1208304151 Q * hparker Quit: Read error: 104 (Peer reset by connection) 1208305605 M * PhatJ anyone got any idea how I can get ubuntu gutsy setup as guest on debian. adding --debug shows a lot of stuff, but nothing obvious shows a failure 1208307325 Q * Infinito Quit: Leaving 1208307573 Q * dowdle Remote host closed the connection 1208308940 Q * bronson_ Ping timeout: 480 seconds 1208309169 J * bronson ~bronson@adsl-68-122-117-135.dsl.pltn13.pacbell.net 1208309743 J * dowdle ~dowdle@71-32-18-216.blng.qwest.net 1208319225 J * ktwilight ~ktwilight@87.66.199.120 1208319503 Q * ktwilight_ Ping timeout: 480 seconds 1208319797 Q * PhatJ Ping timeout: 480 seconds 1208320122 J * PhatJ ~PhatJ@24-231-253-65.dhcp.aldl.mi.charter.com 1208322843 J * sharkjaw ~gab@64.28.12.166 1208323736 J * Slydder ~chuck@194.59.17.53 1208324934 M * padde seems dhozac has some problems with his server... can't get a connection to rpm.hozac.com 1208325004 Q * besonen_mobile_ Ping timeout: 480 seconds 1208325005 M * Slydder morning all 1208325084 J * meandtheshell ~sa@d91-128-17-34.cust.tele2.at 1208327389 J * yarihm ~yarihm@mtec-hg-docking-2-dhcp-130.ethz.ch 1208327955 Q * quasisane Remote host closed the connection 1208329306 J * cryptronic ~oli@851022214.cryptronic.de 1208329309 M * cryptronic hi all 1208329356 M * cryptronic is someonw here with a centos nodesystem, who can show me the content of /proc/version 1208329796 J * rgl ~rgl@bl8-132-113.dsl.telepac.pt 1208330237 J * JonB ~NoSuchUse@77.75.164.169 1208331956 J * mrfree ~mrfree@host1-89-static.40-88-b.business.telecomitalia.it 1208332872 Q * meandtheshell Quit: Leaving. 1208333239 J * snod ~snod@213.239.200.36 1208333240 J * meandtheshell ~sa@d91-128-17-34.cust.tele2.at 1208333369 J * hipe ~hipe@BSN-77-69-204.dsl.siol.net 1208333408 M * snod hi there, i have an etch vserver running in an etch host. i installed java but i get a SIGSEGV when i try to execute it. any ideas what might be wrong? 1208333474 Q * meandtheshell 1208333480 M * daniel_hozac and installing it on a regular etch box works fine? 1208333494 M * daniel_hozac padde: there's some routing issue i haven't yet managed to track down. 1208333507 M * daniel_hozac or, bothered to track down is more like it... 1208333522 M * padde daniel_hozac: i see. thanks for the update :) 1208333549 M * hipe hello, i have lenny host (amd64) and a vserver and cant run a x server in a vserver 1208333583 M * padde daniel_hozac: if i can be of any help (traceroutes from another host... temporary webspace, ...), tell me 1208333610 M * daniel_hozac i've got plenty of accounts on other boxes, so i'm good ;) 1208333627 M * padde daniel_hozac: i thought so ;) 1208333630 M * snod daniel_hozac: i guess so, gonna give it a try 1208333686 M * daniel_hozac hipe: have you followed the instructions for doing so? you need to give it a whole bunch of capabilities, unhide stuff in /proc, etc. 1208333719 M * snod yeah. java works on the etch host 1208333784 M * hipe , i followed the instructions on wiki 1208333812 M * cryptronic daniel_hozac: could you just paste me the content of /proc/version of a centos system running with linux vserver? 1208333820 M * daniel_hozac cryptronic: why? 1208333865 M * cryptronic because i have troubles with a software which is only supported on centos getting this running on an debian nodesystem with centos guest 1208333933 M * daniel_hozac and my /proc/version would help because...? 1208333946 M * cryptronic because the software is checking /proc/version 1208333962 J * jsambrook ~jsambrook@aelfric.plus.com 1208333979 M * daniel_hozac then you'd probably want /proc/version from a _regular_ CentOS box. 1208333988 M * daniel_hozac Linux version 2.6.22.19-vs2.3.0.34.1 (mockbuild@x86-64.builders.hozac.com) (gcc version 4.1.2 20070626 (Red Hat 4.1.2-14)) #1 SMP Mon Mar 17 02:33:42 EDT 2008 1208334110 M * hipe i followed the instructions on old vserver wiki 1208334134 M * daniel_hozac hipe: and what problem are you facing? 1208334194 M * hipe I cannot see any ipcs in the vserver from host (shm) 1208334277 M * heanol is it possible to have a loopback device in each vserver? 