Re: [hlds_linux] [hlds] Linux srcds performance unacceptable
Our dod:s servers isnt having any problems, one has been up since dod:s was relesed, it has been problem, but not now ! Intel cpu:s P4 + 3040Xeon our scrim servers with sourcetv is having high cpu-usage, but that is a sourcetv problem. Peter - Original Message - From: "Rom1 [FFF-Dod]" <[EMAIL PROTECTED]> To: "Half-Life dedicated Linux server mailing list" Sent: Saturday, May 03, 2008 2:52 PM Subject: Re: [hlds_linux] [hlds] Linux srcds performance unacceptable Hello, I can confirm this. Updates are juste eating more and more cpu. Even on DoD:S ... For my servers a core2duo E6300 was really fine 1.5 years ago. Now it's not enough at all ! And how about Colmar ? This map is just unplayable because of tickrate drops ... On player's side : my old Athlon was fine with 25/30fps 2 years ago. On the same map I get 10fps max today ... Maybe the marketing push to more and more eyecandy things but what about an eyecandy game you can't play anymore because of mandatories updates ? Le samedi 3 mai 2008 à 12:59:35, vous écriviez : > This cpu usage is a rising problem. Friend of mine had 5 months ago a > server that was able to run 24 slots. around 3 months ago, it had to be > lowered to 22 slots after certain update. Now after Goldrush update, it > is down to 20 and even that is not enough because in some maps, the lag > is seen very clearly. > CPU just maxes out and if we have this problem, others will have it too. > Is there any way to reduce this usage because it really is a huge issue. > I took a test. Joined 10 randomly picked servers and only 1 of them was > able to run atleast 50 ticks at all times. half could run only 30 at > minimum and rest kept dropping under 30 and the worst part was round > start. Lag is problem, especially on over 20 slot servers right now and > someone needs to do something before those servers go down. Not everyone > has enough money to upgrade their hardware after 6 months. > -ics > Nephyrin Zey kirjoitti: >> I might've spoken a bit too soon on this - the servers do still get up >> into the high nineties CPU usage and stay there for a while - it seems >> like after they've been running for a while their overall CPU usage >> drifted back up to about the same as before the patch. This might be >> the first ever case of "CPU Leaks" :-/ The server also still incurs >> minor memory leaks, which seem to be about the same since before the >> patch. >> >> I run two 32x linux servers that are nearly always full and >> occasionally empty out at night. Here are some pretty graphs: >> >> http://www.nephyrin.net/NemuCPU.png >> >> Shows Load Average. On thursday it was lower despite both servers >> being full (???). The big spike today is the patch. The servers were >> both *instantly full* after the patch (among the first online) and >> have been *constantly full* since, yet the CPU usage over the past >> three hours has drifted up to pre-patch levels... >> http://nephyrin.net/NemuMem.png >> >> Memory Usage over a longer period - note that i went from one to two >> servers on tuesday, so the memory leaks have actually remained about >> the same as before. >> >> If anyone from Valve wants a login to my stats system let me know. It >> has tons and tons of pretty graphs dating to back when the server >> started. >> >> - Neph >> >> On Fri, May 2, 2008 at 5:00 PM, Nephyrin Zey <[EMAIL PROTECTED]> >> wrote: >> >>> Two patches ago levels == how it ran on the patch before the patch >>> before goldrush. Both the patch before goldrush and the goldrush patch >>> hurt performance, and its now at levels similar to prior to both of >>> those. >>> >>> Still a bit higher than it was when i first started hosting servers, >>> and higher than it should be, but better. >>> >>> - Neph >>> >>> >> >> ___ >> To unsubscribe, edit your list preferences, or view the list archives, >> please visit: >> http://list.valvesoftware.com/mailman/listinfo/hlds_linux >> > ___ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: > http://list.valvesoftware.com/mailman/listinfo/hlds_linux -- Cordialement, Rom1Courriel : [EMAIL PROTECTED] ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] [hlds] Linux srcds performance unacceptable
Hello, I can confirm this. Updates are juste eating more and more cpu. Even on DoD:S ... For my servers a core2duo E6300 was really fine 1.5 years ago. Now it's not enough at all ! And how about Colmar ? This map is just unplayable because of tickrate drops ... On player's side : my old Athlon was fine with 25/30fps 2 years ago. On the same map I get 10fps max today ... Maybe the marketing push to more and more eyecandy things but what about an eyecandy game you can't play anymore because of mandatories updates ? Le samedi 3 mai 2008 à 12:59:35, vous écriviez : > This cpu usage is a rising problem. Friend of mine had 5 months ago a > server that was able to run 24 slots. around 3 months ago, it had to be > lowered to 22 slots after certain update. Now after Goldrush update, it > is down to 20 and even that is not enough because in some maps, the lag > is seen very clearly. > CPU just maxes out and if we have this problem, others will have it too. > Is there any way to reduce this usage because it really is a huge issue. > I took a test. Joined 10 randomly picked servers and only 1 of them was > able to run atleast 50 ticks at all times. half could run only 30 at > minimum and rest kept dropping under 30 and the worst part was round > start. Lag is problem, especially on over 20 slot servers right now and > someone needs to do something before those servers go down. Not everyone > has enough money to upgrade their hardware after 6 months. > -ics > Nephyrin Zey kirjoitti: >> I might've spoken a bit too soon on this - the servers do still get up >> into the high nineties CPU usage and stay there for a while - it seems >> like after they've been running for a while their overall CPU usage >> drifted back up to about the same as before the patch. This might be >> the first ever case of "CPU Leaks" :-/ The server also still incurs >> minor memory leaks, which seem to be about the same since before the >> patch. >> >> I run two 32x linux servers that are nearly always full and >> occasionally empty out at night. Here are some pretty