1187829937 Q * bzed Quit: Leaving 1187830872 J * Radiance 9eff72de20@halt.1984world.eu 1187832651 Q * Piet_ Quit: Piet_ 1187833208 J * DoberMann_ ~james@AToulouse-156-1-79-49.w86-196.abo.wanadoo.fr 1187833315 Q * DoberMann[ZZZzzz] Ping timeout: 480 seconds 1187846106 J * friendly12345 ~friendly@ppp121-44-237-56.lns2.mel4.internode.on.net 1187848796 J * sharkjaw ~gab@158.36.44.106 1187849447 J * duckx ~Duck@tox.dyndns.org 1187849669 Q * quasisane Ping timeout: 480 seconds 1187851430 J * ftx_ ~ftx@dslb-084-062-227-163.pools.arcor-ip.net 1187851617 Q * duckx Remote host closed the connection 1187851691 J * meandtheshell ~markus@85.127.105.177 1187852569 Q * Baby Remote host closed the connection 1187852894 J * dna ~dna@107-212-dsl.kielnet.net 1187853170 Q * AStorm Remote host closed the connection 1187853218 J * AStorm ~astralsto@tor-irc.dnsbl.oftc.net 1187853637 J * arachnis1 arachnist@088156185052.who.vectranet.pl 1187853690 J * esa ~esa@ip-87-238-2-45.adsl.cheapnet.it 1187853708 Q * eSa| Ping timeout: 480 seconds 1187853813 Q * arachnist Ping timeout: 480 seconds 1187854076 J * Baby ~miry@195.37.62.208 1187854223 M * matti Morning :-) 1187854331 M * arachnis1 hi 1187854339 N * arachnis1 arachnist 1187854375 M * matti :) 1187855677 N * DoberMann_ DoberMann[PullA] 1187856157 Q * emtty Remote host closed the connection 1187856996 Q * balbir Quit: Leaving 1187857004 J * balbir ~balbir@122.167.95.137 1187857078 Q * balbir 1187857097 J * balbir ~balbir@122.167.95.137 1187857169 Q * balbir 1187857187 J * balbir ~balbir@122.167.95.137 1187857999 J * emtty ~eric@dynamic-acs-24-154-34-241.zoominternet.net 1187858907 J * yarihm ~yarihm@whitehead2.nine.ch 1187859085 J * Pazzo ~ugelt@reserved-225136.rol.raiffeisen.net 1187859885 J * pmenier ~pmenier@LNeuilly-152-22-72-5.w193-251.abo.wanadoo.fr 1187861025 J * dreamind ~dreamind@p54A78073.dip0.t-ipconnect.de 1187861068 M * dreamind Hi 1187861236 M * Supaplex ahoy 1187861255 M * dreamind hi Supaplex 1187862336 J * quasisane ~user@c-76-118-191-64.hsd1.nh.comcast.net 1187864149 Q * dreamind Quit: dreamind 1187864835 J * markus__ ~chatzilla@mail.netcare.at 1187864855 M * markus__ Hi! Can I mount rbind a directory into a vserver without stopping vserver, editing fstab, starting vserver, i.e. on the fly? 1187865187 M * FaUl iirc thats possible 1187865195 M * FaUl but not quite sure now 1187865212 M * FaUl how 1187865259 M * Borg- vnamespace -e bind 1187865268 M * Borg- s/bind/mount/ 1187865454 J * meebey meebey@booster.qnetp.net 1187865463 M * meebey setattr --~hide /proc/bus/usb 1187865485 M * meebey /proc/bus/usb: Invalid argument 1187865490 M * meebey isnt that supposed to work? 1187865523 M * meebey hm worked on /proc/bus but not /proc/bus/usb 1187865596 M * meebey inside the vserver: 1187865600 M * meebey ls /proc/bus -l 1187865600 M * meebey total 0 1187865603 M * meebey hmpf :( 1187866599 M * daniel_hozac meebey: /proc/bus/usb is a separate filesystem. 1187866615 M * daniel_hozac see http://linux-vserver.org/USB_Devices 1187867898 J * AlanCox ~linus@bl7-141-245.dsl.telepac.pt 1187867900 M * AlanCox ) 1187867906 A * AlanCox hi :) 1187868496 A * harry doesn't buy it... 1187868503 M * harry you're not ac, nor lt 1187868505 A * harry food 1187868564 J * Piet ~piet@tor.noreply.org 1187868573 J * phedny ~mark@ip56538143.direct-adsl.nl 1187868854 M * matti :) 1187869264 J * lilalinux_ ~plasma@dslb-084-059-017-211.pools.arcor-ip.net 1187869264 Q * pmenier Read error: Connection reset by peer 1187869575 Q * lilalinux Ping timeout: 480 seconds 1187869871 N * sid3windr RikBobbaers 1187869872 M * RikBobbaers :> 1187869875 N * RikBobbaers sid3windr 1187870924 J * Freax ~linus@bl7-140-10.dsl.telepac.pt 1187871218 Q * AlanCox Ping timeout: 480 seconds 1187871251 Q * AStorm Quit: WeeChat 0.2.5 1187871307 J * AStorm ~astralsto@tor-irc.dnsbl.oftc.net 1187871466 Q * ktwilight Read error: Connection reset by peer 1187871503 J * ktwilight ~ktwilight@67.207-66-87.adsl-static.isp.belgacom.be 1187871541 Q * ftx_ Read error: Connection reset by peer 1187871566 J * ftx_ ~ftx@dslb-084-060-247-218.pools.arcor-ip.net 1187871641 M * meebey daniel_hozac: thanks for the hint/link 1187871849 Q * markus__ Quit: ChatZilla 0.9.78.1 [Firefox 2.0.0.6/2007072518] 1187872959 Q * meandtheshell Quit: Leaving. 1187873237 J * lilalinux__ ~plasma@dslb-084-058-236-164.pools.arcor-ip.net 1187873343 J * meandtheshel1 ~markus@85.127.117.156 1187873471 J * lilalinux ~plasma@dslb-084-059-008-032.pools.arcor-ip.net 1187873614 Q * lilalinux_ Ping timeout: 480 seconds 1187873793 Q * lilalinux__ Ping timeout: 480 seconds 1187873845 P * meandtheshel1 Leaving. 1187874012 J * meandtheshell ~markus@85.127.117.156 1187876087 J * lilalinux_ ~plasma@dslb-084-058-236-239.pools.arcor-ip.net 1187876335 J * fatgoose_ ~samuel@206-248-132-45.dsl.teksavvy.com 1187876468 Q * lilalinux Ping timeout: 480 seconds 1187876588 J * dreamind ~dreamind@p54A78073.dip0.t-ipconnect.de 1187876591 Q * dreamind 1187876655 Q * sharkjaw Quit: Leaving 1187876668 Q * fatgoose Ping timeout: 480 seconds 1187876870 Q * Freax Quit: I'll Be back 1187876897 Q * Guy- Remote host closed the connection 1187876897 J * AlanCox ~linus@bl7-140-10.dsl.telepac.pt 1187876963 J * Guy- ~korn@elan.rulez.org 1187877066 J * Curalton ~thomas@newerth.com 1187877209 Q * gerrit Ping timeout: 480 seconds 1187877796 J * gerrit ~gerrit@c-67-169-199-103.hsd1.or.comcast.net 1187879083 A * ard loves vservers 1187879088 M * ard tyz 32419 0.0 0.0 2936 700 ? R 16:21 0:00 bash -c : ( ) { : | : & } ; : & 1187879101 M * ard do not have a chance anymore to kill my system :-) 1187879375 M * ard top - 16:29:32 up 44 min, 6 users, load average: 293.17, 252.45, 159.67 1187879418 M * daniel_hozac neat. 1187879428 M * mnemoc daniel_hozac: hi, do you use debian's diet 0.30-7? 