1208334288 M * cryptronic thx daniel_hozac 1208334289 M * daniel_hozac hipe: chcontext --xid -- ipcs 1208334319 M * daniel_hozac heanol: define loopback device. 1208334333 N * DoberMann[ZZZzzz] DoberMann 1208334350 M * hipe vika:/etc/vservers# chcontext --xid 49152 -- ipcs 1208334350 M * hipe ------ Shared Memory Segments -------- 1208334350 M * hipe key shmid owner perms bytes nattch status 1208334350 M * hipe ------ Semaphore Arrays -------- 1208334350 M * hipe key semid owner perms nsems 1208334351 M * hipe ------ Message Queues -------- 1208334351 M * hipe key msqid owner perms used-bytes messages 1208334365 M * hipe ups - sorry 1208334370 M * heanol daniel_hozac: interface lo with address 127.0.0.1 1208334371 M * heanol :P 1208334380 M * daniel_hozac please use paste.linux-vserver.org for anything longer than 3 lines. 1208334387 M * daniel_hozac heanol: 2.3 has that. 1208334454 M * heanol ah 1208334496 M * hipe , http://paste.linux-vserver.org/11917 1208335683 J * pmenier ~pme@LNeuilly-152-22-72-5.w193-251.abo.wanadoo.fr 1208335823 J * MatBoy ~MatBoy@wiljewelwetenhe.xs4all.nl 1208336158 M * heanol without a loopback.. do you usually set localhost to the ip of the vserver? 1208336213 M * daniel_hozac that makes some services happier. 1208336719 M * hipe daniel_hozac: shm problem? 1208336732 M * daniel_hozac i didn't see any problem... 1208336756 M * hipe from vserver is not possible to see hosts shm? 1208336763 M * daniel_hozac no, of course not. 1208336814 M * hipe is it possible to have x server in host and vserver app connected to it? 1208336821 M * daniel_hozac sure. 1208336844 M * daniel_hozac just set DISPLAY=:0.0, as usual. 1208336861 M * hipe ok. 1208337604 J * quasisane ~sanep@c-75-68-59-175.hsd1.nh.comcast.net 1208338054 M * snod daniel_hozac: java works now. i think the problem was a incompatibility between java and some debian libs. i added security.debian.org to sources.list and upgraded some packages. now it seems to work 1208338067 M * daniel_hozac cool. 1208338207 Q * mrfree Quit: Leaving 1208338550 P * cryptronic 1208340048 M * bXi the vserver command is fomr the util-vserver package right? 1208340056 M * daniel_hozac yes. 1208340181 J * bonbons ~bonbons@2001:960:7ab:0:2c0:9fff:fe2d:39d 1208340454 Q * jsambrook Quit: Leaving. 1208340499 J * jsambrook ~jsambrook@aelfric.plus.com 1208340574 Q * jsambrook 1208340608 J * jsambrook ~jsambrook@aelfric.plus.com 1208340660 Q * jsambrook Remote host closed the connection 1208340692 J * jsambrook ~jsambrook@aelfric.plus.com 1208341092 Q * jsambrook Remote host closed the connection 1208341162 J * jsambrook ~jsambrook@aelfric.plus.com 1208341234 Q * jsambrook Remote host closed the connection 1208341299 J * jsambrook ~jsambrook@aelfric.plus.com 1208342488 J * friendly ~friendly@ppp59-167-94-13.lns2.mel6.internode.on.net 1208344138 M * hipe any ideas for a performance benchmark - cpu and disk for a vserver? 1208344319 M * daniel_hozac same as you'd use elsewhere? 1208344319 Q * derjohn Quit: by(t)e 1208344348 J * derjohn ~derjohn@80.69.41.3 1208344472 Q * JonB Quit: This computer has gone to sleep 1208344541 M * pmjdebruijn hipe: vserver's overhead is really neckligible 1208344591 Q * yarihm Quit: This computer has gone to sleep 1208346336 J * JonB ~NoSuchUse@77.75.164.169 1208346732 J * yarihm ~yarihm@vpn-global-151-dhcp.ethz.ch 1208346736 J * cryptronic ~oli@p54A3B03F.dip0.t-ipconnect.de 1208347499 N * Bertl_zZ Bertl 1208347504 M * Bertl morning folks! 1208347739 M * JonB hi Bertl 1208348053 J * ftx ~ftx@dslb-084-062-234-171.pools.arcor-ip.net 1208348182 J * meandtheshell ~sa@d91-128-17-34.cust.tele2.at 1208348208 Q * ftx 1208348564 Q * Aiken Quit: Leaving 1208348587 Q * dowdle Remote host closed the connection 1208348728 J * Julius ~julius@p57B25683.dip.t-dialin.net 1208350184 Q * JonB Quit: This computer has gone to sleep 1208350689 Q * meandtheshell Quit: Leaving. 