1184198769 M * AStorm Bertl, hmm, testsuite for VServer would have to check basic startup on some distro images
1184198775 M * AStorm quota management
1184198784 M * AStorm shutdown, killing
1184198797 M * AStorm of course vps and vpstree
1184198820 M * AStorm barrier and unification
1184198902 M * Bertl and a number of more detailed checks too
1184198963 M * AStorm Mhm.
1184198980 M * AStorm scheduler behaviour would be the hardest to check
1184199031 M * AStorm I have to ask Ingo about stackable classes for CFS
1184199107 M * AStorm because this is what VServer does, right?
1184199125 M * AStorm The hard CPU scheduler would have to be ported as a scheduling class
1184199149 M * Bertl yes, that would be an option, I guess
1184199161 M * AStorm I'll try to write a simple throttler for CFS
1184199170 M * Bertl although I'm unsure about the 'hold' functionality (if that can be done with a scheduling class)
1184199179 M * AStorm yes, it can
1184199185 M * AStorm it means sleeping indefinitely
1184199208 M * AStorm the "classes" are much more like separate schedulers in CFS
1184199212 M * AStorm much more powerful
1184199247 M * AStorm so, yes, you can easily hold tasks
1184199260 M * AStorm just like you've done now
1184199277 M * AStorm the scheduler would have to be rewritten so as to not piggyback on Vanilla one
1184199391 M * AStorm full entitlement based one (if you want to reuse the current one)
1184199473 M * AStorm hmm
1184199501 M * AStorm the tasks ran in the vserver would have to use it too for scheduling to work
1184199611 M * AStorm the current design of CFS infrastructure is that it only grabs a task to be scheduled
1184199633 M * AStorm (returned by the tick function)
1184199747 M * AStorm scheduling class can alter the task's load estimation
1184199978 M * AStorm multiCPU behaviour is done by the scheduler
1184200489 M * Bertl hmm, what if I want to have different scheduling decisions based on the cpu?
1184200494 M * Bertl (as we currently have)
1184200636 M * AStorm hmm, well, there are affinity classes, but this doesn't affect the scheduling class
1184200648 M * AStorm s/affinity classes/affinity flags/
1184200656 Q * bzed Quit: Leaving
1184200664 M * AStorm You can decide based on the CPU task takes
1184200679 M * AStorm but that will interfere with load balancing in a weird way
1184200745 M * AStorm though I think you can mess with load estimator
1184201227 J * DoberMann_ ~james@AToulouse-156-1-78-185.w86-196.abo.wanadoo.fr
1184201333 Q * DoberMann[ZZZzzz] Ping timeout: 480 seconds
1184201853 J * oxylin ~jpeeters@chv78-2-88-161-189-78.fbx.proxad.net
1184202839 Q * oxylin Quit: Ex-Chat
1184203059 Q * slack101 Ping timeout: 480 seconds
1184203080 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184206076 M * Bertl okay, off for now .. back later
1184206081 N * Bertl Bertl_oO
1184207374 Q * slack101 Ping timeout: 480 seconds
1184207395 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184211449 Q * derjohn Ping timeout: 480 seconds
1184211626 J * derjohn ~derjohn@80.69.41.3
1184211699 Q * slack101 Ping timeout: 480 seconds
1184211720 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184214725 Q * Piet_ Quit: Piet_
1184215511 M * Bertl_oO okay, back now, but almost off to bed .. anything important?
1184215515 N * Bertl_oO Bertl
1184215959 Q * slack101 Ping timeout: 480 seconds
1184215980 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184218021 M * Bertl okay, off to bed now ... have a good one everyone!
1184218026 N * Bertl Bertl_zZ
1184219878 N * DoberMann_ DoberMann
1184220279 Q * slack101 Ping timeout: 480 seconds
1184220300 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184222785 N * DoberMann DoberMann[PullA]
1184222862 Q * quasisane Ping timeout: 480 seconds
1184224023 J * dna ~naucki@92-243-dsl.kielnet.net
1184224335 J * hosungs ~chatzilla@ee3222.kaist.ac.kr
1184224354 M * hosungs hello
1184224433 M * hosungs is anyone out there?
1184224557 Q * slack101 Ping timeout: 480 seconds
1184224578 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184224614 J * Baby ~miry@195.37.62.208
1184224614 M * hosungs hello
1184224755 J * bzed ~bzed@dslb-084-059-106-081.pools.arcor-ip.net
1184224924 J * hosungs_ ~hosungs@ee3222.kaist.ac.kr
1184224961 M * hosungs_ hello?
1184224980 M * hosungs just me???
1184225005 P * hosungs_ 
1184225063 J * hosungs_ ~hosungs@ee3222.kaist.ac.kr
1184225072 P * hosungs_ 
1184225219 J * hosungs_ ~hosungs@ee3222.kaist.ac.kr
1184225238 N * DoberMann[PullA] DoberMann
1184225298 P * hosungs 
1184225319 J * DavidS ~david@p57A4A23B.dip0.t-ipconnect.de
1184225463 M * hosungs_ should i ask my question anyway? okay. i'm trying linux vserver for the first time with 2.6 kernel. i built the kernel myself and installed the util-vserver. they seemed installed okay. however, after i rebooted with the new kernel and tried the testme.sh script, i get this error: "chcontext: tools were built without legacy API support; can not continue... chcontext failed!" can anyone tell me what to do?
1184225500 J * ema ~ema@rtfm.galliera.it
1184227057 J * rgl ~Rui@84.90.10.107
1184227059 M * rgl hello
1184227076 M * rgl can I normal use use the quota command?
1184227683 M * hosungs_ Okay. I searched the chat log, and found a (maybe) clue: --enable-apis=NOLEGACY. I recompiled the util-vserver with that ./configure option, and now that error (tools were built without legacy API support) is gone.
1184227742 M * hosungs_ However, now I'm getting a new error: "chcontext: vc_new_s_context(): Function not implemented" Again, I got this error from testme.sh...
1184227807 M * hosungs_ I suppose everyone has gone to bed or something. I'll be back later. Hope to get in touch with VServer experts. Bye for now.
1184228371 M * mattzerah what version of utils are you using hosungs_ ?
1184228447 N * DoberMann DoberMann[PullA]
1184228935 Q * DavidS Quit: Leaving.
1184229142 J * meandtheshell ~markus@85.127.103.255
1184229443 Q * Aiken Remote host closed the connection
1184229486 Q * emtty Read error: Connection reset by peer
1184229936 J * Aiken ~james@ppp121-45-220-241.lns2.bne1.internode.on.net
1184230488 M * derjohn hosungs_, do you have a vserver patched kernel running? What does testme.sh say ?
1184230860 J * lilalinux ~plasma@dslb-084-058-201-078.pools.arcor-ip.net
1184231927 J * HeinMueck ~Miranda@host-88-217-199-211.customer.m-online.net
1184232495 Q * Aiken Remote host closed the connection
1184232885 J * DavidS ~david@p57A4A23B.dip0.t-ipconnect.de
1184232995 Q * esa Remote host closed the connection
1184233022 J * esa ~esa@ip-87-238-2-45.adsl.cheapnet.it
1184233139 Q * slack101 Ping timeout: 480 seconds
1184233160 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184233489 M * rgl oh.  the last night warning on guest shutdown about quota was because the scripts at /etc/rcS.d are not run in the guest.  so, the /etc/init.d/quota script wasn't being run to activate the quotas inside the guest.    so, I guess its normal for vserver to only run the current runlevel scripts (at /etc/rc3.d/*)?