1187879440 M * daniel_hozac i don't use Debian, so... no :) 1187879459 M * daniel_hozac why do you ask? 1187879472 M * mnemoc http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=435538 <-- you are here :p 1187879496 M * daniel_hozac yes, i'm CCed on all util-vserver bugs. 1187879513 M * mnemoc because i found a bug on write_pad() and i wanted to know if you had a patch :p 1187879553 M * mnemoc but, as you don't use it... you probably don't experience the bug :p 1187879556 M * daniel_hozac http://svn.linux-vserver.org/projects/dietlibc 1187879574 M * daniel_hozac it's Debian-only? 1187879587 M * mnemoc yes, introduced between 0.30-4 and 0.30-7 1187879616 M * daniel_hozac okay. 1187879724 M * waldi ard: it is also not able to kill my xen machine ... 1187879747 P * friendly12345 1187879883 M * ard waldi : of course it is not. But there is a difference in amount of resources used in normal cases :-) 1187879943 J * Julius ~julius@p57B278F8.dip.t-dialin.net 1187880721 J * lilalinux__ ~plasma@dslb-084-059-001-181.pools.arcor-ip.net 1187880924 J * lilalinux ~plasma@dslb-084-058-252-230.pools.arcor-ip.net 1187881038 Q * lilalinux_ Ping timeout: 480 seconds 1187881204 Q * lilalinux__ Ping timeout: 480 seconds 1187881227 P * Curalton 1187881526 J * gebura ~gebura@96.201.101-84.rev.gaoland.net 1187881532 M * gebura hello all 1187881551 M * gebura i have a little problem with vserver , debian and postfix 1187881589 M * gebura on fews vservers, i want to use postfix in "local-only" mode 1187881630 M * gebura but config provided by dpkg-reconfigure (debconf) seems depend on loopback interface 1187881660 M * gebura has somebody got same problem ? 1187881674 M * grobie hi gebura 1187881693 M * grobie are you using etch? 1187881712 M * gebura hi 1187881714 M * gebura yes i am 1187881746 M * gebura i see that in my log 1187881749 M * gebura "Aug 23 14:38:30 web postfix/master[3833]: fatal: /etc/postfix/master.cf: line 18: no valid IP address found: smtp" 1187881881 M * gebura i am thinking about 3 way to fix the issue 1187881893 M * gebura 1) configuring manualy postfix 1187881906 M * gebura 2) adding a loopback interface if possible 1187881919 M * gebura 3) using another MTA like nullmailer (prety simple) 1187881928 M * gebura the best way for me is the 1 1187881937 M * gebura but if you have an other idea... 1187882315 N * Bertl_oO Bertl 1187882323 M * Bertl morning folks! 1187882339 M * gebura hello Bertl 1187882375 M * grobie gebura: adding a loopback isn't always possible, but you can run dpkg-reconfigure and use a different ip instead of 127.0.0.1 1187882390 M * Bertl reading the back log now ... 1187882426 M * grobie gebura: well, adding a loopback interface might not be possible for you in fact ;) 1187882434 M * Bertl gebura: you can check out the 2.3.x devel branch, specifically 2.3.0.17+ 1187882464 M * Bertl gebura: that will give your guest the illusion of a loopback device 1187882494 M * gebura tks but, it is on a production server , i don't want to use devel branch 1187882516 M * Bertl k, then I'd go for the manual reconfiguration (which is the sane way to handle it) 1187882864 M * grobie gebura: afaik you only need to change the inet_interfaces in main.cf 1187883033 M * Bertl yes, and probably localhost in /etc/hosts too, but that's not postfix related 1187883059 M * gebura on /etc/hosts , it's done 1187883112 M * gebura what value can i use for inet_interfaces ? 1187883217 M * Bertl the actual interfaces visible inside the guest 1187883332 M * gebura like inet_interfaces dummy0 ? 1187883336 M * gebura i will try, thanks 1187883418 M * Bertl if you see dummy0 inside your guest, yes 1187883478 M * gebura hum, that don't work 1187883502 M * gebura but hostname defined in /etc/hosts work 1187883515 M * gebura many thanks :) 1187883565 J * Freax ~linus@bl7-143-88.dsl.telepac.pt 1187883568 M * Bertl excellent, maybe add that info to the wiki? 1187883630 Q * fatgoose_ Quit: fatgoose_ 1187883779 Q * Aiken Remote host closed the connection 1187883848 Q * AlanCox Ping timeout: 480 seconds 1187883866 M * gebura that's a good idea 1187883923 M * gebura but i don't know where can i write 1187883939 M * gebura http://linux-vserver.org/Documentation > configuration ? or guides and howto ? 1187884025 M * gebura what do you think is better ? 1187884055 J * Satriani ~linus@bl7-141-254.dsl.telepac.pt 1187884082 J * Piet_ ~piet@tor.noreply.org 1187884353 Q * Satriani Read error: Connection reset by peer 1187884413 Q * Freax Ping timeout: 480 seconds 1187884469 Q * Piet Ping timeout: 480 seconds 1187884656 M * Bertl gebura: IIRC, we have a page called problematic programs 1187884681 M * gebura oups 1187884683 M * Bertl gebura: it might be good to put it there and state that this is only a problem for a specific configuration 1187884696 M * gebura i have started to write another page 1187884702 M * gebura i finish and try to transfer 1187884708 M * Bertl okay, tx 1187885146 J * bonbons ~bonbons@2001:960:7ab:0:20b:5dff:fec7:6b33 1187885323 M * gebura http://linux-vserver.org/Postfix_local_only_problem 1187885328 M * gebura what do you think about ? 