IRCLogs • ArmaNelgTron.tk

Searching from 2023-09-05 00:00:00 to 2023-09-05 23:59:59.999999.
Query completed in 1.18 seconds
[2023-09-05 04:33:00] <-- Armanelgtron has quit (No Ping reply in 180 seconds.)
[2023-09-05 04:37:01] <-- Armanelgtron has quit (No Ping reply in 270 seconds.)
[2023-09-05 04:37:33] --> Armanelgtron has joined the channel
[2023-09-05 04:37:33] -!- Topic for #armagetron is "Armagetron Advanced | http://www.armagetronad.org/ | Welcome to IRC"
[2023-09-05 04:37:33] -!- Topic set by ChanServ!services@services.oftc.net on 2022-12-21 00:36:08 UTC
[2023-09-05 04:37:35] -!- larich.oftc.net set mode #armagetron +nt
[2023-09-05 04:37:35] -!- Channel #armagetron created on 2021-04-20 19:56:37 UTC
[2023-09-05 04:38:17] --> Armanelgtron has joined the channel
[2023-09-05 04:38:18] -!- molybdenum.libera.chat set mode #armagetron +nt
[2023-09-05 04:38:18] -!- Channel #armagetron created on 2021-05-20 17:23:14 UTC
[2023-09-05 06:22:07] --> monr0e has joined the channel
[2023-09-05 07:17:10] <Lucifer_arma> heh, I've got my cluster running probably around half strength right now, and postgres is taking it like a pro.  MySQL would have melted down already.
[2023-09-05 07:17:11] <armagetronbridge> 10irc:Lucifer_arma| heh, I've got my cluster running probably around half strength right now, and postgres is taking it like a pro.  MySQL would have melted down already.
[2023-09-05 07:18:14] <armagetronbridge> 10irc:Lucifer_arma| I'm not sure it's actually running at full strength though, because I think I screwed up the rate limiter on one of the jobs.  It should be running more jobs more quickly, but it's not.  The first job in the dependency list, the one the other depend on, should be running 25 times/minute, but it seems to be running less often
[2023-09-05 07:18:14] <Lucifer_arma> I'm not sure it's actually running at full strength though, because I think I screwed up the rate limiter on one of the jobs.  It should be running more jobs more quickly, but it's not.  The first job in the dependency list, the one the other depend on, should be running 25 times/minute, but it seems to be running less often
[2023-09-05 07:18:40] <armagetronbridge> 08discord:delinquent| sure you got enough resources? It is a pi after all
[2023-09-05 07:18:40] <armagetron-bridge> 08discord:delinquent| sure you got enough resources? It is a pi after all
[2023-09-05 10:27:21] <Lucifer_arma> @delinquent: The database isn't a pi
[2023-09-05 10:27:22] <armagetronbridge> 10irc:Lucifer_arma| @delinquent: The database isn't a pi
[2023-09-05 10:28:02] <Lucifer_arma> it's a Intel(R) Pentium(R) Silver N5030 CPU @ 1.10GHz
[2023-09-05 10:28:02] <armagetronbridge> 10irc:Lucifer_arma| it's a Intel(R) Pentium(R) Silver N5030 CPU @ 1.10GHz
[2023-09-05 10:28:59] <armagetronbridge> 08discord:delinquent| oh that's fairly recent. Anuc?
[2023-09-05 10:29:00] <armagetron-bridge> 08discord:delinquent| oh that's fairly recent. Anuc?
[2023-09-05 10:29:06] <armagetronbridge> 08discord:delinquent| * a nuc?
[2023-09-05 10:29:06] <armagetron-bridge> 08discord:delinquent| * a nuc?
[2023-09-05 10:29:26] <Lucifer_arma>  CPU max MHz:         3100.0000
[2023-09-05 10:29:26] <armagetronbridge> 10irc:Lucifer_arma| CPU max MHz:         3100.0000
[2023-09-05 10:29:32] <Lucifer_arma> laptop
[2023-09-05 10:29:32] <armagetronbridge> 10irc:Lucifer_arma| laptop
[2023-09-05 10:29:50] <Lucifer_arma> some lenovo thing I got for like $200 :)
[2023-09-05 10:29:50] <armagetronbridge> 10irc:Lucifer_arma| some lenovo thing I got for like $200 :)
[2023-09-05 10:30:31] <Lucifer_arma> it outperforms the pis, but it's still, uh, resource-starved
[2023-09-05 10:30:31] <armagetronbridge> 10irc:Lucifer_arma| it outperforms the pis, but it's still, uh, resource-starved
[2023-09-05 10:31:08] <Lucifer_arma> I could still use a dedicated database machine with serious RAM, 8+ cores, etc.
