1344212879 J * bergerx_ ~bergerx@176.240.194.77 1344212998 Q * bergerx Read error: Operation timed out 1344216305 Q * nenolod Ping timeout: 480 seconds 1344216316 J * nenolod ~nenolod@67.202.104.180 1344224757 J * FireEgl ~FireEgl@173-25-83-57.client.mchsi.com 1344227972 J * vn585 ~R3h@89.205.104.146 1344229191 Q * clopez Ping timeout: 480 seconds 1344237057 J * nlm ~nlm@host111.190-225-209.telecom.net.ar 1344237409 Q * nlm_ Ping timeout: 480 seconds 1344239528 J * kir ~kir@swsoft-msk-nat.sw.ru 1344239708 P * kir 1344240616 J * fisted_ ~fisted@xdsl-84-44-145-64.netcologne.de 1344240796 Q * fisted Ping timeout: 480 seconds 1344241380 N * Bertl_zZ Bertl 1344241385 M * Bertl morning folks! 1344242036 J * ensc ~irc-ensc@p54ADFBEC.dip.t-dialin.net 1344242449 Q * ensc_ Ping timeout: 480 seconds 1344245590 J * clopez ~clopez@fanzine.igalia.com 1344247519 Q * FireEgl Read error: Connection reset by peer 1344247642 J * BenG ~bengreen@cpc10-aztw24-2-0-cust114.aztw.cable.virginmedia.com 1344250890 M * _br_ Good morning Bertl 1344250982 Q * BenG Quit: I Leave 1344251630 J * FireEgl ~FireEgl@173-25-83-57.client.mchsi.com 1344252154 J * ccxCZ ~ccxCZ@178.238.36.182 1344253559 M * vn585 Is there any page where I can see all the possible vserver configurations with explanation? 1344253880 M * Bertl yes, the great flower page 1344253895 M * Bertl just google for 'flower page' the first hit should be the configuration overview 1344256637 M * vn585 thanks 1344259037 M * vn585 when I try to hard limit CPU on the VPS I get /usr/lib64/util-vserver/vserver.functions: line 1610: /dev/cgroup/testvps/cpu.cfs_period_us: Permission denied 1344259045 M * vn585 what is wrong? 1344259315 M * Guy- Bertl: I'm running 3.4.2-vs2.3.3.4, and inside a guest, I see this: lr-x------ 1 root root 64 Aug 6 15:20 /proc/1/fd/3 -> /var/lib/vservers/ilms-be 1344259330 M * Guy- where /var/lib/vservers/ilms-be is the path to the root of the guest, on the host 1344259333 M * Guy- is this expected? 1344259461 M * daniel_hozac i would say no :) 1344259485 M * Guy- a number of other processes have the same fd (likely inherited from init) 1344259528 M * daniel_hozac vn585: does that file exist? 1344259577 M * Bertl Guy-: what kernel/util-vserver version? 1344259616 M * Guy- Bertl: kernel 3.4.2-vs2.3.3.4, util-vserver 0.30.216-pre3034, initstyle plain 1344259681 M * Guy- (and instead of system V init, I use runit-init inside the guest) 1344259701 M * vn585 daniel_hozac: no, it not exist 1344259785 M * Bertl Guy-: so there should be no relation to the outside 1344259811 M * Bertl and the fd points to the root of the guest? 1344259870 M * daniel_hozac vn585: then you don't have the cpu system enabled, or the name of the file has changed in your kernel. 1344259905 M * Guy- Bertl: yes 1344259917 M * Guy- Bertl: (that path doesn't even exist inside the guest) 1344259966 M * Bertl well, I consider it somewhat unusual that init keeps a filedescriptor to a directory 1344259978 M * Guy- it inherits it and doesn't close it, probably 1344259993 M * Bertl where should it come from? 1344260008 M * Bertl I really doubt that util-vserver opens a filedescriptor there 1344260012 M * Guy- some part of util-vserver would be my suspect 1344260060 M * vn585 daniel_hozac: what option in the kernel I need to enable? I use 3.0.37 kernel at this moment. 1344260102 M * Guy- Bertl: I looked at the source of runit; it doesn't open anything other than /dev/console, at least afaict 1344260110 M * Guy- and it doesn't go out of its way to close inherited FDs 1344260142 M * Bertl daniel_hozac: any ideas on fd=3 being used by util-vserver? 