1187885333 M * gebura my english is poor 1187885342 M * gebura maybe somebody will have to check for mistake 1187885401 M * Bertl will be fixed up at some point ... maybe leave it a separate page and just link it from the problematic programs page 1187885606 M * gebura this is done 1187885609 M * gebura i have to go 1187885611 M * gebura good bye 1187885615 M * gebura and thanks for your help 1187885618 Q * gebura Quit: Quitte 1187886480 M * FaUl can i somehow disable that loopback-virtualisation/ 1187886653 J * bzed ~bzed@dslb-084-059-110-141.pools.arcor-ip.net 1187886787 M * Bertl FaUl: yes, per context actually 1187887543 M * FaUl ok, how do i do that? 1187887655 M * Bertl there is a flag for that, called *LBACK 1187887672 M * Bertl check the source for details, you probably need to use the bit value for now 1187887796 M * Supaplex moin Bertl 1187888027 M * FaUl very well 1187888246 M * FaUl sigh, still no newer version of util-vserver in debian unstable 1187888255 M * FaUl i guess i have to build my own 1187888277 M * Bertl you need the prerelease versions anyway 1187888292 M * Bertl otherwise 2.3.0.x will not work as expected 1187888305 M * FaUl yea, i guess so 1187888347 M * FaUl mh, that utils are missing on the wiki-mainpage 1187888395 M * FaUl ay 1187888404 M * FaUl someone changed the default-layout of the great flower-page 1187888407 M * Bertl http://people.linux-vserver.org/~dhozac/t/uv-testing/ 1187888426 M * FaUl not the one on the nongnu-page? 1187888456 M * ard Hmmmm.... Does hard_sched makes the token-bucket scheduler active or is that already active? 1187888502 M * FaUl anyway, i'll try that later 1187888513 M * FaUl the most important thing is that i got the kernel working 1187888520 M * FaUl testing can be done later now 1187888532 M * FaUl btw have the sun and the sgi reached you now? 1187888550 M * Bertl I returned from Boston yesterday, so no yet :) 1187888783 A * ard is wondering how to get the FairUse scheduling active 1187888797 M * Supaplex FaUl: you know bow to use uupdate right? 1187888797 M * ard hard_sched definetely works... dnetc consumes 2% cpu :-) 1187889096 M * ard Hmmmm, /me just wanted to say to gebura that with a single ip the postfix_local_only fix the postfix is not local at all :-( (just like mysql default install) 1187889581 M * FaUl Bertl: ah, ok 1187889616 J * lilalinux_ ~plasma@dslb-084-059-001-204.pools.arcor-ip.net 1187889764 J * lilalinux__ ~plasma@dslb-084-058-235-043.pools.arcor-ip.net 1187889988 Q * lilalinux Ping timeout: 480 seconds 1187890138 Q * lilalinux_ Ping timeout: 480 seconds 1187890209 M * Bertl +*-/ 1187890342 J * the-dude ~martijn@senturparks.xs4all.nl 1187890847 M * ard Hmzzz... fair cpu usage needs CONFIG_VSERVER_IDLETIME? 1187890944 A * ard hits head against wall ... 1187890976 M * Bertl ard: yep 1187891009 M * ard I thought I enabled that, but obviously I did not :-( 1187891205 M * ard Hmmm... Now to put that in the wiki... 1187891285 M * daniel_hozac it's already on the wiki. 1187891330 M * ard It would be nice to have it at http://linux-vserver.org/CPU_Scheduler 1187891336 M * the-dude can someone tell me why you prefer vserver instead of xen? 1187891351 M * daniel_hozac the-dude: resource consumption. 1187891370 M * ard and the sharing of basic things like filesystems 1187891372 M * the-dude like memory? 1187891547 M * the-dude ard, do you mean that you can start serveral machine's with one base or so ? 1187891548 M * daniel_hozac the-dude: memory, disk space, you name it. 1187891548 M * the-dude yeah but xen does that too right 1187891548 M * ard the-dude : things like that: a single bash for all your vservers 1187891561 M * daniel_hozac the-dude: sharing? no. 1187891585 M * ard the concept of xen and vserver are that different that you can't share a single bash over all your vservers 1187891594 M * the-dude daniel_hozac, why would you consider sharing memory ? don't you wnat each machine too have it's own memory ? 1187891619 M * ard no... 1187891626 M * the-dude ard, yeah ok I understand that, and it seems handy 1187891631 M * daniel_hozac the-dude: no, i want to be able to run as many guests as possible on one machine. 1187891632 M * ard Just like internet... you share your bandwidth with others 1187891655 M * the-dude oh ok I see that is a big bennefit 1187891665 M * ard one user can peak, and another can peak. As long as nobody does that at the same time there are no problems 1187891689 M * the-dude ok but what if one machine peaks and another machine needs too peak too? 1187891702 M * ard if they do however, unlike bandwidth sharing you *can* put minimal numbers on your vservers 1187891728 M * ard if both peak, you can give one precedence, or let them share 1187891766 M * the-dude ok are you running production machine's or would you consider running any? 1187891767 M * ard f.i. for cpu, you can give a garantuee of a minimun amount of cpu cycles, and everything above will be fair use... 1187891786 M * ard I am running production... 1187891794 M * the-dude on what scale? 1187891807 M * ard currently small 1187891826 M * daniel_hozac lycos-vds is running Linux-VServer. 1187891831 M * the-dude how many? 1187891838 M * the-dude yeah I heard about it 1187891840 M * ard a lot of servers around here have vserver capable kernels, but we have to have some procedure written out to handle vservers 1187891866 M * ard for another company I only use vservers 1187891872 M * daniel_hozac then there's also planet-lab.org. 1187891909 M * the-dude vserver is uml right ? 