1208350804 M * Hollow daniel_hozac: how do i configure the loopback stuff for 2.3 kernels? i have enabled lback_remap, but i still get connection refused when trying to connect to the loopback inside 1208350822 M * daniel_hozac how did you configure the kernel? 1208350841 M * Hollow CONFIG_VSERVER_AUTO_LBACK=y 1208350841 M * Hollow CONFIG_VSERVER_AUTO_SINGLE=y 1208350862 M * daniel_hozac disable CONFIG_VSERVER_AUTO_SINGLE for your next build, and set ~SINGLE_IP in nflags for now. 1208350891 M * daniel_hozac otherwise your services will bind to your first address, and 127.x.y.1 will be left unbound. 1208350915 M * Hollow so loopback does not count as a second device? 1208350916 M * Bertl but only with a single ip :) 1208350922 M * Hollow or second ip 1208350935 M * daniel_hozac it's a separate kind of IP. 1208350940 M * Hollow ok 1208350965 M * Hollow yay, works, thanks :) 1208350966 M * Bertl otherwise the single ip special casing would not be possible 1208351020 M * daniel_hozac Bertl: should we make CONFIG_VSERVER_AUTO_SINGLE depend on !CONFIG_VSERVER_AUTO_LBACK? having both set to auto doesn't make much sense. 1208351055 M * Bertl maybe, but OTOH, having both doesn't really hurt 1208351084 M * daniel_hozac except that people think the automatic lback stuff is broken :) 1208351093 M * Hollow :) 1208351093 M * Bertl if we disable the CONFIG_VSERVER_AUTO_SINGLE, you have to give SINGLE_IP in nfalgs 1208351155 M * Bertl but yeah, maybe folks do not care that much about efficiency nowadays .. 1208351232 M * Bertl (note for those reading this: I am referring to the overhead the non-single ip case has over the single one :) 1208351381 M * Bertl btw, is it just me or do distros (and applications) tend to get bloated (on linux) recently? 1208351448 M * Bertl I do remember having a desktop machine with 32MB of ram, recently I tried to install (several) linux distros on a 32MB machine (server, no apps) and not even the package management did work as expected 1208351458 M * daniel_hozac hehe. 1208351501 M * Bertl but actually that is harmless, compared to what monsters proprietary software developers produce ... 1208351530 M * heanol with grsec and vserver, you're only supposed to have the gradm-util and policies in the host? 1208351555 M * daniel_hozac yes. 1208351611 M * heanol right, thanks 1208351627 J * jakesolid ~jakesolid@h-64-105-248-183.miatflad.covad.net 1208351714 J * JonB ~NoSuchUse@77.75.164.169 1208351778 M * jakesolid How can I allocate more locks for the vservers inside a host machine? 1208351807 M * jakesolid i'm getting messages like the following: 1208351841 M * jakesolid [Sun Apr 13 04:09:01 2008] [crit] (28)No space left on device: mod_rewrite: could not create rewrite_log_lock 1208351862 M * JonB are there inodes left on the device? 1208351913 M * JonB df -i 1208351918 M * daniel_hozac that's most likely IPC shared memory. 1208351941 M * Bertl ipcs -s | grep apache | perl -e 'while () { @a=split(/s+/); print `ipcrm sem $a[1]`}' 1208351948 M * jakesolid yes 1208351998 M * jakesolid but i'm having problems with locks 1208352019 M * heanol why is it required to put an inode limit too when limiting the disk usage? 1208352043 M * JonB heanol: because some files are big, and other files are small 1208352088 M * Bertl heanol: inodes are a resource which could be abused/depleted too, and thus lead to DoS 1208352108 M * heanol ok 1208352129 M * JonB heanol: like making 23985739475139657975347354 empty files 1208352157 M * heanol it's just not as easy to determine how many inodes are normal to use like with diskspace :P 1208352173 M * heanol with disk space, i know the limit across my vservers is the 200GB disk :) 1208352178 M * Bertl df -i 1208352181 M * heanol JonB: wouldn't they still take up space? 1208352194 M * Bertl no, inodes are not accounted as disk space 1208352197 M * heanol Bertl: ah 1208352209 M * JonB heanol: yes, they would take an inode 1208352225 J * alex__ ~alex@62-249-237-101.no-dns-yet.enta.net 1208352228 M * alex__ hi guys 1208352236 M * alex__ on a guest im trying to setup auth / ident 1208352239 M * alex__ but im getting this error: 1208352261 M * alex__ error: binding to port 113: bind(): Address already in use 1208352263 M * jakesolid so how can i add more locks? 