Re: [hlds_linux] Poor TF2 performance on a dedicated server
[quote] Be sure that your client is configured like this, no matter what your real network connection is (That is, set it to 10M/max). I've had lots of users complain of lag and this fixed it for them; http://whisper.ausgamers.com/wiki/index.php/Bad_choke_solution [/quote] The only thing that that does is setting the rate to 10.000 HKEY_CURRENT_USER\Software\Valve\Steam value: rate rate 1 can be set in the console or in autoexec.cfg as well. Altho my experience is that on TF2 a rate of 1 is too low, I'd advice a minimum of 2 or 25000 From: Jesse Molina je...@opendreams.net To: Half-Life dedicated Linux server mailing list hlds_linux@list.valvesoftware.com Sent: Sunday, 8 April 2012, 4:09 Subject: Re: [hlds_linux] Poor TF2 performance on a dedicated server I don't know about your particular situation, but for the money you paid for this server class CPU and the motherboard, you could have gotten much better performance out of a desktop CPU and board. You probably should have gone with a CPU with fewer cores but a higher clock frequency. There are very few tools in the srcds process itself that will help you troubleshoot issues outside of memory exhaustion and configuration problems, so don't look there. You need to be using net_graph 5 on the client. What is your app ping like? When you get over 80ms, you will start to see a choke effect that is very similar to TCP window exhaustion. It seems to be built into the server, where if it does not get client feedback in time, it will choke off future updates. As far as I know, there is nothing that can be done about this. Do yourself a favor and do this; On my Windows 7 system, the path of my TF2 cfg directory is this; C:\Program Files (x86)\Steam\SteamApps\username\team fortress 2\tf\cfg In this directory, create or edit the file named autoexec.cfg Put the following into the file; Code: //netgraph script alias graph graph1 alias graph1 net_graphpos 1 ; net_graphproportionalfont 0 ; net_graph 4 ; alias graph graph2 alias graph2 net_graphpos 1 ; net_graphproportionalfont 1 ; net_graph 4 ; alias graph graph3 alias graph3 net_graphpos 1 ; net_graphproportionalfont 0 ; net_graph 1 ; alias graph graph4 alias graph4 net_graphpos 1 ; net_graphproportionalfont 1 ; net_graph 1 ; alias graph graph5 alias graph5 net_graphpos 1 ; net_graph 0 ; alias graph graph1 bind p graph This script makes it so that when you press p on your keyboard, it cycles through the net_graph in four different styles; graph plus large-text stats graph plus small-text stats large-text stats with no graph small-text stats with no graph On the netgraph, pay attention to the choke, sv, loss, and var values. When sv dips, your CPU is probably pegged out. Choke is the server holding back packets. Loss is obvious. Var is basically jitter. Be sure that your client is configured like this, no matter what your real network connection is (That is, set it to 10M/max). I've had lots of users complain of lag and this fixed it for them; http://whisper.ausgamers.com/wiki/index.php/Bad_choke_solution You will see a lot of bad advice out there about compiling your kernel, realtime, and other garbage. THere are a lot of very eager-to-please noob kids who want to run servers, but they don't know squat about being a sysadmin. You are running a very modern kernel on amd64; that's good. Could be some BIOS thing. Set it to defaults and don't fark with it unless you know what you are doing. Go back to the first thing I wrote on this email, and kick yourself for wasting money AND getting slower hardware than you could have had. Those Opterons are good for wide multi-threaded multi-user type applications, but that isn't what srcds is. Good luck frog wrote: We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. -- # Jesse Molina # Mail = je...@opendreams.net # Page = page-je...@opendreams.net # Cell = 1.602.323.7608 # Web = http://www.opendreams.net/jesse/ ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
Correct. But, the real problem is that users are being honest at some point and their honesty is causing them pain. You live in Denver, have 256K ADSL back in 2006, then you move to Chicago and have 50/20M cable in 2010, but you're settings are still from 2005 when you had 256K ADSL and the result is that you are punished with crappy game performance. Average users don't mess with autoexec.cfg or the rate commands. But yea, feel free to just use the commands like he said below. Good discussion. Mart-Jan Reeuwijk wrote: [quote] Be sure that your client is configured like this, no matter what your real network connection is (That is, set it to 10M/max). I've had lots of users complain of lag and this fixed it for them; http://whisper.ausgamers.com/wiki/index.php/Bad_choke_solution [/quote] The only thing that that does is setting the rate to 10.000 HKEY_CURRENT_USER\Software\Valve\Steam value: rate rate 1 can be set in the console or in autoexec.cfg as well. Altho my experience is that on TF2 a rate of 1 is too low, I'd advice a minimum of 2 or 25000 From: Jesse Molinaje...