graphs: >> >> http://www.nephyrin.net/NemuCPU.png >> >> Shows Load Average. On thursday it was lower despite both servers >> being full (???). The big spike today is the patch. The servers were >> both *instantly full* after the patch (among the first online) and >> have been *constantly full* since, yet the CPU usage over the past >> three hours has drifted up to pre-patch levels... >> http://nephyrin.net/NemuMem.png >> >> Memory Usage over a longer period - note that i went from one to two >> servers on tuesday, so the memory leaks have actually remained about >> the same as before. >> >> If anyone from Valve wants a login to my stats system let me know. It >> has tons and tons of pretty graphs dating to back when the server >> started. >> >> - Neph >> >> On Fri, May 2, 2008 at 5:00 PM, Nephyrin Zey <[EMAIL PROTECTED]> wrote: >> >>> Two patches ago levels == how it ran on the patch before the patch >>> before goldrush. Both the patch before goldrush and the goldrush patch >>> hurt performance, and its now at levels similar to prior to both of >>> those. >>> >>> Still a bit higher than it was when i first started hosting servers, >>> and higher than it should be, but better. >>> >>> - Neph >>> >>> >> >> ___ >> To unsubscribe, edit your list preferences, or view the list archives, >> please visit: >> http://list.valvesoftware.com/mailman/listinfo/hlds_linux >> > ___ > To unsubscribe, edit your list preferences, or view the list archives, please > visit: > http://list.valvesoftware.com/mailman/listinfo/hlds_linux -- Cordialement, Rom1Courriel : [EMAIL PROTECTED] ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] [hlds] Linux srcds performance unacceptable
This cpu usage is a rising problem. Friend of mine had 5 months ago a server that was able to run 24 slots. around 3 months ago, it had to be lowered to 22 slots after certain update. Now after Goldrush update, it is down to 20 and even that is not enough because in some maps, the lag is seen very clearly. CPU just maxes out and if we have this problem, others will have it too. Is there any way to reduce this usage because it really is a huge issue. I took a test. Joined 10 randomly picked servers and only 1 of them was able to run atleast 50 ticks at all times. half could run only 30 at minimum and rest kept dropping under 30 and the worst part was round start. Lag is problem, especially on over 20 slot servers right now and someone needs to do something before those servers go down. Not everyone has enough money to upgrade their hardware after 6 months. -ics Nephyrin Zey kirjoitti: > I might've spoken a bit too soon on this - the servers do still get up > into the high nineties CPU usage and stay there for a while - it seems > like after they've been running for a while their overall CPU usage > drifted back up to about the same as before the patch. This might be > the first ever case of "CPU Leaks" :-/ The server also still incurs > minor memory leaks, which seem to be about the same since before the > patch. > > I run two 32x linux servers that are nearly always full and > occasionally empty out at night. Here are some pretty graphs: > > http://www.nephyrin.net/NemuCPU.png > > Shows Load Average. On thursday it was lower despite both servers > being full (???). The big spike today is the patch. The servers were > both *instantly full* after the patch (among the first online) and > have been *constantly full* since, yet the CPU usage over the past > three hours has drifted up to pre-patch levels... > http://nephyrin.net/NemuMem.png > > Memory Usage over a longer period - note that i went from one to two > servers on tuesday, so the memory leaks have actually remained about > the same as before. > > If anyone from Valve wants a login to my stats system let me know. It > has tons and tons of pretty graphs dating to back when the server > started. > > - Neph > > On Fri, May 2, 2008 at 5:00 PM, Nephyrin Zey <[EMAIL PROTECTED]> wrote: > >> Two patches ago levels == how it ran on the patch before the patch >> before goldrush. Both the patch before goldrush and the goldrush patch >> hurt performance, and its now at levels similar to prior to both of >> those. >> >> Still a bit higher than it was when i first started hosting servers, >> and higher than it should be, but better. >> >> - Neph >> >> > > ___ > To unsubscribe, edit your list preferences, or view the list archives, please > visit: > http://list.valvesoftware.com/mailman/listinfo/hlds_linux > ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] [hlds] Linux srcds performance unacceptable
I might've spoken a bit too soon on this - the servers do still get up into the high nineties CPU usage and stay there for a while - it seems like after they've been running for a while their overall CPU usage drifted back up to about the same as before the patch. This might be the first ever case of "CPU Leaks" :-/ The server also still incurs minor memory leaks, which seem to be about the same since before the patch. I run two 32x linux servers that are nearly always full and occasionally empty out at night. Here are some pretty graphs: http://www.nephyrin.net/NemuCPU.png Shows Load Average. On thursday it was lower despite both servers being full (???). The big spike today is the patch. The servers were both *instantly full* after the patch (among the first online) and have been *constantly full* since, yet the CPU usage over the past three hours has drifted up to pre-patch levels... http://nephyrin.net/NemuMem.png Memory Usage over a longer period - note that i went from one to two servers on tuesday, so the memory leaks have actually remained about the same as before. If anyone from Valve wants a login to my stats system let me know. It has tons and tons of pretty graphs dating to back when the server started. - Neph On Fri, May 2, 2008 at 5:00 PM, Nephyrin Zey <[EMAIL PROTECTED]> wrote: > Two patches ago levels == how it ran on the patch before the patch > before goldrush. Both the patch before goldrush and the goldrush patch > hurt performance, and its now at levels similar to prior to both of > those. > > Still a bit higher than it was when i first started hosting servers, > and higher than it should be, but better. > > - Neph > ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux