PRPnet servers, right? I'm pretty sure that the LLRnet ones update hourly. If they aren't updating hourly, then something wrong with that. (The PRPnet servers, on the other hand, only update daily at this point...still working on that.)

I'm not sure why the LLRnet servers aren't updating hourly. The status page (and therefore the interim results) are being updated every 15 minutes as usual, so the problem must be on the DB end. I'll email Dave about it.