1187891913 M * ard no 1187891923 J * Chad92 ~Chad92@72.13.172.147 1187891927 M * ard it is like a chroot on steroids 1187891929 M * ard a super jail :-) 1187891937 M * the-dude wow big mistake *g* 1187891959 M * ard that's why you can share resources very easy 1187891985 M * ard I guess you would be able to run it on a WRT54G f.i. :-) 1187892008 M * the-dude ard, No mine did not like the rain last year ;) 1187892038 M * ard you can use it as a virtualization of hardware, or as a further hardening of a single task system 1187892053 A * ard now has to build a new kernel :-(... 1187892062 M * the-dude but does lycos run on vserver? 1187892068 M * daniel_hozac well, it's more like process isolation than hardware virtualization. 1187892124 M * Bertl the-dude: yes, they do for certain products 1187892177 M * the-dude thx, how is it compared too xen with friendlyness? 1187892193 M * daniel_hozac friendlyness? 1187892211 A * ard thinks vserver is very friendly :-) 1187892222 M * the-dude xen has those great tools too manage vm's 1187892227 M * ard I just install a new kernel, and I am able to go vserver 1187892252 M * ard Around here we usually make the tools ourselves... 1187892265 M * the-dude *g* what a suprise 1187892299 M * ard (I mean the company I am working for... I have contributed nothing to vserver whatsoever :-) ) 1187892351 M * the-dude what happens when one of the machine's crashes? 1187892363 M * ard That can't happen 1187892367 M * ard the whole server crashes 1187892369 M * ard or none :-) 1187892373 M * the-dude *g* 1187892383 M * ard you can have a runaway process in one of the contexts 1187892405 M * ard but if you limit your contexts good enough, you just kill the processes in that context 1187892436 M * ard :() { :|: & } ; : # made me take a hard look at limits 1187892458 M * ard don't try that at your bash, it will kill your system, unless you have good limits :-) 1187892460 M * the-dude ok I can understand the pro's but what about the cons ? 1187892478 M * ard what cons? :-) 1187892487 M * the-dude everything has cons :) 1187892495 M * ard You can also run windows-XP using qemu... 1187892503 M * ard that's a big con ;-) 1187892524 M * the-dude but qemu doesn't need a vserver kernel 1187892546 M * the-dude and why should I consider running xp brrrr 1187892579 P * Chad92 In Soviet Russia, iMac parts you! 1187892583 M * ard But that's about the biggest con: you cannot upgrade a kernel for a single "server" 1187892595 M * ard because there is only one 1187892617 M * the-dude ok so they all use the same kernel, but I don't think thats a real problem 1187892636 M * ard Things like kvm and xen have life migration. I would never use it, but that's a difference 1187892665 M * ard and all other differences between jails and system virtualization 1187892683 J * bzed_ ~bzed@dslb-084-059-107-157.pools.arcor-ip.net 1187892720 M * ard The biggest CON so far is that it's not in the mainstream kernel 1187892736 M * the-dude I started with xen cause my collegue said it was fun, so I started too work with it, but I think it is wise too look too other options aswell 1187892863 P * meebey 1187892877 M * ard I never tried to use it. When I saw the horrible renaming of interfaces and such :-). A collegeau of mine runs it, but he will use vserver for the next install because of the resource usage 1187892913 M * the-dude ard, do you mean the xenbridge interface? 1187892945 M * ard probably... I mean that they rename eth0 to something else, and then put fake eth0's in place 1187892975 M * ard I'd rather they don't do that, and just give them sane names... 1187892983 M * the-dude No the interfaces stay normal but it will add a bridge/routing device 1187893037 M * ard The real interfaces won't work anymore. They create new interfaces that are connected to the bridge, and call them eth0 1187893043 M * ard (unless that has changed) 1187893064 M * ard I'd rather call my interface xen0:, because then I know what it is :-) 1187893092 M * the-dude thats possible if you want to 1187893108 Q * bzed Ping timeout: 480 seconds 1187893113 M * the-dude but I don't think it is really important 1187893116 M * ard yes I know, but it's somehow a default to rename them 1187893129 M * the-dude what servers are you running inside vserver? 1187893145 M * ard and as a network guy that want's to know what's happening that is really confusing 1187893184 M * ard some webservers, mailservers, code-development (c-compilers) envirenments and such 1187893231 M * ard subversion and trac for a company that develops software for BMW cars 1187893260 M * the-dude did you ever stress tested it ? 1187893321 M * ard Yes, I've got a rather large test environment for www.speurders.nl (about 60 servers in a live environment) running on a single server :-) 1187893369 M * the-dude dutchy 1187893370 M * ard I once hit a bug on one of my servers 1187893379 M * ard but daniel_hozac already fixed that 1187893388 A * ard is dutch 1187893404 A * the-dude is also very dutch 1187893416 M * ard then why are we not talking dutch? 1187893430 M * daniel_hozac because this is an English channel? :) 1187893433 M * the-dude cause it will be unpolite for the otherss? 1187893453 M * the-dude but speuders is running on a vserver? 1187893454 M * ard ah, your not on #alioth... 1187893463 M * ard the acceptence environment is 1187893476 M * ard and a lot of speurders servers are running vserver enabled kernels 1187893493 M * ard that's for a future security hardening step 1187893502 M * ard vserver is also good for that :-) 1187893530 M * ard It limits the networking of your jail in very nice ways 1187893540 M * the-dude I see but are they just running vserver enabled kernels or are they using the capablity's of vserver? 