1184233554 M * rgl gag.  normal runlevel on debian/ubuntu is 2 not 3.  vserver defaults to 3 then?
1184233942 M * DavidS debian by default has all runlevels the same, if you use "update-rc.d defaults" to install a service, it will be started on all runlevels
1184233971 M * harry not all
1184233984 M * harry btw, rgl ... you can change default runlevel in /etc/inittab
1184234004 M * harry DavidS: i don't want my services started in runlevel 0, 1 or 6 ;)
1184234016 M * rgl harry, yup.  I did it now :)
1184234067 M * DavidS harry: yeah "all" meaning in this case 2..5 
1184234120 J * cedric ~cedric@80.70.39.67
1184234621 M * AStorm Hmm, where should I put the mtab for the VServer?
1184235027 M * rgl AStorm, /etc/vservers/ocelot/apps/init/mtab
1184235035 M * rgl (ocelot is my vserver name)
1184235085 M * AStorm mhm, apps/init
1184235151 M * rgl AStorm, http://www.nongnu.org/util-vserver/doc/conf/configuration.html :)
1184235170 M * AStorm hmm
1184235179 M * AStorm thanks for the doc
1184235204 M * rgl its the flower power page :D
1184235237 M * rgl AStorm, are you using quota?
1184235244 M * AStorm rgl, not yet
1184235270 M * rgl AStorm, do you known if a normal user can run the quota command?
1184235315 M * AStorm he can't
1184235327 M * AStorm it requires direct access to the filesystem
1184235327 M * rgl I think it should, or else he doesn't known how much quota he's using.  though, for doing that, the /aquota.user needs have read permissions to everyone.
1184235341 M * AStorm uhm, he can read the quota
1184235346 M * AStorm df will tell you (I think)
1184235353 M * rgl df?
1184235363 M * AStorm uhm, no, sorry
1184235466 M * rgl humm, I can't change the permissions of /aquota.user.  humm, maybe that can only be done when quotas are off.
1184237205 J * Aiken ~james@ppp121-45-220-241.lns2.bne1.internode.on.net
1184237442 Q * slack101 Ping timeout: 480 seconds
1184237463 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184238145 Q * ktwilight_ Read error: Connection reset by peer
1184238668 M * rgl its a good ideia to run quotacheck before the guest run quotaon?
1184238773 J * ktwilight ~ktwilight@12.104-66-87.adsl-dyn.isp.belgacom.be
1184239294 J * quasisane ~user@c-75-67-252-184.hsd1.nh.comcast.net
1184239313 M * arachnist  /w 24
1184239437 J * Ramjar ~ramjar@195.159.98.150
1184239444 J * renihs ~penguin@83-65-34-34.arsenal.xdsl-line.inode.at
1184239465 M * Ramjar why do i sometimes lose my IP in vserver guest? suddenly (once in a month or something) i lose my IP (whole interface in guest disappears)
1184239480 M * Ramjar (kernel linux-2.6.18-vserver-2.1.1-r2)
1184239708 M * ard maybe another vserver has the same ip, and that one has been started and brought down
1184239718 M * ard is the ip still there in the root server or not?
1184239729 M * ard if not, then that must be your problem :-)
1184239904 M * Ramjar well no one else has that IP. but the strange thing is that the whole infterface disappears.
1184239935 M * Ramjar if someone else /another guest has that ip the interface shouldn't disappear.
1184239990 M * Ramjar is it able to find a gentoo-installation with vserver preinstalled? *easy as Centos or debian installation?* i have reinstalled with Centos Xen, but i really wanna get vserver to work.
1184240075 M * LaZaR i have a gentoo readme in germany for vserver
1184240406 Q * Aiken Quit: Leaving
1184241724 Q * slack101 Ping timeout: 480 seconds
1184241745 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184243177 Q * lilalinux Remote host closed the connection
1184243533 J * lilalinux ~plasma@dslb-084-058-201-078.pools.arcor-ip.net
1184244053 Q * ema Quit: leaving
1184244413 Q * HeinMueck Ping timeout: 480 seconds
1184245139 M * daniel_hozac LaZaR: is the one on gentoo.org not good enough for you?
1184245297 J * flea ~flea@a83-132-13-23.cpe.netcabo.pt
1184245308 M * daniel_hozac Ramjar: Gentoo doesn't really come with anything pre-installed... you might want to read http://www.gentoo.org/proj/en/vps/vserver-howto.xml
1184245364 M * flea howdy ppl
1184245370 M * flea daniel_hozac, Bertl_zZ [[]]
1184245479 M * daniel_hozac Ramjar: as for your problem, that might happen if the primary IP address is removed, and you don't have promote_secondaries enabled.
1184245505 M * Ramjar daniel_hozac thnx. i have done that once. but it thaks a loooong time :) so i just thought maybe someone have done a preinstall or something.