@opendreams.net To: Half-Life dedicated Linux server mailing listhlds_linux@list.valvesoftware.com Sent: Sunday, 8 April 2012, 4:09 Subject: Re: [hlds_linux] Poor TF2 performance on a dedicated server I don't know about your particular situation, but for the money you paid for this server class CPU and the motherboard, you could have gotten much better performance out of a desktop CPU and board. You probably should have gone with a CPU with fewer cores but a higher clock frequency. There are very few tools in the srcds process itself that will help you troubleshoot issues outside of memory exhaustion and configuration problems, so don't look there. You need to be using net_graph 5 on the client. What is your app ping like? When you get over 80ms, you will start to see a choke effect that is very similar to TCP window exhaustion. It seems to be built into the server, where if it does not get client feedback in time, it will choke off future updates. As far as I know, there is nothing that can be done about this. Do yourself a favor and do this; On my Windows 7 system, the path of my TF2 cfg directory is this; C:\Program Files (x86)\Steam\SteamApps\username\team fortress 2\tf\cfg In this directory, create or edit the file named autoexec.cfg Put the following into the file; Code: //netgraph script alias graph graph1 alias graph1 net_graphpos 1 ; net_graphproportionalfont 0 ; net_graph 4 ; alias graph graph2 alias graph2 net_graphpos 1 ; net_graphproportionalfont 1 ; net_graph 4 ; alias graph graph3 alias graph3 net_graphpos 1 ; net_graphproportionalfont 0 ; net_graph 1 ; alias graph graph4 alias graph4 net_graphpos 1 ; net_graphproportionalfont 1 ; net_graph 1 ; alias graph graph5 alias graph5 net_graphpos 1 ; net_graph 0 ; alias graph graph1 bind p graph This script makes it so that when you press p on your keyboard, it cycles through the net_graph in four different styles; graph plus large-text stats graph plus small-text stats large-text stats with no graph small-text stats with no graph On the netgraph, pay attention to the choke, sv, loss, and var values. When sv dips, your CPU is probably pegged out. Choke is the server holding back packets. Loss is obvious. Var is basically jitter. Be sure that your client is configured like this, no matter what your real network connection is (That is, set it to 10M/max). I've had lots of users complain of lag and this fixed it for them; http://whisper.ausgamers.com/wiki/index.php/Bad_choke_solution You will see a lot of bad advice out there about compiling your kernel, realtime, and other garbage. THere are a lot of very eager-to-please noob kids who want to run servers, but they don't know squat about being a sysadmin. You are running a very modern kernel on amd64; that's good. Could be some BIOS thing. Set it to defaults and don't fark with it unless you know what you are doing. Go back to the first thing I wrote on this email, and kick yourself for wasting money AND getting slower hardware than you could have had. Those Opterons are good for wide multi-threaded multi-user type applications, but that isn't what srcds is. Good luck frog wrote: We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. -- # Jesse Molina # Mail = je...@opendreams.net # Page = page-je...@opendreams.net # Cell = 1.602.323.7608 # Web = http://www.opendreams.net/jesse/ ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view
Re: [hlds_linux] Poor TF2 performance on a dedicated server
Thats why I solve that crap from the server: // setting min and max rate's (bytes per second TO the client) sv_maxrate 6 sv_minrate 3 // setting min and max updaterate (number of updates per second TO the client) sv_maxupdaterate 67 sv_minupdaterate 33 // setting min and max cmdrate (number of updates per second FROM the client) sv_maxcmdrate 67 sv_mincmdrate 33 // setting limits for the interpolation on the client, x = over x updates there can be interpolation sv_client_max_interp_ratio 2 sv_client_min_interp_ratio 1 // personally have the min_interp on 2 as well. // set cl_predict on connected clients to 1 (only while connected) sv_client_predict 1 Reason for mincmdrate and minupdaterate to be 33 is that its exactly half the tickrate, most people play on defaults (updaterate 20, cmdrate 30) so that lifts them to half the tickrate of the server, which makes their experience better, and easier for the server. And put a plugin in that kicks players for: - client rate value over 10 (over 100 k will make the packets from your server shorter and shorter up to a point where the packet data vs the content is bigger, yes even if you have sv_maxrate set) - rate values with a + in front (rate +5, cl_cmdrate +66, cl_updaterate +66 etc) - ping over 150 ms Yeah, I know, the settings are not friendly for dial up, but last time I was around any dial up is a long time ago, and I dont know anybody still using dial up. Pretty sure that that dinosaur is extinct in my country. And anyways, with dial up the connection is bad anyway for gaming. From: Jesse Molina je...