1187893574 M * ard On live servers it is not really interesting to use vservers for resource sharing 1187893606 M * ard so they are not using the capabilities. But having a vserver enabled kernel is interesting enough to test :-) 1187893614 M * ard especially on speurders... 1187893641 M * the-dude yeah but what are you testing? just the stability of the kernel? 1187893648 M * ard It means you can prepare your cluster for vserver, and when not satisfied just keep them as they are 1187893676 M * ard in the acceptance environment the code of www.speurders.nl and all database interactions and dataflows 1187893697 M * ard On the live server it is important that the machines just survive 1187893708 M * ard We really push the limits of servers... *really* 1187893797 M * ard (especcially in the firewalling department :-) ) 1187893846 M * the-dude stresstesting iptables or so ? 1187893898 M * ard Anyway: we are able to serve all pictures of www.speurders.nl with 4 apache servers and 2 proxy servers. Marktplaats needs 300+ servers for that :-) 1187893916 M * ard iptables is also stressed 1187893956 M * ard And the nicer routing tables for LVS: 1187893958 M * ard ard@loesje:~$ sudo ip ro sh table tolvs 1187893958 M * ard local default dev lo scope host 1187893983 M * the-dude how much traffic are we talking about ? 1187894008 M * ard Hmmm... that's the kind of questions I am not allowed to answer :-( 1187894022 M * the-dude ok no worries 1187894035 M * ard But we do have 2 10G connections at amsix :-) 1187894050 M * the-dude But I think speuders takes less traffic then markplaats 1187894063 M * ard that's probably true... 1187894115 M * ard When I was working for speurders (setting up the infra) I think it was about 1/3. But a lot of pages have been optimized to make the size/page even smaller 1187894135 M * ard an ex collegeau is working at marktplaats now :-) 1187894164 M * the-dude I know some guys who did some for marktplaats as well 1187894191 M * the-dude speuders is also apache and mysql ? 1187894208 M * ard Ask them for Cees de Groot, he probably knows the difference between speurders and marktplaats 1187894214 M * ard yes 1187894224 M * ard and no netapp's like marktplaats... 1187894226 M * ard :-) 1187894247 M * ard apache,mysql,squid 1187894254 M * ard and lvs for loadbalancing 1187894284 M * ard recently the management forced fredhopper (search engine) down my throat 1187894293 M * ard that's the last what I did for them ;-) 1187894312 M * the-dude squid as a reverse proxy or so ? 1187894323 M * ard yes 1187894349 M * the-dude sounds like some nice hardware :) 1187894390 M * ard We loadbalance images on 2 proxies (*2 for redundancy), each set get's the odd or even pictures 1187894408 M * ard The proxies cache that and %4 request that from the imageservers 1187894438 M * ard You need about 8G per proxy server to cache the pictures of a day's worth of advertisements 1187894450 M * the-dude lvs is software loadbalancing? 1187894455 M * ard yes 1187894471 M * ard It outperforms our hardware loadbalancers 1187894482 M * the-dude any experiance with with pound ? 1187894485 M * ard and it is rocksolid (unlike the hardware balancers) 1187894503 M * ard No... I am using that at home... 1187894522 M * the-dude which hardware, so I will know what not to buy 1187894534 Q * yarihm Quit: This computer has gone to sleep 1187894541 M * ard the proxies are DELL-1850's 1187894546 M * ard with 8G 1187894589 M * ard for image servers you just need a setup that can do at least 50 uncached requests per second for the worst cases 1187894605 M * ard 50 uncached requests per second is *really* a lot 1187894605 M * the-dude hmmz interesting 1187894649 M * the-dude how many db-servers? 1187894653 M * ard You need to take the seek time into account of your disks 1187894659 M * ard Heh 1187894664 M * ard > 30 :-( 1187894675 M * ard mainly due to *bad* programming 1187894687 M * ard and no intent to fix taht 1187894721 M * the-dude apache+php? 1187894722 M * ard And for fredhopper we even have *really* heavy hardware 1187894746 M * ard not only that, but also the method of replication. 1187894769 M * ard The unwillingness to not use autoincrements in mysql 1187894772 M * ard stuff like that 1187894798 Q * igraltist Ping timeout: 480 seconds 1187894820 M * the-dude not using increments had impact on that large scale ? 1187894909 M * ard If you do not use increments you will *never* have a replication problem ;-) 1187894909 M * ard If you do not have a potential replication problem you can rearrange replication and structures on the fly without downtime 1187894948 M * the-dude master slave config ? 1187894963 M * ard master-master ring replications and such 1187894975 Q * lilalinux__ Remote host closed the connection 1187894982 M * ard and on that master you can have a bunch of RO slaves 1187894985 M * the-dude ok I have some experiance with master slave configurations 1187895021 M * ard For HA you need at least a master-master ring replication to be able to do maintenance 1187895083 M * the-dude I should read about it, or do you feel like explaining it in a nutshell? 1187895130 M * ard Eh... this moment I need to really start compiling my kernel... I am still at work... 1187895179 M * the-dude *g* no worries 1187895183 M * the-dude thx for the info so far 1187895289 Q * Pazzo Quit: ... 