1184245509 M * Ramjar but now i knoe :)
1184245572 M * daniel_hozac well, that's the Gentoo-spirit ;)
1184245779 M * flea Ramjar: you have a stage3 prepared for vserver, with the correct baselayout version
1184245803 M * flea Ramjar: you just have to grab it and untar it and you have youserf the base you need
1184245806 M * daniel_hozac it's a stage4.
1184245810 M * daniel_hozac but that's for guests.
1184245818 M * flea or a stage4... it also has it.
1184245822 M * flea I use the stage3
1184245833 M * daniel_hozac (i had assumed this was about the host)
1184245844 M * flea sorry, maybe I'm off topic
1184245853 M * flea *oops* :S
1184245854 M * rgl woah.  stage4 is new for me heheh.    when I used gentoo there was only stage3-
1184246216 M * Ramjar oh
1184246226 M * Ramjar can i run debian etc on vserver guest?
1184246243 M * daniel_hozac yes.
1184246253 M * daniel_hozac the guest distro doesn't really depend on the host's.
1184246266 M * Ramjar i see. so i can run windows to :P
1184246276 M * daniel_hozac Windows runs on Linux now?
1184246294 M * Ramjar hehe yes. didnt you know? *just kidding*
1184246344 M * Ramjar but can i run solaris on vserver? (unix)
1184246356 M * daniel_hozac does it run on Linux?
1184246363 M * Ramjar hehe no, i know :)
1184246378 M * Ramjar okey so it is spec. Linux and nothing else
1184246383 M * Ramjar me like vserver anyway :)
1184246567 M * eyck- hmm, with vmware,qemu or vbox you can run solaris or win$$ on linux
1184246658 M * eyck- so, in this way, you can run solaris on vserver
1184247350 Q * quasisane Quit: ERC Version 5.2 (IRC client for Emacs)
1184247828 M * AStorm eyck-, hmm, as long as you don't have to give special permissions to them
1184247849 M * AStorm which you shouldn't have to (just create some device nodes)
1184248515 M * AStorm Hmm, are there any good howtos for syslog-ng configuration in vserver setting?
1184248612 M * LaZaR comment out: destination console_all { file("/dev/tty12"); };
1184248621 M * LaZaR #log { source(src); destination(console_all); };
1184248640 M * LaZaR remove pipe(/proc/kmsg) from source src { unix-stream("/dev/log"); internal(); };
1184248760 M * AStorm uhm, I want to put logging inside another vserver
1184248765 M * AStorm and network the other syslogs
1184248791 M * LaZaR hm dont know about this
1184248794 M * AStorm you know, external logging
1184248814 M * daniel_hozac so, do it?
1184248834 M * daniel_hozac nothing vserver-specific about that...
1184248922 M * AStorm I was hoping for some nice howto to do that in 30s :P
1184248949 M * AStorm oh, there are some :P
1184248966 M * daniel_hozac how unexpected.
1184248977 M * arachnist onaka ga suita
1184249081 M * AStorm daniel_hozac, hmm, can vservers sniff each other's networking?
1184249148 M * daniel_hozac they can't sniff at all.
1184250141 M * AStorm blah, you're right, no NET_ADMIN rights
1184250149 M * daniel_hozac more like no CAP_NET_RAW.
1184250224 M * AStorm If the vserver had that right, could it sniff?
1184250230 M * daniel_hozac yes.
1184250266 Q * flea 
1184250307 Q * slack101 Ping timeout: 480 seconds
1184250328 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184250388 M * AStorm I wonder if I could unify just a given list of files
1184250402 M * daniel_hozac sure.
1184250441 M * AStorm what will unify do on different files? does it check for it?
1184250448 M * daniel_hozac check for what?
1184250457 M * daniel_hozac if the files differ?
1184250477 M * AStorm yes
1184250484 M * daniel_hozac also, you know that vunify is deprecated in favor of vhashify, right?
1184250502 M * AStorm I know, but I don't want to hash everything
1184250510 M * daniel_hozac so, create an exclude list.
1184250519 M * AStorm uhm, you don't understand my use case
1184250531 M * daniel_hozac no, as i've said, i'm not psychic. :)
1184250532 M * AStorm I want to refresh the unification after system upgrade
1184250540 M * AStorm I have a list of modified files
1184250549 M * AStorm don't want to check hashes on everything
1184250553 M * AStorm or exclude other files
1184250563 M * daniel_hozac it won't check hashes on already unified files.
1184250569 M * AStorm hmm
1184250575 M * AStorm but the hashes will differ...
1184250581 M * daniel_hozac unless you use the refresh option.
1184250585 M * AStorm I want it to overwrite the files
1184250587 M * AStorm :P
1184250589 M * AStorm cp -a?
1184250600 M * daniel_hozac sure.
1184250612 M * AStorm hmm, no problem with it then.
1184250712 M * AStorm hmm
1184250737 M * AStorm could the next version of util-vserver have the --overwrite option for vhashify?
1184250749 M * daniel_hozac overwrite what?
1184250756 M * AStorm which would find different files, overwrite and unify
1184250770 M * daniel_hozac you realize hashify is a one-guest operation, right?
1184250770 M * AStorm (except the exclude list, of course)
1184250779 M * AStorm daniel_hozac, yes, it is
1184250790 M * daniel_hozac so an "overwrite" doesn't make sense.
1184250810 M * AStorm hmm, wait wait. So, it only unifies files on a single guest?
1184250815 M * AStorm Not between guests?
1184250818 M * daniel_hozac yes.
1184250841 M * AStorm Hmm
1184250860 M * AStorm So, I'll have to write a tool to update COW guests then?
1184250881 M * daniel_hozac "rsync"?
1184250910 M * AStorm uhm, then remark the files as COW
1184250938 M * daniel_hozac it really sounds like you want vunify though.
1184250954 M * daniel_hozac it has the whole reference-guest concept.
1184250972 M * AStorm Exactly
1184251050 Q * renihs Remote host closed the connection
1184251050 M * AStorm the question still stands - does it check (in manual mode) for file identity?
1184251075 M * daniel_hozac i have no idea. i've never used it, nor looked at the code.
1184251083 M * AStorm Hmm. So I'll have to.
1184251114 M * daniel_hozac for it to make sense, it should be overwriting the destination guest with whatever's in the reference one though.
1184251148 M * AStorm no, it ignores different files
1184251157 M * AStorm (according to the code)
1184251249 M * AStorm checks the stat data
1184251292 M * AStorm So, it's cp -a for me then :P
1184251318 M * daniel_hozac hmm?
1184251567 J * rgl_ ~Rui@84.90.10.107
1184251572 M * AStorm cp -a + vunify call
1184251591 M * AStorm of course, the vserver will have to be shut down for the update
1184252018 Q * rgl Ping timeout: 480 seconds
1184252685 N * Bertl_zZ Bertl
1184252689 M * Bertl morning folks!
1184253135 M * AStorm hmm, evening here ;-)
1184253169 M * daniel_hozac morning Bertl!
1184253212 M * daniel_hozac any ideas regarding the disk limit testing?
1184253278 M * Bertl well, what I saw from your patch, looks good
1184253306 M * Bertl haven't found the time to actually try it, will do so tonight
1184253345 M * Bertl IMHO the problem will be with reservations
1184253390 M * daniel_hozac yeah, that stuff is tricky.
1184253422 M * Bertl that is actually what stopped me last time when I tried to do a fully automated test suite for this :)
1184253444 M * Bertl but I ahve to admit, it was really half hearted ...
1184253650 M * daniel_hozac hehe.
1184254386 J * emtt1 ~eric@dynamic-acs-24-154-33-109.zoominternet.net
1184254529 M * Bertl wb emtt1!
1184254589 M * rgl_ Hey Bertl!   I figured what the problem was with yesterday problem I was having with quota when the guest was shutingdown *G*
1184254609 Q * slack101 Ping timeout: 480 seconds
1184254630 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184254632 M * Bertl rgl_: let's hear!