1208352287 M * daniel_hozac increase shm_max or whichever IPC-resource you're low on. 1208352362 M * alex__ daniel_hozac, with my ident issue on the guest, should i tie my ident to a specific port on the master/host? 1208352380 M * daniel_hozac to a specific IP. 1208352537 M * alex__ ok 1208352635 M * alex__ daniel_hozac, ah yeah, would you know where thats set? 1208352697 M * daniel_hozac depends on how you run the server. 1208352803 M * alex__ its on debian 1208352816 M * alex__ i cant seem to find the auth daemon cfg 1208352823 M * alex__ on /etc .... 1208352868 M * daniel_hozac probably it's just *inetd. 1208352905 Q * friendly Quit: Leaving. 1208352996 M * alex__ yeah 1208352998 M * alex__ i found ident.conf 1208353004 M * alex__ now hmm, to add the ip address..... 1208353441 M * Bertl if your identd is not able to set that, you can wrap it in a chbind call 1208353455 M * Bertl (on the host, to limit it to host ips) 1208353521 M * Guy- alex__: I don't think you can do it with inetd without chbind 1208353534 M * Guy- alex__: you might want to look at ipsvd or rlinetd 1208353566 M * daniel_hozac xinetd is perfectly capable of doing it too. 1208353715 M * alex__ hrm 1208353716 M * alex__ ok 1208353727 M * alex__ anyone with debian knowledge on this? 1208353885 M * alex__ okay ive install xinetd 1208354401 Q * sharkjaw Remote host closed the connection 1208354544 Q * balbir Ping timeout: 480 seconds 1208354665 M * alex__ hmm interesting 1208354672 M * alex__ ive install xidentd on both master and guest 1208354677 M * alex__ using this guide: 1208354682 M * alex__ http://www.xinetd.org/sample.shtml 1208354743 M * alex__ i have binded the master onto IP: 87.118.112.90 1208354752 M * alex__ and guest on: 87.118.113.90 1208354777 M * alex__ but on guest comes up with the error: bind failed (Address already in use (errno = 98)). service = ident 1208354938 Q * virtuoso Ping timeout: 480 seconds 1208355198 J * balbir ~balbir@122.167.181.56 1208355498 Q * FireEgl Quit: Leaving... 1208355695 N * Bertl Bertl_oO 1208355698 M * Bertl_oO bbl 1208355756 M * daniel_hozac alex__: and netstat -pnlt on the host shows that it's bound to the right address? 1208355793 M * alex__ yeah got it working :) 1208355798 M * alex__ now, another subject 1208355802 M * alex__ im getting this bugger: 1208355810 M * alex__ km31919:~# ntpdate 0.debian.pool.ntp.org 1208355810 M * alex__ 16 Apr 16:23:32 ntpdate[29488]: the NTP socket is in use, exiting 1208355824 M * daniel_hozac you can't run NTP in a guest. 1208355830 M * alex__ for some reason my ntpdate has been buggered, noitced this when my time was out by 90seconds! 1208355836 M * alex__ thats on the master ... 1208355892 M * alex__ ahhh got it 1208356214 P * jakesolid 1208356600 J * grobie ~grobie@master.schnuckelig.eu 1208357726 Q * Slydder Quit: Leaving. 1208357993 J * virtuoso ~s0t0na@ppp78-37-250-182.pppoe.avangarddsl.ru 1208358625 J * dowdle ~dowdle@scott.coe.montana.edu 1208360550 J * Infinito_ ~argos@201-2-47-206.gnace701.dsl.brasiltelecom.net.br 1208361564 J * julius_ ~julius@p57B24E06.dip.t-dialin.net 1208361994 Q * Julius Ping timeout: 480 seconds 1208362368 Q * yarihm Ping timeout: 480 seconds 1208362398 J * yarihm ~yarihm@guest-docking-nat-1-029.ethz.ch 1208362499 N * DoberMann DoberMann[PullA] 1208362645 Q * julius_ Quit: Verlassend 1208363014 Q * JonB Quit: This computer has gone to sleep 1208363176 Q * Infinito_ Remote host closed the connection 1208364711 J * ftx ~ftx@dslb-084-062-234-171.pools.arcor-ip.net 1208364860 Q * ftx Remote host closed the connection 1208365320 Q * pmenier Quit: Konversation terminated! 1208365599 J * JonB ~NoSuchUse@77.75.164.169 1208365762 M * SpComb http://pb.paivola.fi/605 <-- I built a custom script that replaces `vserver ... build`, but I guess I forgot something, as `vserver ... start` raises that error 1208365852 M * daniel_hozac why did you do that? 1208365927 M * SpComb well, vserver-build doesn't offer an option to not touch the vdir at all (although an empty template/rsynch could proably fool it into doing that) 1208366078 M * daniel_hozac and /proc, /dev and /tmp just don't work with your solution? 