1187895399 J * Rich_Estill ~restill@c-24-11-195-139.hsd1.mi.comcast.net 1187896303 Q * coderanger Quit: coderanger 1187896393 N * bzed_ bzed 1187896759 J * xe ex@81.219.196.129 1187896759 Q * xe 1187897725 J * jmcaricand ~jmcarican@d83-179-201-62.cust.tele2.fr 1187898101 Q * meandtheshell Quit: Leaving. 1187898255 Q * jmcaricand Quit: KVIrc 3.2.4 Anomalies http://www.kvirc.net/ 1187898295 N * puck tabcomplete 1187898467 Q * Piet_ Remote host closed the connection 1187898897 J * yarihm ~yarihm@84-75-109-39.dclient.hispeed.ch 1187899366 N * tabcomplete puck 1187899898 M * the-dude ard, did you use Ultra Monkey? 1187900281 Q * bonbons Ping timeout: 480 seconds 1187900460 J * bonbons ~bonbons@2001:960:7ab:0:20b:5dff:fec7:6b33 1187900864 Q * ftx_ Ping timeout: 480 seconds 1187901681 Q * bonbons Ping timeout: 480 seconds 1187901693 J * bonbons ~bonbons@2001:960:7ab:0:20b:5dff:fec7:6b33 1187902276 M * ard the-dude : keepalived 1187902276 Q * Julius Remote host closed the connection 1187902357 M * the-dude ard, does mysql-ndb still use memory for storage? 1187902370 M * ard with : http://217.196.41.9/~ard/vlarp/ and http://217.196.41.9/~ard/vlan/ 1187902377 M * ard we don't use that 1187902382 M * ard and yes... 1187902404 M * the-dude is that why you have 30 mysql servers :) 1187902432 M * ard no 1187902444 M * ard that's because of *BAD* programming/design 1187902482 M * ard some stuff was accelerated by putting that in daemons written in C 1187902490 M * the-dude I like the mysqlcluster idea but I doubt if all our databases will fit into memory 1187902516 M * ard but bad programming would have crashed mysql-nbd 1187902529 M * the-dude but it seems like a really nice solution 1187902535 M * the-dude why cause of load ? 1187902548 M * ard Hmmm, nbd sounds to me like a single lock would lock the cluster 1187902567 M * ard a single node fail will kill your sie 1187902568 M * ard eh 1187902568 M * ard site 1187902586 M * ard the old replication style is very stupid, and therefore very predictable 1187902607 M * ard if you use that in the right way, you don't have to use 30+ servers :-) 1187902628 M * ard on the other hand: marktplaats has 300+ last time I heard :-) 1187902638 M * the-dude so you are using a differant technique for mysql clustering? 1187902667 M * ard plain old ring replication will suffice 1187902679 M * ard 2 servers in a ring, and per server a few slaves 1187902719 M * the-dude ok i did not figure that out yet 1187902719 M * ard But at some point your swamped with updates on the slave, so you have to filter out the updates to the slaves 1187902740 M * ard If you do that part correct, you don't need 30+ 1187902745 M * the-dude but doesn't that make the slave's inconsistent? 1187902764 M * ard that depends on what you call consistent 1187902777 M * ard they lag 1..30s with the master 1187902796 M * ard so that should not be a problem 1187902798 M * the-dude ok so what do you mean by filter out the updates 1187902822 M * ard you don't need al the tables on every slave 1187902844 M * ard a group of slaves should be dedicated to a specific task 1187902862 M * ard the slaves can have different indexes from the master 1187902887 M * the-dude so if both masters in de rings crash then what happens? 1187902902 M * ard then you did something wrong :-) 1187902907 M * the-dude *g* 1187902908 M * ard only 1 will crash :-) 1187902937 M * the-dude and the master are connected to each other? 1187902938 M * ard we have both masters at different locations, connected with a 20G link 1187902977 M * ard if you have a split brain, smart programming will make sure you won't have a headache when the split stops 1187902986 M * ard but smart programmers are hard to find :-( 1187902999 M * the-dude *g* I was going to say something like that 1187903021 M * the-dude but are the masters identical to eachother? 1187903046 M * ard but this solution can survive a split brain, and can even work independent from eachother for an hour or so 1187903073 M * ard If you do it correctly: both masters are indeed R/W masters and replicate eachother 1187903097 M * ard in other words they are the same (unless some looser f*cks the sql statements) 1187903107 M * the-dude isn't that dangerous? 1187903119 M * ard you have to have a clue bat at all times with you when communicating with programmers 1187903121 M * the-dude if one breaks the otherone breaks to right 1187903137 M * ard that's correct 1187903145 M * ard they don't break, but the replication will stop 1187903157 M * the-dude and then how do you respond ? 1187903172 M * ard I used to fix it 1187903211 M * the-dude fix the replication , and use binfiles to recover it ? 1187903216 M * ard But in the speurders case the lack of clue was too big to be able to have two masters... so there is a single master and a standby master 1187903245 M * ard skip the statement and repair what was wrong 1187903284 M * the-dude ok ok cause if there is some loadbalancing in front of the mysql and the application keeps running you will get 2 different database 1187903313 M * ard only if you don't program for clusters 1187903356 M * ard you will use the slaves only as a RO database, you will only allow correct non-auto-incrementing queries on the master 1187903364 M * ard stuff like that 1187903372 M * ard it's not hard, but you have to think 1187903412 M * ard a programmer should not be allowed to think up queries without consulting peers or let it be audited by peers 1187903460 M * ard auditing queries has prevented a lot of downtime (and prevented a lot of mysql injections :-) ) 1187903472 M * the-dude I still am somehow suprised by the stupid bugs programmers can come up with 1187903498 M * the-dude easy way out :( 1187903504 M * ard heh 1187903562 M * ard I once had a programmer that at 4:00 AM confessed that he didn't realise that in a cluster that picture won't be available on every server without syncing it 1187903599 M * ard That programmer in the end decided for himself that it was best to look for another job 8-D 1187903622 M * ard Although he did survive another year of harsh remarks 1187903636 M * the-dude yeah well some of our programmers still try to deny that we are using mysql master slave combination 1187903706 M * the-dude what do you think of a master slave setting? 