1344260219 M * daniel_hozac sounds possible, i'll take a look. 1344260295 M * Bertl the fact that it is shown with the host path might be a bordercase in the 'within' check 1344260752 M * vn585 guys can some one please tell me how to find if the "cpu.cfs_period_us" file name is changed in the 2.0.27 kernel? 1344261026 M * vn585 I mean 3.0.37 kernel :) 1344261106 M * Bertl check the kernel documentation (i.e. Documentation in the kernel source tree) for cgroup related details 1344261530 Q * clopez Ping timeout: 480 seconds 1344262066 M * ensc|w daniel_hozac: vsysctl is guilty by not closing curdir 1344262140 M * Bertl hey ensc|w! LTNS! how's going? 1344262487 M * vn585 Is hard limit available at all in this kernel version? Because I'm not able to see anything like "CONFIG_CFS_HARD_LIMITS" in the .config file. 1344262541 M * ensc|w Bertl: well; still the old job 1344262580 M * Bertl vn585: mainline didn't develop hard limits for some time and I'm not sure they are present/working in recent kernel, but it seems, nobody complained hard enough (to the cgroups folks :) 1344262677 M * vn585 is there any other way to hard limit CPU usage in a VPS on 3.x kernel? 1344262723 M * Bertl nope, we dropped the token bucket scheduler in favor of mainline development 1344262742 M * Bertl but you can still do cpusets and thus achieve some kind of hard limits 1344262754 M * Guy- vn585: is there a good reason to limit cpu usage other than for billing? 1344262755 M * Bertl (given that you have enough cores to group :) 1344262778 M * Bertl Guy-: yes, if you want to ensure that certain resources are always available 1344262808 M * Guy- for the host? 1344262833 M * Guy- or for other guests? 1344262872 M * Guy- still, a hard limit, without borrowing, doesn't seem reasonable to me - if the CPU were otherwise idle, why not allow it to be used? 1344262895 M * Guy- I can understand wanting a limit if it's not idle, certainly 1344262943 M * Bertl it's a question of response time for certain types of applications 1344262961 M * Bertl (i.e. near real time response and/or high availability) 1344263026 M * Guy- OK, I can see that, but that would still work with borrowing, wouldn't it? 1344263055 M * Bertl if the scheduler would be (a lot) better, yes, of course 1344263167 M * Guy- so you're saying mingo's celebrated O(1) scheduler reacts too slowly to changes in load for borrowed CPU time to be cut back quickly if needed? 1344263474 M * Bertl all the current mainline shedulers are pretty bad when it comes to short response times 1344263475 M * vn585 Why I need it? I do not need it? Just playing around with vservers qemu and openvz for now. 1344263478 M * vn585 :) 1344263516 M * Bertl Guy-: try to play a movie on a server running 20 cpu intensive tasks, even with highest priority it will lag 1344263539 M * Guy- and with SCHED_RR or SCHED_FIFO? 1344263603 M * Bertl realtime scheduling is better, still if a cpu is in use and the process(es) need to be evicted/moved, you can get significant latencies on a newly starting process 1344263627 M * Bertl (that's why almost all real time processes reserve all resources in advance and then idle around) 1344264070 A * trippeh routinely watches 1080p HD video on machine with make -j$highnum in background;) 1344264102 M * trippeh Works fine. ;) 1344264127 M * Bertl lucky you, probably hardware support for decoding 1344264404 M * Guy- Bertl: thanks for the explanation 1344264472 M * Bertl np, but of course, hard limit is somehting mainly used for selling resources ... 