@opendreams.net To: Mart-Jan Reeuwijk mreeu...@yahoo.com; Half-Life dedicated Linux server mailing list hlds_linux@list.valvesoftware.com Sent: Sunday, 8 April 2012, 14:46 Subject: Re: [hlds_linux] Poor TF2 performance on a dedicated server Correct. But, the real problem is that users are being honest at some point and their honesty is causing them pain. You live in Denver, have 256K ADSL back in 2006, then you move to Chicago and have 50/20M cable in 2010, but you're settings are still from 2005 when you had 256K ADSL and the result is that you are punished with crappy game performance. Average users don't mess with autoexec.cfg or the rate commands. But yea, feel free to just use the commands like he said below. Good discussion. Mart-Jan Reeuwijk wrote: [quote] Be sure that your client is configured like this, no matter what your real network connection is (That is, set it to 10M/max). I've had lots of users complain of lag and this fixed it for them; http://whisper.ausgamers.com/wiki/index.php/Bad_choke_solution [/quote] The only thing that that does is setting the rate to 10.000 HKEY_CURRENT_USER\Software\Valve\Steam value: rate rate 1 can be set in the console or in autoexec.cfg as well. Altho my experience is that on TF2 a rate of 1 is too low, I'd advice a minimum of 2 or 25000 From: Jesse Molinaje...@opendreams.net To: Half-Life dedicated Linux server mailing listhlds_linux@list.valvesoftware.com Sent: Sunday, 8 April 2012, 4:09 Subject: Re: [hlds_linux] Poor TF2 performance on a dedicated server I don't know about your particular situation, but for the money you paid for this server class CPU and the motherboard, you could have gotten much better performance out of a desktop CPU and board. You probably should have gone with a CPU with fewer cores but a higher clock frequency. There are very few tools in the srcds process itself that will help you troubleshoot issues outside of memory exhaustion and configuration problems, so don't look there. You need to be using net_graph 5 on the client. What is your app ping like? When you get over 80ms, you will start to see a choke effect that is very similar to TCP window exhaustion. It seems to be built into the server, where if it does not get client feedback in time, it will choke off future updates. As far as I know, there is nothing that can be done about this. Do yourself a favor and do this; On my Windows 7 system, the path of my TF2 cfg directory is this; C:\Program Files (x86)\Steam\SteamApps\username\team fortress 2\tf\cfg In this directory, create or edit the file named autoexec.cfg Put the following into the file; Code: //netgraph script alias graph graph1 alias graph1 net_graphpos 1 ; net_graphproportionalfont 0 ; net_graph 4 ; alias graph graph2 alias graph2 net_graphpos 1 ; net_graphproportionalfont 1 ; net_graph 4 ; alias graph graph3 alias graph3 net_graphpos 1 ; net_graphproportionalfont 0 ; net_graph 1 ; alias graph graph4 alias graph4 net_graphpos 1 ; net_graphproportionalfont 1 ; net_graph 1 ; alias graph graph5 alias graph5 net_graphpos 1 ; net_graph 0 ; alias graph graph1 bind p graph This script makes it so that when you press p on your keyboard, it cycles through
Re: [hlds_linux] Poor TF2 performance on a dedicated server
The rate is used to compute the time when the next update can be sent to a client. A a huge value means that the updates can be sent almost anytime. But the server won't send more than min(66, sv_maxupdaterate) updates to a client because it has only 66 frames per second, so a player can use 999 for rate without problems (is something like: nextUpdateTime = currentTime + 0). Anyway you can limit the maximum value using sv_maxrate, so there's no need to kick players for this. Also the values for convars are retrieved as an int or as a double. If the convar has an invalid value (+50 or X50 or XyZ or whatever) then the result is 0 (but a + sign should be valid if the parser is smart, same like - is valid for negative values). Kicking players for this is silly when you can enforce the limits using sv_minrate, sv_maxrate, sv_mincmdrate, sv_maxcmdrate, sv_minupdaterate and sv_maxupdaterate. -Original Message- From: hlds_linux-boun...@list.valvesoftware.com [mailto:hlds_linux-boun...@list.valvesoftware.com] On Behalf Of Mart-Jan Reeuwijk Sent: Sunday, April 08, 2012 4:08 PM To: Half-Life dedicated Linux server mailing list Subject: Re: [hlds_linux] Poor TF2 performance on a dedicated server ... And put a plugin in that kicks players for: - client rate value over 10 (over 100 k will make the packets from your server shorter and shorter up to a point where the packet data vs the content is bigger, yes even if you have sv_maxrate set) - rate values with a + in front (rate +5, cl_cmdrate +66, cl_updaterate +66 etc) ... ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
I love your server settings mart-jan! I hate when servers force me to use cl_updaterate 66 but at the same time force me to use bullshit like rate 3 _' 2012/4/8 hlds h...@gmx.com: The rate is used to compute the time when the next update can be sent to a client. A a huge value means that the updates can be sent almost anytime. But the server won't send more than min(66, sv_maxupdaterate) updates to a client because it has only 66 frames per second, so a player can use 999 for rate without problems (is something like: nextUpdateTime = currentTime + 0). Anyway you can limit the maximum value using sv_maxrate, so there's no need to kick players for this. Also the values for convars are retrieved as an int or as a double. If the convar has an invalid value (+50 or X50 or XyZ or whatever) then the result is 0 (but a + sign should be valid if the parser is smart, same like - is valid for negative values). Kicking players for this is silly when you can enforce the limits using sv_minrate, sv_maxrate, sv_mincmdrate, sv_maxcmdrate, sv_minupdaterate and sv_maxupdaterate. -Original Message- From: hlds_linux-boun...@list.valvesoftware.com [mailto:hlds_linux-boun...@list.valvesoftware.com] On Behalf Of Mart-Jan Reeuwijk Sent: Sunday, April 08, 2012 4:08 PM To: Half-Life dedicated Linux server mailing list Subject: Re: [hlds_linux] Poor TF2 performance on a dedicated server ... And put a plugin in that kicks players for: - client rate value over 10 (over 100 k will make the packets from your server shorter and shorter up to a point where the packet data vs the content is bigger, yes even if you have sv_maxrate set) - rate values with a + in front (rate +5, cl_cmdrate +66, cl_updaterate +66 etc) ... ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
When sv_maxrate is set to w/e reasonable, for sake of the convo lets say its 6, and the client has a large value set like: rate 100 network packets get very small ~64 byte etc, which makes that the server is pumping out mainly packets, but not much data. Unfortunately, that influences the other players as well. I agree that the rate setting will limit the data transfer, and with any normal values it will work. I have no idea why OB servers gets influenced so much by it. I just observe that it is so. On the +66 settings, it also does influence. Unfortunately valve has not heard of sanitizing input to variables on the client. The + values are used for a number of things by players: - avoiding ping kicks, for the ping calculation goes wrong - arbitrarely adjust rate settings. The + sign is somehow interpreted badly by the game engine and results in a negative effect. and but a + sign should be valid if the parser is smart ... I lolled. The server rate settings work if the rates on the client are set normal, without the high rate 1M setting or the + values. If such players are on your server, the rate value needs to be higher and the server performance is less. And I'm really not telling anything new. and anyways, for clients I advise: // full, good connection rate 5 cl_cmdrate 67 cl_updaterate 67 cl_interp 0.034 cl_interp_ratio 2.0 // half, mediocre rate 3 cl_cmdrate 33 cl_updaterate 33 cl_interp 0.064 cl_interp_ratio 2.0 // default, but better rate: rate 25000 cl_cmdrate 30 cl_updaterate 20 cl_interp 0.100 cl_interp_ratio 2.0 default rate are 3500, 5000, 7500 and 1, depending on the internet connection setting in steam client. From: hlds h...@gmx.com To: 'Mart-Jan Reeuwijk' mreeu...@yahoo.com; 'Half-Life dedicated Linux server mailing list' hlds_linux@list.valvesoftware.com Sent: Sunday, 8 April 2012, 15:37 Subject: RE: [hlds_linux] Poor TF2 performance on a dedicated server The rate is used to compute the time when the next update can be sent to a client. A a huge value means that the updates can be sent almost anytime. But the server won't send more than min(66, sv_maxupdaterate) updates to a client because it has only 66 frames per second, so a player can use 999 for rate without problems (is something like: nextUpdateTime = currentTime + 0). Anyway you can limit the maximum value using sv_maxrate, so there's no need to kick players for this. Also the values for convars are retrieved as an int or as a double. If the convar has an invalid value (+50 or X50 or XyZ or whatever) then the result is 0 (but a + sign should be valid if the parser is smart, same like - is valid for negative values). Kicking players for this is silly when you can enforce the limits using sv_minrate, sv_maxrate, sv_mincmdrate, sv_maxcmdrate, sv_minupdaterate and sv_maxupdaterate. -Original Message- From: hlds_linux-boun...@list.valvesoftware.com [mailto:hlds_linux-boun...@list.valvesoftware.com] On Behalf Of Mart-Jan Reeuwijk Sent: Sunday, April 08, 2012 4:08 PM To: Half-Life dedicated Linux server mailing list Subject: Re: [hlds_linux] Poor TF2 performance on a dedicated server ... And put a plugin in that kicks players for: - client rate value over 10 (over 100 k will make the packets from your server shorter and shorter up to a point where the packet data vs the content is bigger, yes even if you have sv_maxrate set) - rate values with a + in front (rate +5, cl_cmdrate +66, cl_updaterate +66 etc) ... ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