1208366281 M * SpComb not sure what you mean with those, but I don't need them in my vservers 1208366310 M * daniel_hozac that's what skeleton creates. 1208366316 M * SpComb vserver-build seems to be geared towards building general-purpose servers, I want to build servers that are entirely unsuitable for general-purpose use 1208366378 M * SpComb the vservers ran fine without /proc and /tmp, and I can just create the /dev entries myself, not that much effort 1208366396 M * Bertl_oO I'd suggest to use the skeleton build method, and then simply remove the vdir 1208366439 M * Bertl_oO that'll ensure that the config is done properly and leaves you all the options of not having any 'conventional guest' 1208366528 M * Bertl_oO make sure that the init style reflects your guest's startup mechanism and that the guest has a proper xid assigned 1208366532 J * brc bruce@megarapido.cliquerapido.com.br 1208366612 M * bonbons Bertl_oO: looks like with 2.2.0.7 patch some guest apps can cause traffic 127.0.0.1 -> guest address (clamdscan it seems) 1208366629 M * daniel_hozac hmm? 1208366634 M * daniel_hozac i.e. a source address of 127.0.0.1? 1208366635 M * bonbons will investigate and see if I find the missing check 1208366637 M * bonbons yep 1208366647 M * daniel_hozac that's expected if you don't set CONFIG_VSERVER_REMAP_SADDR. 1208366670 M * bonbons sure? I would expect address not available for that case 1208366695 M * Bertl_oO why shouldn't it be available? 1208366704 M * daniel_hozac the destination address is the important one. 1208366705 M * bonbons not inside the guest 1208366724 M * Bertl_oO the guest ip should not be available inside the guest? 1208366746 M * bonbons no, 127.0.0.1 sould not be available to the guest 1208366756 M * daniel_hozac then set CONFIG_VSERVER_REMAP_SADDR. :) 1208366757 M * bonbons the destination is the guest's address as expected 1208366937 M * bonbons so no way to get apps to fail when they try to use 127.0.0.1? 1208366951 M * daniel_hozac no. 1208366954 M * daniel_hozac they'll always work. 1208366968 M * daniel_hozac it'll just be rewritten. 1208367039 M * bonbons hm, IMHO not a clean solution 1208367059 M * daniel_hozac making apps using 127.0.0.1 work is unclean? 1208367199 M * bonbons no, allowing apps to use 127.0.0.1 tough it's not in the list of addresses available to the guest (using it if marked available is fine of course) 1208367323 M * SpComb seems it was was just a case of me using symlinks to "../foo" in my config 1208367338 M * SpComb (save_ctxinfo: open("../lockfile"): File exists) 1208367447 M * Bertl_oO bonbons: draft up a working patch, and we'll discuss it :) 1208367460 M * bonbons ok :) 1208367604 M * daniel_hozac bonbons: you might want to look at 2.3. you can disable the loopback completely there. 1208367743 M * bonbons daniel_hozac: 2.3 is not yet scheduled for my server, some more testing to be done on another box first 1208367810 M * daniel_hozac you might want to prioritize that then :) 1208368039 J * Slydder ~chuck@dslb-088-072-030-031.pools.arcor-ip.net 1208368188 M * bonbons maybe, though I can't reproduce use of 127.0.0.1 with netcat, weird! 1208368215 M * daniel_hozac so you're connecting to 127.0.0.1? 1208368419 M * bonbons it always happens to connect to/from guest's address (as seen by netfilter) 1208368464 M * bonbons and that even though I tell it 127.0.0.1 1208368474 M * daniel_hozac nc -l 0.0.0.0 6666 & nc 127.0.0.1 6666 should reproduce, IMHO. 1208368590 M * bonbons yep, that does 1208368734 Q * _gh_ Ping timeout: 480 seconds 1208368734 Q * hijacker__ Read error: Connection reset by peer 1208368805 J * hijacker__ ~hijacker@213.91.163.5 1208368812 M * yarihm Bertl_oO, daniel_hozac , in our company we'd require that all vservers on a host have - in case of ressource starvation - the same amount of physical memory available and the same amount of swap. to get that, I guess it would be needed to have the processes that are allocated in a region above the soft-limit be swapped while the others below the soft-limit be not. daniel_hozac said once that this might not be very difficult so i convinced 1208368812 M * yarihm my boss that instead of migrating to something else (which costs a lot given our deployment) we ask for a quote what it would cost to hire someone implement this. is this a reasonable approach to the problem? 