Team drive #5: k=400-600 n=600K-1M - Page 10

kar_bon · 2010-08-18, 12:49

There's work for about a day left on Port 9000, ~11000 pairs to test!

mdettweiler · 2010-08-18, 16:51

Quote:

Originally Posted by kar_bon

There's work for about a day left on Port 9000, ~11000 pairs to test!

Thanks for the heads-up. Reserving 890K-900K for port 9000.

gd_barnes · 2010-08-18, 18:16

Better would have been an n=3K or 5K range since the rally will end shortly after the server would have dried without adding pairs. We'll see how it goes in the near future.

mdettweiler · 2010-08-18, 22:30

Quote:

Originally Posted by gd_barnes

Better would have been an n=3K or 5K range since the rally will end shortly after the server would have dried without adding pairs. We'll see how it goes in the near future.

I thought about that; however, as I mentioned earlier in a PM, if we don't encounter the elusive "too many connections" PRPnet bug before the rally ends, then I'm going to put out a request for some of the bigger contributors to come back a few days later to try to repeat the high-load conditions and hopefully trigger the bug. (PrimeGrid will be doing a 48-hour rally starting Friday, so I'll wait until that's over.) If we end up doing that, then we should be able to clean through much of the glut of work in the server.

Note that if it becomes necessary, it's not hard for me to remove a portion of the range from the server. I don't even have to stop the server to do it--it's just one quick MySQL command.

gd_barnes · 2010-08-18, 22:47

OK sounds good.

Lennart · 2010-08-18, 23:10

Quote:

Originally Posted by mdettweiler

I thought about that; however, as I mentioned earlier in a PM, if we don't encounter the elusive "too many connections" PRPnet bug before the rally ends, then I'm going to put out a request for some of the bigger contributors to come back a few days later to try to repeat the high-load conditions and hopefully trigger the bug. (PrimeGrid will be doing a 48-hour rally starting Friday, so I'll wait until that's over.) If we end up doing that, then we should be able to clean through much of the glut of work in the server.

Note that if it becomes necessary, it's not hard for me to remove a portion of the range from the server. I don't even have to stop the server to do it--it's just one quick MySQL command.

If you shall try to get the error load some smaller n into the server and you have a much bigger chance to get the error again.

Lennart

gd_barnes · 2010-08-19, 03:19

Quote:

Originally Posted by Lennart

If you shall try to get the error load some smaller n into the server and you have a much bigger chance to get the error again.

Lennart

That had crossed my mind and it is a very good idea. We'd like to leave port 9000 with the pairs that it currently has but we could set up a new server for one of the drives. The issue is having the server set up to go properly into the stats database within just a few days.

Max and Dave,

I think it would be good to run on the 11th drive for a while, which includes k=2000-3000 for n=~590K. With so many k, we could load an n=5K range into a PRPnet server and it could potentially last for a long time, even with all of us on it. Could you guys have all of the behind-the-scenes work set up for such a thing?

Max,

If we do the above, I think what I'd like to do is delete n=585K-590K out of LLRnet port 6000 and load that range into the new PRPnet server. That would take little additional time on my part and would leave less of a gap than loading n=590K-595K into the PRPnet server.

Lennart,

If we do this, I'll put most of my machines on it. That will add to the load and hopefully recreate the problem.

Gary

mdettweiler · 2010-08-19, 04:12

Quote:

Originally Posted by gd_barnes

That had crossed my mind and it is a very good idea. We'd like to leave port 9000 with the pairs that it currently has but we could set up a new server for one of the drives. The issue is having the server set up to go properly into the stats database within just a few days.

Max and Dave,

I think it would be good to run on the 11th drive for a while, which includes k=2000-3000 for n=~590K. With so many k, we could load an n=5K range into a PRPnet server and it could potentially last for a long time, even with all of us on it. Could you guys have all of the behind-the-scenes work set up for such a thing?

Max,

If we do the above, I think what I'd like to do is delete n=585K-590K out of LLRnet port 6000 and load that range into the new PRPnet server. That would take little additional time on my part and would leave less of a gap than loading n=590K-595K into the PRPnet server.

Lennart,

If we do this, I'll put most of my machines on it. That will add to the load and hopefully recreate the problem.

Gary

It looks like we may not need to do that after all; as of now the "too many connections" bug is in the process of happening. See the rally thread for more details.

gd_barnes · 2010-08-19, 05:19

I'm glad to have that out of the way.

gd_barnes · 2010-08-20, 22:30

Reserving n=892K-900K for port 3000.

Max, can you delete n>=892001 from port 9000? Thanks.

Vaughan is going to hang around on port 3000 for a little while and I will stay on there with a couple of quads for a little over a week. I'll maintain 4 to 6 quads on it until we get to n>890K to fill in the gap.

mdettweiler · 2010-08-20, 23:56

Quote:

Originally Posted by gd_barnes

Max, can you delete n>=892001 from port 9000? Thanks.

Done.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Team drive #7 k=800-1001 n=600K-1M	gd_barnes	No Prime Left Behind	127	2011-07-15 14:25
Team drive #6 k=600-800 n=600K-1M	gd_barnes	No Prime Left Behind	89	2011-03-10 12:34
Team drive #1: k=400-1001 n=333.2K-600K	gd_barnes	No Prime Left Behind	675	2009-02-24 16:37
Team drive #3: k=300-400 n=260K-600K	gd_barnes	No Prime Left Behind	255	2008-11-12 10:43
Team drive #4, 15 k's < 300 for n=600K-1M	gd_barnes	No Prime Left Behind	38	2008-10-22 16:20

2010-08-18, 12:49	#100
kar_bon Mar 2006 Germany 2³×3×11² Posts	There's work for about a day left on Port 9000, ~11000 pairs to test!

2010-08-18, 18:16	#102
gd_barnes May 2007 Kansas; USA 3³·5·7·11 Posts	Better would have been an n=3K or 5K range since the rally will end shortly after the server would have dried without adding pairs. We'll see how it goes in the near future.

2010-08-18, 22:47	#104
gd_barnes May 2007 Kansas; USA 3³·5·7·11 Posts	OK sounds good.

2010-08-19, 05:19	#108
gd_barnes May 2007 Kansas; USA 3³×5×7×11 Posts	I'm glad to have that out of the way.

2010-08-20, 22:30	#109
gd_barnes May 2007 Kansas; USA 3³×5×7×11 Posts	Reserving n=892K-900K for port 3000. Max, can you delete n>=892001 from port 9000? Thanks. Vaughan is going to hang around on port 3000 for a little while and I will stay on there with a couple of quads for a little over a week. I'll maintain 4 to 6 quads on it until we get to n>890K to fill in the gap.