[2023-09-05 10:31:08] <armagetronbridge> 10irc:Lucifer_arma| I could still use a dedicated database machine with serious RAM, 8+ cores, etc.
[2023-09-05 10:32:08] <armagetron-bridge> 08discord:delinquent| tbf I ran mailcow, apache, and MS-SQL on an elderly 2001 laptop for a good three years - right up until last year in fact. Prolly not a massive issue unless it's horribly starved for memory
[2023-09-05 10:32:08] <armagetronbridge> 08discord:delinquent| tbf I ran mailcow, apache, and MS-SQL on an elderly 2001 laptop for a good three years - right up until last year in fact. Prolly not a massive issue unless it's horribly starved for memory
[2023-09-05 10:33:08] <Lucifer_arma> if the app works out, I'll have the money for some serious hardware.  All that really matters right now is that the hardware I have will do the job, and it wasn't looking good with mysql.
[2023-09-05 10:33:08] <armagetronbridge> 10irc:Lucifer_arma| if the app works out, I'll have the money for some serious hardware.  All that really matters right now is that the hardware I have will do the job, and it wasn't looking good with mysql.
[2023-09-05 10:33:41] <Lucifer_arma> I'm still running a stock postgres install, there's room for optimization.  I'm also still writing all the cluster management code, so there's room to optimize my actual queries
[2023-09-05 10:33:41] <armagetronbridge> 10irc:Lucifer_arma| I'm still running a stock postgres install, there's room for optimization.  I'm also still writing all the cluster management code, so there's room to optimize my actual queries
[2023-09-05 10:36:20] <Lucifer_arma> my next big task is to find out why my scheduler takes up 99% CPU like it does.  If it's actually working, that's fine, but if it's doing all that in the network, then I need to do some network optimizations
[2023-09-05 10:36:20] <armagetronbridge> 10irc:Lucifer_arma| my next big task is to find out why my scheduler takes up 99% CPU like it does.  If it's actually working, that's fine, but if it's doing all that in the network, then I need to do some network optimizations
[2023-09-05 10:36:40] <Lucifer_arma> oh yeah, and write the prediction job.
[2023-09-05 10:36:40] <armagetronbridge> 10irc:Lucifer_arma| oh yeah, and write the prediction job.
[2023-09-05 10:37:39] <Lucifer_arma> Tonight, well, earlier today, I guess (tonight, since I woke up at like 5pm yesterday), I ran it for a few hours for the first time.  So now I'm looking for the kinds of problems that only happen when the app is running for a long time
[2023-09-05 10:37:39] <armagetronbridge> 10irc:Lucifer_arma| Tonight, well, earlier today, I guess (tonight, since I woke up at like 5pm yesterday), I ran it for a few hours for the first time.  So now I'm looking for the kinds of problems that only happen when the app is running for a long time
[2023-09-05 10:52:09] <-- Lucifer_arma has quit (Quit: Core dumped)
[2023-09-05 10:53:07] --> Lucifer_arma has joined the channel
[2023-09-05 14:13:13] <-- Armanelgtron has quit (Ping timeout: 240 seconds)
[2023-09-05 14:13:30] --> Armanelgtron has joined the channel
[2023-09-05 14:13:31] -!- molybdenum.libera.chat set mode #armagetron +nt
[2023-09-05 14:13:31] -!- Channel #armagetron created on 2021-05-20 17:23:14 UTC
[2023-09-05 16:07:32] <Z-Man> Gaa,  too much to read.
[2023-09-05 16:07:32] <Z-Man> Lucifer_arma: Sure a dictatorship is easier, but your dictator is a micromanager and can turn out to be the bottleneck. With the work stealing anarchy, well, fist of all you need a system where jobs get created by jobs; then the network becomes the limit. Not every worker needs to talk to every other; a hypercube connection should do fine (every worker gets a number and communicates only with those workers where the number differs by one 
[2023-09-05 16:07:32] <Z-Man> binary digit).
[2023-09-05 16:07:33] <armagetronbridge> 15irc:Z-Man| Gaa,  too much to read.
[2023-09-05 16:07:33] <armagetronbridge> 15irc:Z-Man| binary digit).