1208368892 M * Bertl_oO well, I'd love to implement that for your boss, given that he pays for the time, but IMHO that's a step in the wrong direction 1208368900 M * yarihm in other words, if Bertl_oO thinks the proposal makes sense, can someone of you be hired to implement that? 1208368915 M * yarihm Bertl_oO, i see. why do you think so? I'm very interested in this :) 1208368940 M * Bertl_oO because there is no real point in swapping out a guest, when there is no system-wide need to swap out anything 1208368949 M * yarihm yes, i completely agree 1208368972 M * Bertl_oO more than that, this behaviour, will leed to a lot of unnecessary swapping 1208368977 M * Bertl_oO *lead 1208368985 M * yarihm it's not that we always want to swap guests once above the soft-limit, but only if physical memory gets low 1208368996 M * Bertl_oO that will happen by default :) 1208369016 M * Bertl_oO i.e. already implemented by not changing anything in the kernel :) 1208369064 M * daniel_hozac i think the point is that guests above the soft-limit should have an increased likelyhood in being swapped out. 1208369080 M * yarihm Bertl_oO, are you sure? we kind of feel that memory distribution is a kind of "most greedy / first come" first serve allocation. the guest / customer that allocates the most and uses it the most intense, will have all his processes in ram while others will have all of theirs in swap. that's not very good in a shared environment 1208369082 M * Bertl_oO something which would make sense to add/improve on is a mechanism which ensures that guests with high memory usage get swapped out earlier than one with low memory usage (relative) 1208369120 M * Bertl_oO daniel_hozac has the better wording for that :) 1208369177 M * Bertl_oO yarihm: but in general, I'd say those guest which are active will also keep more data in memory 1208369197 M * yarihm well, at the moment guests above soft-limit do not have an increased likelyhood in being swapped out, right? (preferrably they would be forced to swap out if some other context needs physical ram up to his soft limit) 1208369232 M * yarihm Bertl_oO, well that makes perfect sense if you are alone on a machine. but we see customers that run all hell on such a vserver and these will ruin the platform for all other customers 1208369237 M * JonB this might leed to unneasesary swapping 1208369270 M * yarihm JonB, well, on an overall performance PoV, but not if you are looking at it from a customer's PoV 1208369310 M * yarihm i have no problem seeing a context use up all physical memory if noone else needs it 1208369335 M * yarihm but what if some other context needs some fast memory, too? 1208369341 M * JonB because even if one guest is very active and uses alot of memory, and there are some idle guests, then you might end up swapping out the active host for some of the idle hosts if a 2. host gets active 1208369358 M * JonB yarihm: put enough memory in the machine/limit the memory usage 1208369379 M * yarihm JonB, well, if noone else needs physical ram, then the context may have it for itself, that's granted 1208369393 M * daniel_hozac there's no such thing as "enough RAM" when it comes to customers :) 1208369402 M * JonB daniel_hozac: i know 1208369437 M * Bertl_oO okay, I have to leave now for a moment, brb 1208369475 M * JonB yarihm: but if someone else uses the RAM, and you run with increased likely hood of the very active guest to be swapped out, then you might swap out the active guest and not the idle guests 1208369494 M * JonB yarihm: swapping out just to swap in again will also kill performance 1208369522 M * yarihm JonB, well, what exactly would you suggest then? say you have a machine with 4GB and you want all customers to have 256MB. that will lead you to 16 customers and you then set the hard-limit to 256MB. OK now, that's obviously bad, because 256MB is really low for nowday's applications. fine you say, set the hard limit to 2GB and add enough swap. fine, but some 5 bastards will set their mysql to allocate 1GB of ram and once the next customer 1208369522 M * yarihm is coming along he will only get swapspace or what? 1208369578 M * JonB yarihm: asuming i have a hugh number of identical physical machines 1208369597 M * JonB i would look at the average number of guests pr. machine 1208369623 M * JonB i might for some time have a machine with few customers, if some other machine made more money 1208369628 M * yarihm we need to be able to set an upper bound on how much physical memory a context can allocate IF (and only IF) available physical memory gets low 1208369650 M * JonB but if the customer keeps on using, i would tell the customer to buy a bigger virtual machine 1208369684 M * JonB a bigger guest 1208369700 M * yarihm but you do see that this does not scale, right? you have to watch all machines and call back customers that over do it. you deal with complaints of the other customers instead of having a system that just sets proper limits 1208369733 M * JonB yarihm: i'm not sure that you can have the system make all customers happy 1208369750 M * yarihm JonB, well, you could have a system that is a bit fairer 1208369756 M * JonB i think the number 1. priority is to avoid swap. It is a waste of resources 1208369811 M * daniel_hozac nobody's arguing that. 1208369823 M * JonB daniel_hozac: i know 1208369829 M * daniel_hozac it's just that once you start swapping, guests over the soft limit should be preferred. 1208369834 M * yarihm JonB, so do I. the question just is who gets swapped out when things get tight. and at the moment, we feel that it's the wrong ones. don't get me wrong, i do not opt for this feature be the default, but it should be available for shared hosting environments 1208369847 M * JonB daniel_hozac: i think that might leed to more swapping 1208369869 Q * PhatJ Remote host closed the connection 1208369886 M * JonB yarihm: how often do RAM usage get tight? 1208369893 M * yarihm JonB, we don't care if that leads to more swapping. we just want those that are over their limits get swapped out, not the ones that are actually not even getting what they pay for 1208369923 M * JonB yarihm: but swapping out will cost system resources too 1208369925 M * yarihm JonB, well, virtually always once people start running big webapps and databases in their vservers 1208369971 M * JonB yarihm: then i think you might oversold your capacity on a single machine 1208369979 M * yarihm JonB, yes, that is unavoidable. but still that is better than those poor bastards only using 50MB of RAM be swapped out so that the one that is way over his limits can have the freedom of physical memory just because he uses it overduly anyway 1208370013 M * JonB yarihm: the small customers will also pay when the CPU uses resources to copy data back and forth 1208370061 M * JonB all customers pays that cpu penalty 1208370064 M * yarihm JonB, no, we don't oversell these machines. they are not thought to be huge application-servers, they are intended to be run as secondary MX, small mailserver, whatever. But in order to have this operable, you will have to punish these that are way over their limits and not grant them the more memory the more they use 1208370103 M * JonB yarihm: increase the price of those that uses too much 1208370133 M * JonB where i rent my vserver guest, they wrote that if i use too much i will have to up a class 1208370148 M * daniel_hozac that's a lot of administration. 