[hlds_linux] Poor TF2 performance on a dedicated server
I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. Any advise would be very gratefully received. Could it be that the hardware is simply not up to it regardless of OS and configuration? Has anyone here manage to get a 24 slot TF2 server to run smoothly on the same or comparable hardware? Happy to supply any further info as required. -- Best regards, frog mailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
What effects are you seeing in-game? On , frog f...@thehh.co.uk wrote: I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. Any advise would be very gratefully received. Could it be that the hardware is simply not up to it regardless of OS and configuration? Has anyone here manage to get a 24 slot TF2 server to run smoothly on the same or comparable hardware? Happy to supply any further info as required. -- Best regards, frog mailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
There has been a lot of lagg coming from valve servers when they are overloaded -or- down If it is continuous then you may have a problem with the server, making my point mute. I run all my servers on debian linux with no issues so it is not debian. do apt-get install htop and then run htop which is a far better view of what is actually happening on the server. (i assume you are running 64 bit debian, correct?) -- if so did you install the two packages 64bit servers need? On Apr 7, 2012, at 10:13 AM, dave.ro...@gmail.com wrote: What effects are you seeing in-game? On , frog f...@thehh.co.uk wrote: I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. Any advise would be very gratefully received. Could it be that the hardware is simply not up to it regardless of OS and configuration? Has anyone here manage to get a 24 slot TF2 server to run smoothly on the same or comparable hardware? Happy to supply any further info as required. -- Best regards, frog mailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
I'm pretty sure it's the server and nothing to do with Valve. I've been using atop ( http://www.atoptool.nl/ ) to monitor the what's being going on, but I've installed htop now too. I'm seeing a full 24 slot server using (wanting to use) 100% of 1 core for sustained periods of 30 seconds or more. On the server there are reports of things like lag spikes of 5-10 seconds, players teleporting around and frequent drops into Yellow figures for the SV Network Graph value which results in some nasty stuttering/hitching. (i assume you are running 64 bit debian, correct?) -- if so did you install the two packages 64bit servers need? Yes it's 64bit Debian Squeeze with 3.3.1 kernel. Which 2 packages might they be? Thanks :) Saturday, April 7, 2012, 6:28:14 PM, you wrote: There has been a lot of lagg coming from valve servers when they are overloaded -or- down If it is continuous then you may have a problem with the server, making my point mute. I run all my servers on debian linux with no issues so it is not debian. do apt-get install htop and then run htop which is a far better view of what is actually happening on the server. (i assume you are running 64 bit debian, correct?) -- if so did you install the two packages 64bit servers need? On Apr 7, 2012, at 10:13 AM, dave.ro...@gmail.com wrote: What effects are you seeing in-game? On , frog f...@thehh.co.uk wrote: I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. Any advise would be very gratefully received. Could it be that the hardware is simply not up to it regardless of OS and configuration? Has anyone here manage to get a 24 slot TF2 server to run smoothly on the same or comparable hardware? Happy to supply any further info as required. -- Best regards, frog mailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
On the server there are reports of things like lag spikes of 5-10 seconds, players teleporting around and frequent drops into yellow figures for the SV Network Graph value which results in some nasty stuttering/hitching. In atop srcds_linux is trying to use 100% (up to 120%) for sustained periods. Saturday, April 7, 2012, 6:13:35 PM, you wrote: What effects are you seeing in-game? On , frog f...@thehh.co.uk wrote: I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. Any advise would be very gratefully received. Could it be that the hardware is simply not up to it regardless of OS and configuration? Has anyone here manage to get a 24 slot TF2 server to run smoothly on the same or comparable hardware? Happy to supply any further info as required. -- Best regards, frog mailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