1344264611 M * trippeh Bertl: Heh nope. 1344264630 M * trippeh Just CONFIG_SCHED_AUTOGROUP=y 1344264774 M * trippeh Automaticly groups processes according to setsid() 1344264820 M * trippeh A group of processes gets scheduled as one process, kinda. 1344264905 M * Bertl defeats the purpose if you have 20 cpu intensive tasks hogging one CPU, but yes, I can imagine that works 1344264976 M * trippeh setsid is generally called when forking a deamon, running a make, launching an app in gnome etc. 1344265002 M * trippeh So any processes they fork, is considered one (with fairness within, of course) 1344265045 M * trippeh A bunch of flash processes in chrome wont kill your interactivity anymore :) 1344265103 M * trippeh or an amok apache wont starve mysql 1344265179 M * trippeh (in the case of prefork mpm) 1344269314 J * bonbons ~bonbons@2001:960:7ab:0:11a7:59d0:1d0:79c9 1344270378 J * trippeh_ atomt@t-1000.ugh.no 1344270381 Q * trippeh_ 1344272647 M * Bertl off for a nap ... bbl 1344272657 N * Bertl Bertl_zZ 1344273619 Q * ensc|w Remote host closed the connection 1344273630 J * ensc|w ~ensc@www.sigma-chemnitz.de 1344278423 J * clopez ~clopez@131.29.165.83.dynamic.mundo-r.com 1344280360 J * petzsch ~markus@p57B65395.dip.t-dialin.net 1344280655 Q * petzsch 1344281206 M * disposable i'm trying to move my vservers from 2.6.26 kernel to 2.6.32. I've read about the barrier attribute and thought i'd be able to handle it. but no matter what i do, I cannot get a vserver started. what am i doing wrong? http://pastebin.com/KAx5HXwc 1344282168 M * daniel_hozac what is in /vservers/source/etc/init.d? 1344282245 M * disposable one more thing that might help; when i try to run 'chroot /vservers/' i get - chroot: cannot run command `/bin/bash': Permission denied 1344282258 M * daniel_hozac why it fails after you set the barrier is because you set it on the wrong directory. 1344282423 M * disposable care to elaborate? 1344282454 M * disposable i have tried this too - t /vservers/' i get - chroot: cannot run command `/bin/bash': Permission denied 1344282464 M * disposable sorry, wrong paste 1344282492 M * disposable i have tried this too - cd -P /etc/vservers/source/vdir; setattr --barrier .. 1344282715 M * daniel_hozac that would likely be accurate. 1344282727 M * daniel_hozac you still need to remove it from the guest directory itself. 1344282744 M * daniel_hozac and if chroot doesn't work, vserver definitely will not work. so you'll want to fix that first. 1344282792 M * disposable daniel_hozac: when you say guest directory, you mean /vservers, right? 1344283074 M * disposable daniel_hozac: i've removed barrier from /vservers, from /vservers/ and still can't start chroot into . is there a way to remove barrier attribute from teh whole directory tree or mount the fs without barriers? 1344283410 M * disposable daniel_hozac: nevermind.... i had the fs mounted as noexec. DOH!!! 1344283723 Q * ntrs Ping timeout: 480 seconds 1344283799 Q * vn585 Ping timeout: 480 seconds 1344284100 J * ntrs ~ntrs@vault08.rosehosting.com 1344284120 J * hijacker_ ~hijacker@cable-84-43-134-121.mnet.bg 1344284501 N * Bertl_zZ Bertl 1344284504 M * Bertl back now ... 1344285361 J * BenG ~bengreen@cpc10-aztw24-2-0-cust114.aztw.cable.virginmedia.com 1344286077 J * fisted ~fisted@xdsl-87-78-11-151.netcologne.de 1344286260 Q * fisted_ Ping timeout: 480 seconds 1344286294 Q * hijacker_ Quit: Leaving 1344287041 Q * bonbons Quit: Leaving 1344287866 Q * BenG Quit: I Leave 1344289432 Q * mrjack 1344290803 J * nlm_ ~nlm@host106.190-30-33.telecom.net.ar 1344291158 Q * nlm Ping timeout: 480 seconds 1344291184 Q * Hunger Ping timeout: 480 seconds 1344292404 Q * DoberMann Ping timeout: 480 seconds 1344296125 J * DoberMann ~james@2a01:e35:8b44:84c0::2