1208370172 M * yarihm JonB, again, that does not scale. these people should feel for theirselves that the server is not getting them what they want. We don't want the other 13 customers first complain and create tickets and make calls so that we can (again by hand) go after that one bastard that causes the mess 1208370183 M * JonB daniel_hozac: yes, it might be 1208370203 M * JonB yarihm: you can monitor the usage and go after people before getting complaints 1208370210 M * yarihm JonB, no it absolutely is and if you offer vservers for 20 bucks the month you can't afford such things 1208370244 M * JonB yarihm: i buy mine for $6, maybe $8 a month 1208370255 M * yarihm JonB, why do you have such a problem with the system providing the option (you can do whatever you want with your vserver, we do not say that you must do it that way) that these users be limited 1208370266 M * JonB yarihm: i dont have a problem 1208370280 M * yarihm JonB, why then are you argueing with me asking for this feature to be implemented? 1208370344 M * JonB yarihm: all i'm trying to say is that this might lead to more swapping, and that you might not be able to solve the problem anyway, they might still use the resources, and then the server just swaps in and out 1208370357 M * yarihm look, my boss is about to switch to openvz because they provide this option even out of the box. I see why he wants to do that: because at the moment it is a hassle, lots of complaints and unfair ressource allocation from a customer POV. I just don't want that switch to happen and think that the money it costs is better invested improving linux-vserver 1208370377 M * JonB yarihm: i agree with that part 1208370379 M * yarihm JonB, yes, I am aware of this. 1208370385 M * JonB yarihm: but try it on 1208370410 M * JonB yarihm: i am sorry if i gave you the impression that i am against 1208370431 M * yarihm JonB, well, no problem, i just feel that we have different requirements :) 1208370491 M * JonB yarihm: maybe we do, i'm not a vserver guest reseller too 1208370501 M * yarihm i personally wouldn't need the feature because i'm all alone on my vserver-hosts. if one context uses too much ressources its the same user taking the ressources of the same user. 1208370520 M * JonB where i really use it is internally, and i have memory enough 1208370526 M * yarihm same here 1208370538 M * yarihm but not where I work :) 1208370557 M * JonB i use it where i work, and it is internally 1208370697 M * yarihm daniel_hozac, do you think this feature might be worth implementing / be along the concepts of linux-vserver? if neither you nor Bertl_oO nor any other developer do, that's fine. I mean I can't implement it :) but if you feel that this might be a good addition, we would pay for it to be implemented. Should I ask on the mailing-list maybe? 1208370753 M * daniel_hozac if i were you, i'd just wait until Bertl_oO gets back. 1208370784 M * daniel_hozac personally, i think this is something that's just been missing for a while, and ought to be corrected. 1208370852 M * yarihm ok. I have to leave now, but given that, I'll come back and ask again. thanks for the support so far! 1208370927 M * yarihm cu guys 1208370931 Q * yarihm Quit: Leaving 1208372457 Q * rgl Quit: Saindo 1208372886 J * larsivi ~larsivi@144.84-48-50.nextgentel.com 1208374310 J * meandtheshell ~sa@d91-129-52-41.cust.tele2.at 1208374725 J * _gh_ ~gerrit@166.129.176.226 1208376174 Q * Linus Quit: I'll by back 1208377750 Q * bonbons Quit: Leaving 1208378113 Q * ritter_ Quit: Leaving 1208378572 N * DoberMann[PullA] DoberMann 1208379009 Q * Slydder Quit: Leaving. 1208379324 J * Infinito ~argos@201-2-47-206.gnace701.dsl.brasiltelecom.net.br 1208380576 J * Aiken ~james@ppp121-45-192-61.lns1.bne1.internode.on.net 1208381985 Q * emag Remote host closed the connection 1208382001 J * emag ~Itoc5OI6@gurski.org 1208382765 N * DoberMann DoberMann[ZZZzzz] 1208382795 J * hparker ~hparker@linux.homershut.net 1208383726 Q * JonB Quit: This computer has gone to sleep 1208384447 Q * cryptronic Quit: Leaving. 1208384743 Q * larsivi Quit: Konversation terminated! 1208384857 Q * MatBoy Quit: Ik ga weg 1208384939 Q * Infinito Quit: Leaving 1208387060 Q * dowdle Remote host closed the connection 1208388597 M * Bertl_oO off to bed now .. have a good one everyone! 1208388602 N * Bertl_oO Bertl_zZ