[2023-09-05 16:07:34] <armagetronbridge> 15irc:Z-Man| Lucifer_arma: Sure a dictatorship is easier, but your dictator is a micromanager and can turn out to be the bottleneck. With the work stealing anarchy, well, fist of all you need a system where jobs get created by jobs; then the network becomes the limit. Not every worker needs to talk to every other; a hypercube connection should do fine (every worker gets a number and communica <clipped message>
[2023-09-05 16:07:35] <armagetronbridge> 15irc:Z-Man| tes only with those workers where the number differs by one
[2023-09-05 16:10:39] <armagetronbridge> 15irc:Z-Man| What you describe as the shooter network communication is basically what they do, at least for "infinite" speed weapons, since Half Life/Counterstrike. What is sent and how it is interpreted by the server is a high art :) Early-ish Call of Duty games had an exploit where you could interrupt the physical network connection, all the other players would stand still and you could jus <clipped message>
[2023-09-05 16:10:39] <Z-Man> What you describe as the shooter network communication is basically what they do, at least for "infinite" speed weapons, since Half Life/Counterstrike. What is sent and how it is interpreted by the server is a high art :) Early-ish Call of Duty games had an exploit where you could interrupt the physical network connection, all the other players would stand still and you could just shoot them, then reconnect.  Because you shot them on your 
[2023-09-05 16:10:39] <Z-Man> client and the server trusted the client a little too much, that counted.
[2023-09-05 16:10:40] <armagetronbridge> 15irc:Z-Man| t shoot them, then reconnect.  Because you shot them on your
[2023-09-05 16:10:41] <armagetronbridge> 15irc:Z-Man| client and the server trusted the client a little too much, that counted.
[2023-09-05 16:14:10] <armagetronbridge> 15irc:Z-Man| Arma's ping uses three exponential decay averagers with varying decay rates and then picks the best one. It also measures ping variations, and if they are high, the faster decaying averager is used. It is very ad-hoc. Maybe just using the median of the three would be better.
[2023-09-05 16:14:10] <Z-Man> Arma's ping uses three exponential decay averagers with varying decay rates and then picks the best one. It also measures ping variations, and if they are high, the faster decaying averager is used. It is very ad-hoc. Maybe just using the median of the three would be better.
[2023-09-05 16:26:09] <armagetronbridge> 15irc:Z-Man| Network logging: There is wireshark, I suppose you can configure it to do what you want. But no, I did not use any of that, the basic network communication was rarely a problem. And for the higher level problems, having a network log would not have helped. I used selective printf logging for that, or visual debugging aids.
[2023-09-05 16:26:09] <Z-Man> Network logging: There is wireshark, I suppose you can configure it to do what you want. But no, I did not use any of that, the basic network communication was rarely a problem. And for the higher level problems, having a network log would not have helped. I used selective printf logging for that, or visual debugging aids.
[2023-09-05 16:48:27] <armagetronbridge> 15irc:Z-Man| Arma's ack/resend mechanism is basically what you described for your mechanism. We send packets, if no ack arrives until Ping * some factor + Ping variation * some factor later, we resend; the receiver always sends acks for received packets, even if it already got them (then they are just not processed again)
[2023-09-05 16:48:28] <Z-Man> Arma's ack/resend mechanism is basically what you described for your mechanism. We send packets, if no ack arrives until Ping * some factor + Ping variation * some factor later, we resend; the receiver always sends acks for received packets, even if it already got them (then they are just not processed again)
[2023-09-05 16:50:49] <armagetronbridge> 15irc:Z-Man| I think your "send relative positions to walls" idea would only do good for a) the non-90 degree turn case, where a slide may put you on a collision course sooner or later or b) when mazing against your own walls, where your client may still have them in a pre-slide position.
[2023-09-05 16:50:49] <Z-Man> I think your "send relative positions to walls" idea would only do good for a) the non-90 degree turn case, where a slide may put you on a collision course sooner or later or b) when mazing against your own walls, where your client may still have them in a pre-slide position.
[2023-09-05 16:52:16] <armagetronbridge> 15irc:Z-Man| Regular other player walls will not slide. The one thing that can change your distance to a wall comparing the client with the server is if a player just cuts you off, or you have prediction on and they turn out of the way, and I think compensating in those cases would do more harm than good.
[2023-09-05 16:52:16] <Z-Man> Regular other player walls will not slide. The one thing that can change your distance to a wall comparing the client with the server is if a player just cuts you off, or you have prediction on and they turn out of the way, and I think compensating in those cases would do more harm than good.