apt-get install lib32gcc1 ia32-libs I installed both on my servers, but it says or. Here is the post: http://wiki.teamfortress.com/wiki/Linux_dedicated_server * * On 4/7/2012 12:00 PM, frog wrote: On the server there are reports of things like lag spikes of 5-10 seconds, players teleporting around and frequent drops into yellow figures for the SV Network Graph value which results in some nasty stuttering/hitching. In atop srcds_linux is trying to use100% (up to 120%) for sustained periods. Saturday, April 7, 2012, 6:13:35 PM, you wrote: What effects are you seeing in-game? On , frogf...@thehh.co.uk wrote: I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. Any advise would be very gratefully received. Could it be that the hardware is simply not up to it regardless of OS and configuration? Has anyone here manage to get a 24 slot TF2 server to run smoothly on the same or comparable hardware? Happy to supply any further info as required. -- Best regards, frog mailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
Yeh, both of those were installed and up to date. Saturday, April 7, 2012, 9:38:34 PM, you wrote: apt-get install lib32gcc1 ia32-libs I installed both on my servers, but it says or. Here is the post: http://wiki.teamfortress.com/wiki/Linux_dedicated_server * * On 4/7/2012 12:00 PM, frog wrote: On the server there are reports of things like lag spikes of 5-10 seconds, players teleporting around and frequent drops into yellow figures for the SV Network Graph value which results in some nasty stuttering/hitching. In atop srcds_linux is trying to use100% (up to 120%) for sustained periods. Saturday, April 7, 2012, 6:13:35 PM, you wrote: What effects are you seeing in-game? On , frogf...@thehh.co.uk wrote: I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. Any advise would be very gratefully received. Could it be that the hardware is simply not up to it regardless of OS and configuration? Has anyone here manage to get a 24 slot TF2 server to run smoothly on the same or comparable hardware? Happy to supply any further info as required. -- Best regards, frog mailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux -- Best regards, frogmailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
I also assume that this server is not at your house, but how do you exactly have the server hosted? Shared, could be someone else causing the lagg? Are you running any mods? Do you have sprays enabled? On 4/7/2012 2:05 PM, frog wrote: Yeh, both of those were installed and up to date. Saturday, April 7, 2012, 9:38:34 PM, you wrote: apt-get install lib32gcc1 ia32-libs I installed both on my servers, but it says or. Here is the post: http://wiki.teamfortress.com/wiki/Linux_dedicated_server * * On 4/7/2012 12:00 PM, frog wrote: On the server there are reports of things like lag spikes of 5-10 seconds, players teleporting around and frequent drops into yellow figures for the SV Network Graph value which results in some nasty stuttering/hitching. In atop srcds_linux is trying to use100% (up to 120%) for sustained periods. Saturday, April 7, 2012, 6:13:35 PM, you wrote: What effects are you seeing in-game? On , frogf...@thehh.co.uk wrote: I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. Any advise would be very gratefully received. Could it be that the hardware is simply not up to it regardless of OS and configuration? Has anyone here manage to get a 24 slot TF2 server to run smoothly on the same or comparable hardware? Happy to supply any further info as required. -- Best regards, frog mailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
It's a dedicated server on a 1000mbit link hosted at Sovereign House data centre in London ( http://www.datacentermap.com/united-kingdom/london/telecity-london-sovereign-house_connectivity.html ) Average incoming is 1.37 MBit/s (max 2.13 MBit/s and outgoing 5.39 MBit/s (max 7.87 MBit/s) It has a basic sourcemod setup linking to a mysql db hosted on the same box. Sprays are enabled, but limited to registered users and the downloads folder is only ever 24 hours old. Saturday, April 7, 2012, 10:08:44 PM, you wrote: I also assume that this server is not at your house, but how do you exactly have the server hosted? Shared, could be someone else causing the lagg? Are you running any mods? Do you have sprays enabled? On 4/7/2012 2:05 PM, frog wrote: Yeh, both of those were installed and up to date. Saturday, April 7, 2012, 9:38:34 PM, you wrote: apt-get install lib32gcc1 ia32-libs I installed both on my servers, but it says or. Here is the post: http://wiki.teamfortress.com/wiki/Linux_dedicated_server * * On 4/7/2012 12:00 PM, frog wrote: On the server there are reports of things like lag spikes of 5-10 seconds, players teleporting around and frequent drops into yellow figures for the SV Network Graph value which results in some nasty stuttering/hitching. In atop srcds_linux is trying to use100% (up to 120%) for sustained periods. Saturday, April 7, 2012, 6:13:35 PM, you wrote: What effects are you seeing in-game? On , frogf...@thehh.co.uk wrote: I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. Any advise would be very gratefully received. Could it be that the hardware is simply not up to it regardless of OS and configuration? Has anyone here manage to get a 24 slot TF2 server to run smoothly on the same or comparable hardware? Happy to supply any further info as required. -- Best regards, frog mailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux -- Best regards, frogmailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