1184254660 M * rgl_ Bertl, it was because the quotas weren't enable when the guest starts up.  the script that enables them was inside rcS.d/ which vserver-util do no run *G*
1184254697 M * Bertl ts ts ... but they were working as expected nevertheless :)
1184254698 M * rgl_ Bertl, util-verver "only" runs the scripts inside the configured runlevel, which by default is 3.
1184254718 M * Bertl rgl_: which is correct, as S is not of any relevance
1184254830 M * rgl_ Bertl, they seemed to be running.  which was odd *G*
1184254878 M * rgl_ Bertl, anyways, the fix was just to symlink the quota init script to run before any of the other sripts.  then, the warning was gone :)
1184254951 M * Bertl yeah, simply adding it to the proper runlevel should suffice too
1184254962 M * Bertl okay, off for dinner ... back shortly
1184255070 J * stefani ~stefani@flute.radonc.washington.edu
1184255991 Q * matti Ping timeout: 480 seconds
1184256241 Q * ensc Ping timeout: 480 seconds
1184256875 J * bonbons ~bonbons@2001:5c0:85e2:0:20b:5dff:fec7:6b33
1184257331 J * ensc ~irc-ensc@p54B4D165.dip.t-dialin.net
1184257364 M * daniel_hozac Bertl: http://people.linux-vserver.org/~dhozac/p/k/delta-cow-fix09.diff fixes an oops caused by chcontext --xid 42 touch <COW-file>, when the current directory is below the barrier.
1184257416 M * daniel_hozac (granted, that should never happen in real life, but nonetheless it's nice to cover our bases)
1184257438 M * DavidS it could happen with vserver-exec, no?
1184257443 M * daniel_hozac no.
1184257450 M * daniel_hozac that's chrooted.
1184257455 M * DavidS ah!
1184257458 M * daniel_hozac (and such doesn't hit the barrier)
1184257461 M * daniel_hozac +as
1184258112 J * lilalinux_ ~plasma@dslb-084-059-011-165.pools.arcor-ip.net
1184258232 J * lilalinux__ ~plasma@dslb-084-058-221-118.pools.arcor-ip.net
1184258484 Q * lilalinux Ping timeout: 480 seconds
1184258604 Q * lilalinux_ Ping timeout: 480 seconds
1184258687 M * DavidS vserver+puppet really rules
1184258712 M * DavidS i'm currently testing my deployment and a full test cycle (including all application setup) looks like this:
1184258715 M * DavidS vserver ldap.edv-bus.at stop && rm -Rf /var/lib/vservers/ldap.edv-bus.at/ /etc/vservers/ldap.edv-bus.at/ && vserver puppetmaster exec puppetca --clean ldap.edv-bus.at && puppetd --test && vserver puppetmaster exec puppetca --sign ldap.edv-bus.at && vserver ldap.edv-bus.at exec puppetd --test
1184258855 M * Bertl back now
1184258902 Q * slack101 Ping timeout: 480 seconds
1184258923 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184258933 M * Bertl daniel_hozac: okay, I guess we have a bunch of patches which want to go into a new release (and some pending issues) no?
1184258973 M * Bertl dilinger: what I mean is, could you put them into a dir so that I can integrate them easily?
1184259069 M * Bertl s/dilinger/daniel_hozac of course :)
1184260527 M * rgl_ DavidS, cool :)
1184260577 M * rgl_ DavidS, can you talk about your setup a bit?  I'm very interested in puppet, and vserver+puppet seems wonderful :D
1184260610 M * DavidS rgl_: svn co http://club.black.co.at:82/svn/manifests/trunk/ /tmp/davids_manifests
1184260611 M * DavidS ;)
1184260634 M * DavidS I have one host with several vserver guests, one of them is the puppetmaster ...
1184260636 M * rgl_ DavidS, gag!  even better :)
1184260716 M * DavidS it's a bit of a mess at the moment, because i'm still migrating to modules, but it should give you an overview .. most of the stuff in modules/ should be fairly usable
1184260724 M * mjt anyone tried to use unionfs with vservers?  It seems that vserver scripts are being confused about its "lower level" directories/branches...
1184260778 M * rgl_ DavidS, are you using debian?
1184260932 M * Bertl mjt: there is not much point in using unionfs, but IIRC, some folks are successfully using that
1184260950 M * mjt not much point?
1184260954 J * Pazzo ~ugelt@195.254.225.136
1184260991 M * Bertl mjt: well, there are two scenarios:
1184261009 M * Bertl - central update of many guests (no individual updates)
1184261019 M * Bertl - individual updates across guests
1184261039 M * slack101 good day Bertl 
1184261041 M * Bertl in the first case, you are better off with a read only bind mount of the shared data
1184261055 M * Bertl in the second case, you are _much_ better off with unification
1184261057 M * DavidS rgl_: yes, exclusively
1184261065 Q * cedric Quit: cedric
1184261079 M * mjt hmm
1184261090 M * mjt unification.. i haven't tried it yet
1184261091 M * rgl_ DavidS, I'm using ubuntu.  so everything should apply smootly :)
1184261118 M * DavidS rgl_: no,since i use "etch" in quite a few places .. but 
1184261121 M * Bertl wb slack101!
1184261131 M * AStorm mjt, well, here unionfs works with vservers as usual
1184261136 M * DavidS but that's perhaps all in the apt and dbp modules
1184261147 M * mjt but i'm a bit worried about upgrading glibc on 50+ guests and running vunify to bring the differences back again
1184261154 M * rgl_ DavidS, I'll have to look deepear into puppet in near future.  no more administration by "hand" for me! heheh
1184261172 M * Bertl mjt: the problem with unionfs is that the guests will diverge with each update, while unification will allow you to merge them transparently each week (or whenever you like)
1184261173 M * DavidS rgl_: see you on #puppet at freenode
1184261195 M * rgl_ DavidS, I known it :)
1184261196 M * mjt AStorm: well it works, but as i mentioned scripts gets confused.. and it works after telling which filesystems not to umount
1184261203 M * Bertl mjt: how do you update glibc on 50+ guests with unionfs?
1184261214 M * mjt i use it a bit differently
1184261223 M * mjt unionfs is readonly too
1184261242 M * AStorm mjt, is not :>
1184261244 M * mjt a thick layer on top of common image
1184261253 M * mjt thin err
1184261254 M * rgl_ DavidS, I known puppet a long time ago ;)
1184261265 M * AStorm Bertl, how do you do that on 50+ guests with vunify?
1184261270 M * AStorm My package manager is not supported :P
1184261285 M * mjt ie, a filesystems with all the components, unionfs with (mostly) configs specific to each guest
1184261292 N * rgl_ rgl
1184261315 M * Bertl mjt: no point in using unionfs then, bind mounts would suffice
1184261329 M * mjt just too many files to bind-mount
1184261332 M * AStorm Bertl, sharing configs is useful though
1184261344 M * mjt alot of stuff scattered inside /etc for example
1184261371 M * AStorm the proper way would be to unify the configs
1184261395 M * AStorm and use COW
1184261411 M * mjt ALL the files are read-only, so there will be no W in COW :)
1184261416 M * Bertl AStorm: there are tools to manage your guests (for most package managers) and there is a wrapper to execute something in a number of guests
1184261454 M * Bertl AStorm, mjt: if you are really concerned about the disk space during an upgrade, you can reunify after each guest upgrade
1184261475 M * mjt not a disk space, but the whole procedure becomes just too complex
1184261497 M * Bertl well, I don't see how it becomes simpler with unionfs
1184261520 M * mjt i chroot into the common root and update/install whatever i want. once.
1184261533 M * Bertl which is exactly the same you do with ro bind mounts
1184261535 M * mjt next i restart the servers if needed
1184261537 M * mjt yes
1184261546 M * Bertl so?
1184261562 M * mjt try to bind-mount 20+ DIFFERENT configs in /etc ;)
1184261566 M * rgl mjt, humm.  that way you share everything in all guest?  like, config and binaries?
1184261574 M * mjt no, not configs
1184261590 M * mjt configs lives in another directory - only the ones which are specific to each guest
1184261596 M * mjt and are union-mounted together
1184261606 M * Bertl mjt: I would simply mount /etc separately
1184261638 M * AStorm Bertl, it's _slow_ :>
1184261639 M * mjt yeah, and think why the hell my timezone is wrong after glibc upgrade for example :)
1184261646 M * AStorm I mean, the unification
1184261657 M * Bertl AStorm: and unionfs is _always_ slow :)
1184261660 M * AStorm I'll write some scripts which will just unify the changed files
1184261664 M * AStorm Bertl, uhm, no
1184261667 M * AStorm not that much :P
1184261681 M * Bertl AStorm: it adds a constant overhead to _any_ operation _all_ the time
1184261689 M * mjt there are common configs (/etc/timezone - maybe not a good example because it doesn't change often), which SHOUL be here in all guests. and there are configs specific to each guest, not shareable
1184261694 M * AStorm Bertl, the package manager in question is pkgcore - similar to portage
1184261710 M * Bertl mjt: that is configuration based on pure luck
1184261724 M * Bertl mjt: what if somebody modifies the 'shared' configs?
1184261726 M * mjt luck??
1184261741 M * mjt it will be propagated to all the guests
1184261742 M * Bertl mjt: then your update will fail for this 
1184261758 M * mjt but i repeat: the whole filesystem is read-only in all guests
1184261777 M * Bertl you are not making much sense here
1184261791 M * Bertl if it is read-only, then ro bind mounts are the way to go
1184261801 M * Bertl there will be absolutely no problem with updates
1184261807 M * mjt there's no
1184261815 M * mjt we're going in circles ;)
1184261822 M * Bertl if you want the guests to be able to modify the files
1184261827 M * mjt no
1184261832 M * Bertl you have to decide how to handle that on updates
1184261863 M * Bertl so, what's the point in using unionfs when everything is ro?
1184261887 M * mjt very easy management of guest-specific stuff
1184261904 M * Bertl sorry, doesn't make sense to me
1184261923 M * DavidS mjt: you modify it from the host, not from one of the guests?
1184261926 M * mjt one dir for each guest with its own files inside -- everything which is here replaces the same thing in common image
1184261940 M * mjt from host, yeah
1184261947 M * AStorm Bertl, it's an overlay
1184261952 M * mjt overlay, sure ;)
1184261956 M * AStorm much like COW
1184261964 M * mjt unionfs does COW
1184261972 M * AStorm mjt, not really
1184261980 M * AStorm it actually writes a separate file
1184261981 M * Bertl mjt: no, actually it doesn't
1184261992 M * AStorm and hides the original behind it
1184261997 M * mjt well, the end effect is like COW
1184262002 M * AStorm more or less
1184262004 M * Bertl nope, not at all
1184262011 M * mjt lol
1184262019 M * Bertl with CoW you get a completely _new_ file
1184262030 M * Bertl where the overlay now has _two_ files
1184262042 M * mjt in both cases we've 2 files
1184262044 M * Bertl and more important, every time you access a file
1184262059 M * Bertl unionfs has to check, what file to present to the user
1184262065 M * mjt yes
1184262071 M * Bertl (which adds the overhead)
1184262082 M * mjt all directory lookups will cost 2x of the simple case
1184262090 M * Bertl something like that
1184262097 M * mjt (in case 2 layers - with 3 layers, cost is 3x etc)
1184262115 M * Bertl thus CoW and unification is significantly faster
1184262117 M * AStorm unionfs is good for overlaying on RO filesystems
1184262123 M * AStorm and that's mostly it
1184262128 M * Bertl like CDs, yes
1184262143 M * mjt i use it alot for that very purpose here
1184262178 M * Bertl well, if you are running from a CD, then performance is not critical anyway :)
1184262241 M * Bertl mjt: trust me, I had a deep look at all kinds of unification/overlay systems
1184262298 M * mjt performance for reading a config file isn't critical either ;)
1184262299 M * Bertl http://vserver.13thfloor.at/TBVFS
1184262307 M * mjt (which is done once at startup)
1184262395 M * AStorm Bertl, well, unionfs will be in the mainline soon
1184262413 M * mjt . o O { aufs }
1184262435 M * mjt unionfs has issues which were discussed several times on lkml
1184262442 M * Bertl a lot of funny stuff is in mainline :)
1184262453 M * AStorm Bertl, yes, and most of the useful stuff isn't :P
1184262481 M * AStorm Write something like the VServer jail that will get accepted :>
1184262493 M * AStorm e.g. a small patch
1184262501 M * mjt Bertl: what's that double-backslashes everywhere on that TBVFS page?  Seems those should be newlines, no?
1184262511 M * AStorm though they love LSMs
1184262518 A * mjt hates LSMs
1184262520 M * Bertl AStorm: Jacques did that 4 years ago :)
1184262554 M * mjt i use vserver as "jail" here.  Like a DMZ on a single host
1184262558 M * Bertl mjt: yes, the wiki changed since, and I don't maintain those pages anymore
1184262584 M * AStorm Bertl, well, it isn't maintained
1184262634 M * Bertl hmm?
1184262674 M * AStorm Was that FServer or something?
1184262710 M * Bertl AStorm: no, Jacques did Linux-VServer before I took over maintainership
1184262718 M * AStorm ah, that
1184262742 M * AStorm hmm, wouldn't chroot + network virtualisation + capabilities be exactly what VServer is?
1184262753 M * AStorm (much like grsecurity-enhanced chroots)
1184262754 M * Bertl not at all
1184262768 M * AStorm if the chroot is secured, of course
1184262768 M * daniel_hozac pid space anyone?
1184262769 M * pusling AStorm: process seperation
1184262772 M * Bertl first, we do network isolation for a good reason (not virtualization)
1184262803 M * Bertl then, pid/user/ipc/uts isolation is required too
1184262805 M * AStorm daniel_hozac, hmm, since when VServer divides pid space?
1184262828 M * AStorm you're right
1184262828 M * Bertl AStorm: do you see the processes of other contexts?
1184262863 M * pusling talking about network isolation - can I as vserver root user add tun/tap devices ?  (eg for aiccu ipv6 tunnels)
1184262881 M * AStorm Bertl, ah :>
1184262885 M * Bertl pusling: nope, not from the guest
1184262894 M * AStorm pusling, not with CAP_NET_ADMIN
1184262897 M * AStorm *without
1184262949 M * daniel_hozac actually, you can add the interface.
1184262954 M * daniel_hozac you just can't set it up or use it.
1184262965 M * Bertl right :)
1184262998 M * daniel_hozac Bertl: so, it looks like breaking a link with COW when you're over your disk limit merely truncates the file.
1184263016 M * daniel_hozac i mean, when breaking the link gets you over the disk limit.
1184263031 M * daniel_hozac that doesn't seem right to me.
1184263035 M * daniel_hozac it should return -ENOSPC, no?
1184263036 M * Bertl which sounds somewhat correct, although unexpected
1184263062 M * daniel_hozac okay, how so?
1184263075 M * Bertl but yeah, a better fault handling would probably be a good idea
1184263115 M * Bertl but I think all we can do is papering over this, until mainline has an in kernel copy function
1184263149 M * Bertl maybe what we can do is the following (for this specific case)
1184263198 M * Bertl check that the space is available (dlimit), allocate it before doing the copy, then let the copy run without checks, and if it succeds, return the handle
1184263217 M * Bertl otherwise return the appropriate error condition on the open
1184263224 Q * slack101 Ping timeout: 480 seconds
1184263225 M * AStorm  /check/d
1184263232 M * AStorm just allocate and see if that fails
1184263243 M * Bertl daniel_hozac: or do you have a better suggestion?
1184263245 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184263257 M * Bertl daniel_hozac: also, how to handle real file-system full conditions?
1184263271 M * daniel_hozac same way, no?
1184263275 M * AStorm Bertl, same as quota full - don't break the link, return ENOSPC
1184263315 M * Bertl daniel_hozac: the problem there is that you cannot do the reservations (without going down to filesystem level)
1184263325 M * AStorm blah
1184263346 M * AStorm fallocate call is not yet implemeted anywhere, heh
1184263348 N * DoberMann[PullA] DoberMann
1184263356 M * AStorm maybe just write and then rename?
1184263372 M * AStorm as in the good old days :P
1184263387 M * Bertl ah, and take the big kernel lock *G*
1184263402 M * AStorm there's no lock anymore?
1184263418 M * DavidS AStorm: that's why it's called OPEN source ;)
1184263421 M * daniel_hozac Bertl: yeah, that's true... i do think it'd be better to fail than trashing the file though.
1184263441 M * Bertl daniel_hozac: no, I think we should handle failures on the cow link break properly
1184263456 M * Bertl daniel_hozac: i.e. if the link break fails for whatever reasons
1184263482 M * Bertl just undo as much damage as possible, and return the file as readonly or with special return code
1184263485 M * daniel_hozac Bertl: exactly.
1184263486 M * Bertl (maybe ENOSPC)
1184263518 M * daniel_hozac we need to get an error return from sendfile.
1184263539 M * Bertl yep, that is probably the tricky part :)
1184263579 M * Bertl but I think this way is actually doable
1184263660 M * daniel_hozac hmm, should be fairly straight-forward.
1184263681 M * daniel_hozac in the if (ret < 0) after vfs_sendfile in cow_break_link, can't we just add || ret != size?
1184263708 M * daniel_hozac well, another if would be required as we'd have to set ret to -ENOSPC, but the idea's the same.
1184263732 M * Bertl give it a try, let me know how it goes ...
1184263750 M * Bertl (read: sounds good :)
1184263758 M * daniel_hozac btw, i have stray <filename><weird character> files. that should never happen, right?
1184263802 M * daniel_hozac could be from my previous crashes, so i'm not ready to blame anything yet... will see if they reappear when i redo the tests.
1184263810 M * Bertl those are good signs of CoW breaks going wrong
1184263829 M * Bertl we do not clean up on CoW failures (yet)
1184263838 M * daniel_hozac oh really?
1184263852 M * Bertl at least it is on my todo list :)
1184263860 M * daniel_hozac yeah, that's right...
1184263992 M * AStorm Hmm, how does one kill a virtual context by hand?
1184263998 M * daniel_hozac vc_ctx_kill
1184264006 M * AStorm uhm, from command line :P
1184264013 M * daniel_hozac vkill
1184264039 M * AStorm and that'd be vkill --xid <thexid>, right?
1184264052 M * Bertl vkill --help :)
1184264057 M * AStorm hmm, I've got myself a dead context
1184264097 M * AStorm brb
1184264119 Q * AStorm Quit: Bye
1184264190 Q * ensc Remote host closed the connection
1184264372 M * daniel_hozac Bertl: what were the open issues you were talking about earlier? XFS sendfile, disk limit removal on superblock destruction (not very important, and i guess would require some testing), ... this COW stuff. is there more?
1184264437 M * Bertl IIRC, you mentioned a patch to fix something (but I do not remember what exactly)
1184264455 M * daniel_hozac http://people.linux-vserver.org/~dhozac/p/k/2.2.0.1/ are the patches i have lined up right now.
1184264457 M * Bertl ah, ext3 disk limits
1184264465 M * Bertl the off by one issue
1184264471 M * daniel_hozac right.
1184264512 M * daniel_hozac btw, there was one thing i was a bit unsure about when looking at the disk limit stuff...
1184264561 M * Bertl okay, I should be back in business tomorrow (almost), I decided to completely upgrade my infrastructure (at both locations) and that did take some time ...
1184264572 M * daniel_hozac in ext3_new_blocks, right before the return in the success branch, there's a DQUOT_FREE_BLOCK with no corresponding DLIMIT_FREE_BLOCK.
1184264593 M * Bertl looks suspicious
1184264607 M * Bertl in general, we always pair DQUOT and DLIMIT
1184264609 M * daniel_hozac (same thing applies to ext4)
1184264622 M * daniel_hozac yeah, that's what i was thinking.
1184264725 M * daniel_hozac i'll fix that too then.
1184264752 M * Bertl I think ext3/4 fixed some 'missing' quota allocations recently
1184264770 M * Bertl so that might be the reason why it is not handled by dlimits (yet)
1184264814 M * daniel_hozac yeah. all the others seemed to have a dlimit friend.
1184264879 M * daniel_hozac http://people.linux-vserver.org/~dhozac/p/k/2.2.0.1/delta-ext-dlimit-fix02.diff fixes all of the issues.
1184264913 M * Bertl ah, excellent, so that is replacing the fix01 then
1184264916 M * daniel_hozac right.
1184264931 M * Bertl (which got removed this moment :)
1184264936 M * daniel_hozac yep :)
1184264948 M * daniel_hozac still present in the upper level directory.
1184265007 M * Bertl how is life trating you, btw?
1184265012 M * Bertl *treating even
1184265057 M * daniel_hozac very well, thanks. how about you?
1184265075 M * Bertl yeah, can't complain so far ...
1184265168 Q * Pazzo Quit: Ex-Chat
1184265176 M * daniel_hozac so what infrastructure were you upgrading? network?
1184265178 J * HeinMueck ~Miranda@dslb-088-065-254-049.pools.arcor-ip.net
1184265212 M * Bertl daniel_hozac: network, servers (at home), disk/raid ...
1184265230 M * daniel_hozac ah, everything. nice.
1184265234 M * Bertl daniel_hozac: and finally software (i.e. moved all from mandrake 8.2 to mandriva 2007.1 :)
1184265242 M * daniel_hozac hehehe.
1184265249 M * daniel_hozac how old is 8.2 now?
1184265265 M * Bertl well, actually it had nothing in common with 8.2 anymore
1184265281 M * daniel_hozac oh, you had upgraded everything?
1184265284 M * Bertl I kept updating parts on demand to newer versions
1184265311 M * Bertl but there is only so much you can upgrade without running into troubles
1184265318 M * daniel_hozac yeah..
1184265340 M * Bertl so I decided to scratch the distro, and reinstall (only keeping improtant data)
1184265345 M * Bertl *important
1184265422 M * daniel_hozac sounds like a good idea.
1184265429 M * Bertl but it was about time now ... in that process I also replaced my k6/2 300 frontends :)
1184265456 M * daniel_hozac with what?
1184265471 M * Bertl (now running Athlon/XP 2GHz)
1184265494 M * Bertl those machines are way too fast for a frontend, btw :)
1184265499 M * daniel_hozac ah, nice upgrade.
1184265520 M * daniel_hozac hehe, definitely.
1184265539 M * Bertl server is an intel core2 duo at 64bit, which is really nice
1184265638 J * AStorm ~astralsto@host-81-190-179-124.gorzow.mm.pl
1184265746 M * daniel_hozac indeed, sounds good. i've only got that in my laptop for now, hope to upgrade the servers soon too...
1184265819 M * mjt hmm. how come - i built a "vserver" manually, without umounting anything, yet /proc/mounts in a vserver does not show "extra" directories mounted on host (such as /var etc) ?
1184265841 M * daniel_hozac it's chrooted, no?
1184265842 M * mjt "manually" = entering sequence of commands on the command line - vnamespace vcontext ...
1184265851 M * daniel_hozac mounts which are not accessible aren't showed.
1184265868 M * mjt but it's a good idea to umount them anyway, right?
1184265897 M * mjt (i was hoping to see what i have to umount - in a vserver's /proc/mounts ;)
1184265912 M * AStorm mjt, you won't be able to unmount because of lack of capabilities :>
1184265925 M * mjt i know
1184265937 M * mjt it has to be done before entering guest
1184266025 M * daniel_hozac unmounting them means you only have to unmount them on the host if you want to e.g. reformat it, extend it, or whatever.
1184266057 M * mjt here's my ad-hoc script: http://paste.linux-vserver.org/4502
1184266062 M * mjt ;)
1184266086 M * mjt dunno what some things mean still - like vuname --set -t context=..., and vattribute
1184266107 M * daniel_hozac so you run it with vnamespace --new ... ?
1184266115 M * mjt yes - around that script
1184266125 M * mjt vnamespace --new ./start.sh
1184266169 M * mjt and now i wonder how to kill the running vserver ;)
1184266179 M * mjt (it's not present in /etc/vservers)
1184266203 Q * rgl Ping timeout: 480 seconds
1184266224 M * daniel_hozac vcontext --migrate --xid .. --chroot -- /etc/vserver-stop?
1184266228 M * mjt aha. killing all processes inside it killed the whole vserver
1184266234 M * daniel_hozac yep.
1184266255 M * mjt i mean if there's some "force-kill" thing, like a power-off button
1184266264 Q * AStorm Quit: Bye
1184266268 M * mjt vpower-off ;)
1184266296 M * daniel_hozac vkill -9 --xid <xid> -- 0?
1184266301 M * daniel_hozac uh, -s 9
1184266370 M * mjt that works ;)
1184266408 M * mjt man vattribute
1184266411 M * mjt err miswin
1184266717 Q * gerrit Ping timeout: 480 seconds
1184267130 M * Bertl daniel_hozac: btw, regarding networking and userspace support
1184267158 J * ensc ~irc-ensc@p54B4D165.dip.t-dialin.net
1184267165 M * Bertl daniel_hozac: am I right that the interface ips are added according to the alphanumerical order the interface dirs have?
1184267231 M * daniel_hozac i think that depends on bash.
1184267236 M * daniel_hozac i.e. it's just a glob.
1184267250 M * daniel_hozac but the idea is that they should be.
1184267255 M * Bertl hmm, okay, that still should be truem I guess
1184267258 M * Bertl *true
1184267273 M * daniel_hozac why do you ask?
1184267280 M * Bertl okay, how complicated would it be to support extended matches there?
1184267295 M * Bertl e.g. networks or ip ranges?
1184267331 M * daniel_hozac networks are already supported, no?
1184267372 M * Bertl hmm, what I meant is, entire networks assigned to a guest
1184267372 M * daniel_hozac but it shouldn't be too hard.
1184267384 M * daniel_hozac yeah, but that would that entail more than ip and prefix?
1184267422 M * Bertl nothing actually, just an additional 'marker' when using the ekrnel interface
1184267432 M * daniel_hozac right.
1184267438 M * Bertl i.e. it would set a flag/mask, that's it
1184267465 M * Bertl so from the userspace PoV you think that would be simple to extend?
1184267486 M * Bertl just asking, because I think for the first shot, I'm going for an ordered list, but with extended matching
1184267503 M * Bertl i.e. ip, range and network (only inclusion, no exclusion)
1184267505 M * daniel_hozac well, kinda. i guess it'd be easiest to just start fresh with another directory tree in that case.
1184267520 Q * slack101 Ping timeout: 480 seconds
1184267541 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184267542 M * daniel_hozac or... hm. i guess it should be pretty easy.
1184267566 M * Bertl this would allow to test in actual setups, without doing any of the fancy parsing tree things
1184267586 M * Bertl and honestly, I think it would suffice for 99% of all cases
1184267607 M * daniel_hozac sure.
1184267845 Q * Blissex Remote host closed the connection
1184269487 J * gerrit ~gerrit@bi01p1.co.us.ibm.com
1184269601 M * Bertl okay, off for now .. have to pick up my SO from the airport :)
1184269606 N * Bertl Bertl_oO
1184271658 N * Bertl_oO Bertl
1184271664 M * Bertl flight got delayed :/
1184271728 M * daniel_hozac that's too bad.
1184271832 Q * slack101 Ping timeout: 480 seconds
1184271847 M * DavidS Bertl: that "just increases the suspense" ;)  ... the more it sucks now, the better it is when it's over
1184271853 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184272843 M * meandtheshell what's an SO?