1187903709 M * ard Use of non-replicated variables and select * from into... and autoincrements and stuff like that? 1187903731 M * the-dude *g* yeah thats sounds very familiar 1187903745 M * ard I would always prefer master-master in which one master can be used as a RO slave 1187903770 M * ard It's so easy to switch the roles so you can optimize the tables without downtime 1187903804 M * the-dude yeah thats true , but I am wondering why we are using master slave configurations 1187903890 M * ard High availability... if the master dies, scrap it, and use the slave :-) 1187903919 M * ard you loose a few transactions, but that's usually not a problem 1187904047 M * the-dude so the server is upgraded to master? 1187904064 M * the-dude sorry 1187904085 M * the-dude the hot standyby which is slave of the master becomes upgraded to master 1187904297 M * ard yes... that's the idea 1187904329 M * the-dude yeah I quess thats what we are using at the moment 1187904392 Q * bonbons Quit: Leaving 1187904406 M * the-dude how do you release new software? 1187904484 M * the-dude we schedule downtime, bring town the application, make a dump , update the database, release the new software, most of the times it is fine, but it takes alot of time to do a rollback 1187904521 M * ard importing a dump takes too much time 1187904534 M * the-dude yep it does 1187904544 M * ard as for the images: restoring a full imageserver takes more than a week 1187904555 M * the-dude it is really timeconsuming job 1187904577 M * ard if it is that important, you do an acceptence test on 1/2 of your cluster, other half shut off 1187904597 M * ard if it is ok, you fix the other half 1187904619 M * the-dude what if it is not :) 1187904648 M * ard you replicate the right 1/2 to the wrong 1/2 1187904664 M * ard raw-copying is always way faster than importing 1187904827 M * the-dude what about stopping the replication, the upgrade the slave if everthing goes allright promote the slave to master, would that be possible and a good idea? 1187904958 M * the-dude any flaws? 1187904958 M * ard depends on your settings 1187904958 M * ard upgrading the master or the slave is really dependent on your situation :-) 1187904958 A * ard is reb00ting for a new kernel on his macb00k 1187904958 M * ard ttl 1187904962 M * the-dude master is being used, the slave does nothing accept being a mysql slave 1187904970 M * the-dude ok ok good luck :) 1187905894 Q * dna Quit: Verlassend 1187906875 J * Ashsong ~chatzilla@orchard.laptop.org 1187906880 M * Ashsong Bertl: ping 1187906964 M * Bertl Ashsong: pong! 1187906981 M * Ashsong Ah, hello. 1187906986 M * Ashsong You had a safe trip home, I trust? 1187907001 M * Bertl yep, tx, I did ... and even all my baggage arrived with me :) 1187907005 M * Ashsong Excellent. 1187907013 M * Ashsong So: my question. 1187907078 M * Ashsong I'm seeing processes freeze after calling vc_ctx_create(). 1187907092 M * Ashsong dmesg reports that the call succeeded; however, the process freezes. 1187907103 M * Ashsong Thoughts on how to start debugging this? 1187907106 M * Ashsong strace, perhaps? 1187907130 M * Ashsong (Also, vkill -c $xid -s 9 fails to kill the process) 1187907137 M * Bertl hmm, freeze as in process not responding or freeze as in getting a zombie? 1187907163 M * Bertl after doing a create, the process is marked as setup process 1187907170 M * Ashsong freeze as in process stops responding to keyboard input and is not killed by vkill. 1187907175 M * Bertl so it is half in and half outside the context 1187907179 M * Ashsong right. 1187907194 M * Bertl you should finalize the setup in the same thread 1187907200 M * Ashsong If I could, I would. :) 1187907211 M * Bertl so it hangs before you get there? 1187907222 M * Ashsong Well, I'm doing this interactively in python. 1187907231 M * Bertl ah, bad choice :) 1187907237 M * Ashsong Would you like me to test any vcmd or vcontext incantations for you? 1187907246 M * Ashsong Bertl: well, it used to work fine. 1187907246 M * Bertl you will lose terminal conectivity as soon as you enter the context 1187907250 J * FireEgl FireEgl@Sebastian.Tcldrop.Com 1187907263 M * Bertl at least if you have a pts 1187907301 M * Bertl the 'frozen' process how does it show up on 'vps auxww' ? 1187907302 M * Ashsong Even after calling create but before turning off the STATE_SETUP flag? 1187907336 M * Bertl yes, for the process, the context tagging is already in place while the context is still in setup 1187907345 J * Aiken ~james@ppp121-45-255-55.lns2.bne4.internode.on.net 1187907355 M * Bertl wb Aiken! 1187907360 M * Aiken hello 1187907364 M * Ashsong "root 1916 1600 0.0 1.6 7572 3876 pts/1 H+ 22.12 0:00 python 1187907366 M * Ashsong " 1187907370 M * Bertl aha! H 1187907381 M * Ashsong What does H indicate? 1187907383 M * Bertl you exhausted your hard scheduler limits then 1187907389 M * Ashsong Hmm. 1187907391 M * Bertl very interesting ... 1187907397 M * Ashsong Indeed. :) 1187907400 M * Ashsong No limits should be set. 1187907403 M * Ashsong (yet) 1187907413 M * Bertl could I get the output of the /proc/virtual//sched 1187907415 M * Ashsong The only vserver call I've made is vc_ctx_create() 1187907423 M * Bertl please use paste.linux-vserver.org 1187907426 M * Ashsong Sure. 1187907454 M * Bertl could as well be a scheduler bug 1187907481 M * micah daniel_hozac: did you see #439133? I reported the compile bug and ted tso (author of ext fs) responded 1187907493 M * micah (he also maintains the debian packages) 1187907494 M * Bertl micah: url? 1187907507 M * Ashsong http://paste.linux-vserver.org/6474 1187907561 M * micah Bertl: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=439133 1187907565 M * Bertl tx 1187907566 M * Ashsong It appears that it should be getting tokens... 1187907640 M * Bertl Ashsong: and the /proc/virtual//status please 1187907723 M * micah (FTBS = fails to build from source) 1187907726 J * Piet_ ~piet@tor.noreply.org 1187907753 M * Ashsong `http://paste.linux-vserver.org/6475 1187907833 M * Bertl b78449c0bf989d00 <-- looks like your problem :) 1187907859 M * Bertl you are using the one argument create call, which passes initial flags (see vcmd for that) 1187907886 M * Bertl obviously you are passing in an unitialized pointer/field, so the flags are bogus 1187907908 M * Ashsong Ah, thanks. 1187907909 M * Bertl if you accidentially specify the hold flag, the context will hang 1187907914 M * Ashsong :) 1187907917 M * Ashsong Handy feature, that. 1187907931 M * Bertl you can change that to reactivate and kill the context, btw 1187907957 M * Ashsong Indeed. Or I can reboot and use the correct call the next time. 1187907966 M * Bertl note: when passing the initial flags, you want at least the ones for setup and admin to be there 1187907972 M * Ashsong Right 1187907978 M * Bertl (see VMCD Howto examples for details) 1187908020 M * Ashsong thanks. 1187908027 M * Bertl np 1187908074 Q * yarihm Quit: Leaving 1187908562 M * Ashsong Bertl: the version of util-vserver-lib that we wrote our wrapper for had a vc_ctx_create() that only took one argument. 1187908575 M * Ashsong Do you know if there are other functions that now take different arguments? 1187908598 M * Bertl yep, was the previous version of this command, we updated a few versions to the latest syscall interface 1187908613 M * Bertl you could sync with the latest util-vserver too, should now use them too 1187908649 M * Bertl IIRC, it was create and migrate, the scheduler calls and the new network interface calls which got added 1187908663 M * Ashsong Thanks. 1187908770 M * Ashsong So I need STATE_SETUP, STATE_INIT, and STATE_ADMIN? 1187908772 M * Ashsong Anything else? 1187908790 M * Bertl not that should be fine to start with 1187908811 M * Bertl you drop STATE_SETUP at the end of your setup procedure 1187908864 M * Ashsong Indeed, we do. 1187909040 M * daniel_hozac Ashsong: how are you calling vc_ctx_create? are you calling the util-vserver wrapper, or the syscall directly? 1187909045 M * daniel_hozac what kernel is this based on? 1187909066 M * Bertl daniel_hozac: the OLPC kernel and the wrapper from the old tools 1187909069 M * Ashsong daniel_hozac: through util-verver-0.30-214-pre2590 1187909069 M * daniel_hozac micah: yeah, i saw it. 1187909077 M * Ashsong on a custom kernel. 1187909089 M * Bertl hmm, those are the new tools ... strange ... 1187909097 M * daniel_hozac Bertl: so roughly equivalent to 2.3.0.17? 1187909099 M * Ashsong Essentially 2. 2.6.22-vsOLPC.0.4.4 + a few things. 1187909116 M * Ashsong Bertl: it was our fault. 1187909125 M * Bertl daniel_hozac: yeah, more like the upcoming 2.3.0.18 1187909129 M * daniel_hozac right. 1187909211 M * daniel_hozac Ashsong: so you were only calling it with one argument from the python wrapper? 1187909230 M * Ashsong daniel_hozac: correct. 1187909240 M * daniel_hozac yeah, that's undefined behaviour :) 1187909258 M * daniel_hozac you're lucky it didn't segfault. 1187909268 M * daniel_hozac (or maybe that's unlucky...) 1187909273 M * Bertl :) 1187909315 M * Ashsong Yeah, well... :) 1187909348 M * daniel_hozac Bertl: besides the OOM-bias, do you think we'll be adding more interfaces to 2.3? or once i've got that implemented and tested, should i release 0.30.214? 1187909401 M * daniel_hozac of course, if we realize we need more interfaces and tool support, releasing 0.30.215 shortly thereafter shouldn't be a big deal. 1187909801 M * Ashsong Bertl: How do I use vcmd to negate a single bit? 1187909827 M * daniel_hozac set flags to 0, mask to the bit you want to remove. note that you can use vattribute too. 1187909828 M * Ashsong Or do I just set hte mask appropriately? 1187909842 M * Ashsong mm, interesting. 1187909844 M * Ashsong Thanks. 1187909871 Q * the-dude Quit: Is that a kangeroo? 1187909988 M * Bertl Ashsong: ~ and ! negates the entire expression 1187910003 M * Bertl Ashsong: so, ~^21 is everything except bit 21 1187910081 M * Ashsong Indeed, so I discovered. :) 1187910091 M * Ashsong Unfortunately, the other flags are fairly messed up too. 1187910143 M * Ashsong Time for a reboot, I think. 1187910444 M * Ashsong Bertl: thanks, supplying flags made everything happy. 1187910454 M * Bertl excellent!