Okay so all those look good. Is it a dedicated 1000mbit connect to your server only, or is it shared? Is there any other info that you can give me, or when the lagg occurs, i.e. when players connect, leave, or just during normal game play? On 4/7/2012 2:27 PM, frog wrote: It's a dedicated server on a 1000mbit link hosted at Sovereign House data centre in London ( http://www.datacentermap.com/united-kingdom/london/telecity-london-sovereign-house_connectivity.html ) Average incoming is 1.37 MBit/s (max 2.13 MBit/s and outgoing 5.39 MBit/s (max 7.87 MBit/s) It has a basic sourcemod setup linking to a mysql db hosted on the same box. Sprays are enabled, but limited to registered users and the downloads folder is only ever 24 hours old. Saturday, April 7, 2012, 10:08:44 PM, you wrote: I also assume that this server is not at your house, but how do you exactly have the server hosted? Shared, could be someone else causing the lagg? Are you running any mods? Do you have sprays enabled? On 4/7/2012 2:05 PM, frog wrote: Yeh, both of those were installed and up to date. Saturday, April 7, 2012, 9:38:34 PM, you wrote: apt-get install lib32gcc1 ia32-libs I installed both on my servers, but it says or. Here is the post: http://wiki.teamfortress.com/wiki/Linux_dedicated_server * * On 4/7/2012 12:00 PM, frog wrote: On the server there are reports of things like lag spikes of 5-10 seconds, players teleporting around and frequent drops into yellow figures for the SV Network Graph value which results in some nasty stuttering/hitching. In atop srcds_linux is trying to use100% (up to 120%) for sustained periods. Saturday, April 7, 2012, 6:13:35 PM, you wrote: What effects are you seeing in-game? On , frogf...@thehh.co.ukwrote: I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. Any advise would be very gratefully received. Could it be that the hardware is simply not up to it regardless of OS and configuration? Has anyone here manage to get a 24 slot TF2 server to run smoothly on the same or comparable hardware? Happy to supply any further info as required. -- Best regards, frog mailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
You say it's a dedi box, have you made sure that any and all power-saving and dynamic overclocking features are turned OFF in BIOS? Both are known to cause hiccups like you describe. On 4/7/2012 5:33 PM, Cameron Munroe wrote: Okay so all those look good. Is it a dedicated 1000mbit connect to your server only, or is it shared? Is there any other info that you can give me, or when the lagg occurs, i.e. when players connect, leave, or just during normal game play? On 4/7/2012 2:27 PM, frog wrote: It's a dedicated server on a 1000mbit link hosted at Sovereign House data centre in London ( http://www.datacentermap.com/united-kingdom/london/telecity-london-sovereign-house_connectivity.html ) Average incoming is 1.37 MBit/s (max 2.13 MBit/s and outgoing 5.39 MBit/s (max 7.87 MBit/s) It has a basic sourcemod setup linking to a mysql db hosted on the same box. Sprays are enabled, but limited to registered users and the downloads folder is only ever 24 hours old. Saturday, April 7, 2012, 10:08:44 PM, you wrote: I also assume that this server is not at your house, but how do you exactly have the server hosted? Shared, could be someone else causing the lagg? Are you running any mods? Do you have sprays enabled? On 4/7/2012 2:05 PM, frog wrote: Yeh, both of those were installed and up to date. Saturday, April 7, 2012, 9:38:34 PM, you wrote: apt-get install lib32gcc1 ia32-libs I installed both on my servers, but it says or. Here is the post: http://wiki.teamfortress.com/wiki/Linux_dedicated_server * * On 4/7/2012 12:00 PM, frog wrote: On the server there are reports of things like lag spikes of 5-10 seconds, players teleporting around and frequent drops into yellow figures for the SV Network Graph value which results in some nasty stuttering/hitching. In atop srcds_linux is trying to use100% (up to 120%) for sustained periods. Saturday, April 7, 2012, 6:13:35 PM, you wrote: What effects are you seeing in-game? On , frogf...@thehh.co.ukwrote: I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. Any advise would be very gratefully received. Could it be that the hardware is simply not up to it regardless of OS and configuration? Has anyone here manage to get a 24 slot TF2 server to run smoothly on the same or comparable hardware? Happy to supply any further info as required. -- Best regards, frog mailto:f...@thehh.co.uk ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
Make sure to run htop with -d 1 for high resolution view. You won't see spikes at the normal resolution. Cameron Munroe wrote: do apt-get install htop and then run htop which is a far better view of what is actually happening on the server. -- # Jesse Molina # Mail = je...@opendreams.net # Page = page-je...@opendreams.net # Cell = 1.602.323.7608 # Web = http://www.opendreams.net/jesse/ ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
I don't know about your particular situation, but for the money you paid for this server class CPU and the motherboard, you could have gotten much better performance out of a desktop CPU and board. You probably should have gone with a CPU with fewer cores but a higher clock frequency. There are very few tools in the srcds process itself that will help you troubleshoot issues outside of memory exhaustion and configuration problems, so don't look there. You need to be using net_graph 5 on the client. What is your app ping like? When you get over 80ms, you will start to see a choke effect that is very similar to TCP window exhaustion. It seems to be built into the server, where if it does not get client feedback in time, it will choke off future updates. As far as I know, there is nothing that can be done about this. Do yourself a favor and do this; On my Windows 7 system, the path of my TF2 cfg directory is this; C:\Program Files (x86)\Steam\SteamApps\username\team fortress 2\tf\cfg In this directory, create or edit the file named autoexec.cfg Put the following into the file; Code: //netgraph script alias graph graph1 alias graph1 net_graphpos 1 ; net_graphproportionalfont 0 ; net_graph 4 ; alias graph graph2 alias graph2 net_graphpos 1 ; net_graphproportionalfont 1 ; net_graph 4 ; alias graph graph3 alias graph3 net_graphpos 1 ; net_graphproportionalfont 0 ; net_graph 1 ; alias graph graph4 alias graph4 net_graphpos 1 ; net_graphproportionalfont 1 ; net_graph 1 ; alias graph graph5 alias graph5 net_graphpos 1 ; net_graph 0 ; alias graph graph1 bind p graph This script makes it so that when you press p on your keyboard, it cycles through the net_graph in four different styles; graph plus large-text stats graph plus small-text stats large-text stats with no graph small-text stats with no graph On the netgraph, pay attention to the choke, sv, loss, and var values. When sv dips, your CPU is probably pegged out. Choke is the server holding back packets. Loss is obvious. Var is basically jitter. Be sure that your client is configured like this, no matter what your real network connection is (That is, set it to 10M/max). I've had lots of users complain of lag and this fixed it for them; http://whisper.ausgamers.com/wiki/index.php/Bad_choke_solution You will see a lot of bad advice out there about compiling your kernel, realtime, and other garbage. THere are a lot of very eager-to-please noob kids who want to run servers, but they don't know squat about being a sysadmin. You are running a very modern kernel on amd64; that's good. Could be some BIOS thing. Set it to defaults and don't fark with it unless you know what you are doing. Go back to the first thing I wrote on this email, and kick yourself for wasting money AND getting slower hardware than you could have had. Those Opterons are good for wide multi-threaded multi-user type applications, but that isn't what srcds is. Good luck frog wrote: We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. -- # Jesse Molina # Mail = je...@opendreams.net # Page = page-je...@opendreams.net # Cell = 1.602.323.7608 # Web = http://www.opendreams.net/jesse/ ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
Re: [hlds_linux] Poor TF2 performance on a dedicated server
On 07/04/2012 17:06, frog wrote: I'm hoping to get some advice from the undoubted wealth of experience subscribed to this list. We've got dedicated server, 6 x 2.3ghz (AMD Opteron 6276), 16GB RAM, 200GB HDD, which struggles to run a full 24 slot TF2 server smoothly. It used to be running CentOS 5.x, and a TF2 server would occasionally spike beyond 100% of a core. Months of experimentation resulted in some improvement, but nothing really satisfying. Also tried all the things suggested in http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide For the last few days, it's been rebuilt with Debian 6.0 with a built-on-the-box kernel of 3.3.1, following the recommendations listed in the fragaholics' wiki. http://wiki.fragaholics.de/index.php/EN:Kernel_Configuration_General Still there has been no improvement. If anything it's worse with sustained periods of srcds_linux requiring more than 100% of a core. The TF2 server is just running a basic sourcemod setup. I would start by just running TF2, no plugins, no mysql et al. If that works, then you know where the problem is. If it doesn't (and unless you can find some configuration or os-related problem) then you know it never will. -- Dan. ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux
