Rally Nov. 21st-23rd - Page 9

IronBits · 2008-11-23, 17:42

The proxy opens a new socket for each client that is connected to that proxy. (at once!) my emphasis
!!! crap !!!
That explains a LOT, I didn't know that part.
The proxy needs to be written differently so it behaves like the dnet pproxy server does.
Thanks for letting us know James!

Based on that, no proxy servers allowed to hit my server please without prior arrangements and testing.

proxy servers are your own personal servers to crunch reserved knpairs for all you super pharmers.

I guess it was never developed for server to server communications, and explains why the server can handle many many clients all at the same time without nary a hiccup.

Maybe we could get someone to look into that code and fix it so it behaves properly.

Lennart · 2008-11-23, 18:34

Quote:

Originally Posted by Beyond

The proxy opens a new socket for each client that is connectd to that proxy. On each failure to connect (send or recieve) it leaves the sockt connection open (there is no 'time_out" to drop the connection) then immediately opens another socket to re-try the connection, multiply this by the number of clients on the proxy and you can see what happens.

Multiply proxies will eliminate the problem if thy are all pointed at different servers/ports.

This is not specific becuse it is a proxy !
It is the same if i run many clients with 5-10 in cache.

/Lennart

mdettweiler · 2008-11-23, 19:22

Quote:

Originally Posted by Lennart

This is not specific becuse it is a proxy !
It is the same if i run many clients with 5-10 in cache.

/Lennart

But at least with a caching client, there's only 5-10 sockets that are being opened at once...rather than 1000.

I keep thinking that it wouldn't be too terribly hard to completely re-write the whole proxy system in Perl or something like that...if only I could ever figure out how the heck to program those pesky sockets.

For that matter, I could even re-write the *client* as a Perl wrapper that runs LLR 3.7.1c via command line, if I could figure out the socket stuff.

Anyone around here who's ever done any programming with sockets before?

mdettweiler · 2008-11-23, 19:22

Oh, and I forgot to mention in my last message: the rally officially ended 23 minutes ago.

henryzz · 2008-11-23, 19:26

Quote:

Originally Posted by mdettweiler

For that matter, I could even re-write the *client* as a Perl wrapper that runs LLR 3.7.1c via command line, if I could figure out the socket stuff.

that would be good that would get round the fact that llrnet hasnt been updated

mdettweiler · 2008-11-23, 19:33

Quote:

Originally Posted by henryzz

that would be good that would get round the fact that llrnet hasnt been updated

My thoughts exactly.

I've tried searching online for information on Perl sockets, but when I tried what I found, it didn't work too well. It also doesn't help too much that I don't know squat about the LUA programming language, so it's somewhat hard for me to even figure out the LLRnet client/server protocol by looking through the source code files...

As an attempt to get around my limited knowledge of LUA, I've also tried recording a number of LLRnet client/server communications with Wireshark (a network packet sniffer), but it was still pretty hard to figure out exactly what characters Wireshark was representing as "." (which is what Wireshark does for any non-printable character). And the LUA code, from what I can tell, was no help on this either. I'm sure it's got to be a lot simpler than I'm making it, though.

Which begs the question: is anyone around here familiar with LUA by any chance? Anyone who could help interpret the LLRnet source code (specifically, the sections of the files client.lua and llrserver.lua that define the protocol)? If only I could see the client/server communication code in some more easily understandable pseudocode, then I might be in business...

em99010pepe · 2008-11-23, 19:36

Results file from C443 uploaded to http://em99010pepe.googlepages.com/c443.

em99010pepe · 2008-11-23, 20:44

I think Lennart has to run llrnet for 24 hours to compensate the proxy issue...lol

gd_barnes · 2008-12-11, 05:35

Karsten,

Can you get us an official rally results page created for this Nov. 21st-23rd rally?

Max, does he need any help from you getting results tabulated?

I'd like to get this done before the weekend of the 19th-21st. I'm still debating about promoting a rally for that weekend.

I'll sticky this thread again until it is done so that I don't forget about it.

Gary

mdettweiler · 2008-12-11, 05:38

Quote:

Originally Posted by gd_barnes

Max, does he need any help from you getting results tabulated?

Not that I'm aware of; I haven't needed to do any filling in of gaps or other such stuff with the results files lately, so the results should all be exactly the same on Karsten's end as on mine.

kar_bon · 2008-12-11, 09:26

i'll try to do the race-page before the weekend.
i think, i have all results for the racing-time.

i have to write a tool to make the statistics automatically:
- counting pairs for user/team/server is simple
- displaying graph had to be done via spreadsheet

the only thing is to equalize the time-zones and the recalculation on the timings (-6 hours or so).

oohhh..... i know how to...

idea:

i quit my work here at the public office, retrain to an independent developer, appoint some experts and all is done!

2008-11-23, 17:42	#89
IronBits I ♥ BOINC! Oct 2002 Glendale, AZ. (USA) 3×7×53 Posts	The proxy opens a new socket for each client that is connected to that proxy. (at once!) my emphasis !!! crap !!! That explains a LOT, I didn't know that part. The proxy needs to be written differently so it behaves like the dnet pproxy server does. Thanks for letting us know James! Based on that, no proxy servers allowed to hit my server please without prior arrangements and testing. proxy servers are your own personal servers to crunch reserved knpairs for all you super pharmers. I guess it was never developed for server to server communications, and explains why the server can handle many many clients all at the same time without nary a hiccup. Maybe we could get someone to look into that code and fix it so it behaves properly. Last fiddled with by IronBits on 2008-11-23 at 17:51

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Let's vote the black 21st move	LaurV	Game 1 - ♚♛♝♞♜♟ - Shaolin Pirates	3	2013-03-09 16:09
December 21st 2012	jasong	jasong	11	2012-12-19 00:28
End of the world May 21st, 2011?	jasong	Lounge	67	2011-05-30 04:15
Rally Jan. 23rd-25th	gd_barnes	No Prime Left Behind	89	2009-01-25 22:59
LLRnet server rally port 300 May 23rd-25th	gd_barnes	No Prime Left Behind	172	2008-06-04 19:21

2008-11-23, 19:22	#92
mdettweiler A Sunny Moo Aug 2007 USA (GMT-5) 3×2,083 Posts	Oh, and I forgot to mention in my last message: the rally officially ended 23 minutes ago.

2008-11-23, 19:36	#95
em99010pepe Sep 2004 2830₁₀ Posts	Results file from C443 uploaded to http://em99010pepe.googlepages.com/c443.

2008-11-23, 20:44	#96
em99010pepe Sep 2004 5416₈ Posts	I think Lennart has to run llrnet for 24 hours to compensate the proxy issue...lol

2008-12-11, 05:35	#97
gd_barnes May 2007 Kansas; USA 10395₁₀ Posts	Karsten, Can you get us an official rally results page created for this Nov. 21st-23rd rally? Max, does he need any help from you getting results tabulated? I'd like to get this done before the weekend of the 19th-21st. I'm still debating about promoting a rally for that weekend. I'll sticky this thread again until it is done so that I don't forget about it. Gary

2008-12-11, 09:26	#99
kar_bon Mar 2006 Germany 2³·3·11² Posts	i'll try to do the race-page before the weekend. i think, i have all results for the racing-time. i have to write a tool to make the statistics automatically: - counting pairs for user/team/server is simple - displaying graph had to be done via spreadsheet the only thing is to equalize the time-zones and the recalculation on the timings (-6 hours or so). oohhh..... i know how to... idea: i quit my work here at the public office, retrain to an independent developer, appoint some experts and all is done!