[2023-09-05 16:53:41] <-- monr0e has quit (Ping timeout: 250 seconds)
[2023-09-05 16:57:10] <armagetron-bridge> 12discord:zman_0| @northernscrub about the error video: Aside from packet loss, it could also be that the client refuses to accept whatever the server sends, because that goes through a whole "simulate again from the time the server sync was sent to now" phase where things can go wrong. Also, server syncs are only fully accepted if you have not turned again in the meantime; accepting them too earl <clipped message>
[2023-09-05 16:57:11] <armagetron-bridge> 12discord:zman_0| y would lead to multiple slides, possibly back and forth because the re-simulation stuff is hard and unreliable. In a perfect world, it would rewind the entire gamestate to do its simulation, but for simplicity, it only does so for your cycle and uses the stored metadata in the walls to check whether they already existed in the past.
[2023-09-05 16:57:11] <armagetronbridge> 12discord:zman_0| @northernscrub about the error video: Aside from packet loss, it could also be that the client refuses to accept whatever the server sends, because that goes through a whole "simulate again from the time the server sync was sent to now" phase where things can go wrong. Also, server syncs are only fully accepted if you have not turned again in the meantime; accepting them too earl <clipped message>
[2023-09-05 16:57:11] <armagetronbridge> 12discord:zman_0| y would lead to multiple slides, possibly back and forth because the re-simulation stuff is hard and unreliable. In a perfect world, it would rewind the entire gamestate to do its simulation, but for simplicity, it only does so for your cycle and uses the stored metadata in the walls to check whether they already existed in the past.
[2023-09-05 18:11:43] <-- Armanelgtron has quit (Ping timeout: 240 seconds)
[2023-09-05 18:11:59] --> Armanelgtron has joined the channel
[2023-09-05 18:12:00] -!- molybdenum.libera.chat set mode #armagetron +nt
[2023-09-05 18:12:00] -!- Channel #armagetron created on 2021-05-20 17:23:14 UTC
[2023-09-05 20:29:45] <-- Netsplit between *.net and *.split. Quit: Z-Man
[2023-09-05 20:30:27] --> Netsplit between *.net and *.split ended. Joined: Z-Man
[2023-09-05 22:56:31] <Lucifer_arma> Found the bottleneck.  Now my cluster is running at full strength.  :)
[2023-09-05 22:56:32] <armagetronbridge> 10irc:Lucifer_arma| Found the bottleneck.  Now my cluster is running at full strength.  :)
[2023-09-05 22:58:21] <Lucifer_arma> Z-Man: with only two workers, that's a bit of overkill.  The micromanager is fine, although if this cluster ever gets to a significant size, I'll probably get more hierachical
[2023-09-05 22:58:22] <armagetronbridge> 10irc:Lucifer_arma| Z-Man: with only two workers, that's a bit of overkill.  The micromanager is fine, although if this cluster ever gets to a significant size, I'll probably get more hierachical
[2023-09-05 23:11:11] <Lucifer_arma> now I'm back to the other question.  Do I setup the third Pi as a database server, a generic worker, or a scheduler?  If I make it a scheduler, I'll put a worker node on it, still, but I'll make the worker only use 2-3 CPUs instead of all 4
[2023-09-05 23:11:11] <armagetronbridge> 10irc:Lucifer_arma| now I'm back to the other question.  Do I setup the third Pi as a database server, a generic worker, or a scheduler?  If I make it a scheduler, I'll put a worker node on it, still, but I'll make the worker only use 2-3 CPUs instead of all 4
[2023-09-05 23:12:10] <Lucifer_arma> If I don't make it the database, then the database will stay on my laptop.  Otherwise, the laptop gets the scheduler.  I could put a worker on my laptop, too, and have it only use two CPUs.
[2023-09-05 23:12:10] <armagetronbridge> 10irc:Lucifer_arma| If I don't make it the database, then the database will stay on my laptop.  Otherwise, the laptop gets the scheduler.  I could put a worker on my laptop, too, and have it only use two CPUs.
[2023-09-05 23:13:30] <Lucifer_arma> If I make it the database, then I need a USB 4 adapter for the ssd that I have, so that it'll be able to access the filesystem at a high speed.  That's probably going to be the deciding factor.
[2023-09-05 23:13:30] <armagetronbridge> 10irc:Lucifer_arma| If I make it the database, then I need a USB 4 adapter for the ssd that I have, so that it'll be able to access the filesystem at a high speed.  That's probably going to be the deciding factor.
View entire month
DISCLAIMER: These logs of public chat may contain some content which may not be appropriate for all audiences. Use at your own risk.
Logs from 2006-2009 pulled from wrtlprnft
Format changes at: 2015-08-25, 2017-02-20, and 2020-03-23. Times (2015 and later) should be Eastern.
. [About this site] [Credits]