1184273052 M * daniel_hozac significant other ;)
1184273071 M * meandtheshell lol
1184273118 M * meandtheshell awesome ... there are two Bertls out there :)
1184273324 M * Bertl meandtheshell: hmm?
1184273352 M * Bertl ah, got it :)
1184273363 M * meandtheshell ;)
1184273445 J * Aiken ~james@ppp121-45-220-241.lns2.bne1.internode.on.net
1184273462 M * coderanger Bertl: Ping
1184273814 J * AStorm ~astralsto@host-81-190-179-124.gorzow.mm.pl
1184274038 M * coderanger Bertl: Never mind, just needed to rtfm
1184274282 M * Bertl coderanger: np
1184274385 N * DoberMann DoberMann[ZZZzzz]
1184274599 Q * gerrit Ping timeout: 480 seconds
1184275265 J * gerrit ~gerrit@bi01p1.co.us.ibm.com
1184275812 Q * DavidS Quit: Leaving.
1184276024 P * stefani I'm Parting (the water)
1184276316 J * markus_ ~chatzilla@chello213047089232.17.14.vie.surfer.at
1184276414 M * daniel_hozac Bertl: btw, http://people.linux-vserver.org/~dhozac/p/m/delta-dlimit-feat02.diff is slightly updated. i'm trying to figure out why 216 is failing (disk limit is reporting 1 more block freed than df again...)
1184276505 M * markus_ How do I properly specify a runlevel other than 3? I was reading http://linux-vserver.org/util-vserver:Documentation#.2Fetc.2Fvservers.2Fvserver-name.2Fapps.2Finit and setting the numner '7' in the file apps/init/runlevel, but then I get this on the console: "Usage: fakerunlevel <runlevel> <utmp_file>\nPut a runlevel record in file <utmp_file>". Do I need some command in that file? There's...
1184276506 M * markus_ ...no documentation about it.
1184276578 Q * bonbons Quit: Leaving
1184276581 M * daniel_hozac it's erroring because 7 isn't between 0 and 6 (inclusive).
1184276608 M * markus_ hmm .. the reason is, I rsynced another machine which is using 7 ...
1184276644 M * daniel_hozac so you'll either have to patch out that sanity check from util-vserver, or make it use another runlevel.
1184276692 M * markus_ Why this check ... :-(
1184276768 Q * HeinMueck Quit: Aah!
1184276811 M * daniel_hozac probably because you normally use a runlevel in that range.
1184276816 M * coderanger Bertl: Another question, is there a good (read: not racey) way to deal with cleaning up after a container?
1184276828 Q * dna Quit: Verlassend
1184276839 M * daniel_hozac coderanger: why would it be racy, and what does cleaning up mean?
1184276853 M * markus_ daniel_hozac: thanks for that hint, I think I can change the runlevel number on the machine itself..
1184276878 M * Bertl coderanger: you can get a notification if you wait on context disposal
1184276910 M * Bertl coderanger: if you do not start a new one (with the same id) before your cleanup is finished, you should be fine
1184276924 M * coderanger Bertl: There is vc_wait_exit, but I would imagine we can't wait before it is created
1184276939 M * daniel_hozac you can use the state change helper as well.
1184276945 M * Bertl coderanger: correct
1184276952 M * coderanger daniel_hozac: Wazat?
1184276964 M * daniel_hozac which would call /sbin/vshelper once the context is destroyed.
1184276977 M * daniel_hozac (path is configurable, of course)
1184276985 M * coderanger daniel_hozac: Is this part of util-vserver or vserver itself?
1184276997 M * Bertl this is a kernel space helper
1184277009 M * Bertl note: you have to do any synchronization in userspace
1184277027 M * coderanger Bertl: Is this documented somewhere I can stare at?
1184277063 M * daniel_hozac kernel/vserver/helper.c? :)
1184277071 M * Bertl yep :)
1184277092 M * coderanger Heh, touche
1184277190 J * quasisane ~sanep@c-75-67-252-184.hsd1.nh.comcast.net
1184277198 M * daniel_hozac note that the helper is called a bit sooner than vc_wait_exit would return.
1184277232 M * daniel_hozac (i.e. before the context is unhashed, as opposed to after)
1184277248 M * coderanger This is for cleaning up the chroot and such, so split-second timing isn't needed
1184277261 M * Bertl who does the cleanups?
1184277268 M * daniel_hozac hmm, you probably want vc_wait_exit then.
1184277280 M * coderanger Rainbow will fork off another process to do it
1184277291 M * Bertl Rainbow is?
1184277296 M * markus_ Is util-vserver maintained anymore? last release somewhen last year ... ?
1184277297 M * daniel_hozac the mounts are still around when the helper is called.
1184277327 M * coderanger Bertl: The security serivce
1184277327 M * daniel_hozac markus_: eh? last release was in may.
1184277350 M * Bertl coderanger: then using the wait functionality is definitely the better way
1184277402 M * markus_ daniel_hozac: ops, I was looking in http://ftp.linux-vserver.org/pub/utils/util-vserver/ and https://savannah.nongnu.org/projects/util-vserver/, but seems old .. wrong place?
1184277417 M * markus_ ok, now I got it, my fault :)
1184277421 M * daniel_hozac markus_: that's where i'm looking. [   ] util-vserver-0.30.213.tar.bz2     03-May-2007 14:26  646K  
1184277446 M * Bertl http://ftp.linux-vserver.org/pub/utils/util-vserver/util-vserver-0.30.213.tar.bz2 ?
1184277449 M * markus_ yes, hard to spot. So, is there a place to report features ?
1184277455 M * daniel_hozac savannah.
1184277463 M * daniel_hozac or here.
1184277508 M * markus_ Here? Ok :) I'ld like to propose for supporting runlevels greater than 6 ...
1184277548 M * daniel_hozac ok, done.
1184277554 M * Bertl markus_: do you have a distro which uses them?
1184277563 M * markus_ sme server does
1184277573 M * Bertl daniel_hozac: you should do the cron-o-john part :)
1184277585 M * daniel_hozac what?
1184277598 M * daniel_hozac (http://svn.linux-vserver.org/projects/util-vserver/changeset/2562)
1184277610 M * Bertl daniel_hozac: you know, travel back in time and fix it ... then explain that it is implemented now :)
1184277615 M * daniel_hozac oooh, hehe.
1184277636 M * markus_ Bertl: it's a "all in one" mail/web server made easy, based on .. hmm ... redhat maybe
1184277663 M * Bertl sounds interesting, and what does it use the extra runlevels for?
1184277729 M * markus_ well, 7 is the main runlevel
1184277741 M * markus_ per inittab it straight starts 7 as default
1184277762 M * daniel_hozac why not 2? or 3? or 4? or 5?
1184277805 Q * meandtheshell Quit: Leaving.
1184277814 M * markus_ uh, I don't know? I didn't wrote it.
1184278268 M * Bertl okay, off to the airport (again) .. back later
1184278274 N * Bertl Bertl_oO
1184278278 M * daniel_hozac good luck!
1184278722 Q * markus_ Quit: ChatZilla 0.9.78.1 [Firefox 2.0.0.4/2007051502]
1184279494 Q * hardwire Ping timeout: 480 seconds
1184279933 Q * lilalinux__ Remote host closed the connection
1184280399 Q * slack101 Ping timeout: 480 seconds
1184280420 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184281909 Q * gerrit Ping timeout: 480 seconds
1184281910 M * Supaplex slack101: sup slacker
1184282341 M * slack101 Supaplex, what it do pimpin 
1184282547 J * gerrit ~gerrit@bi01p1.co.us.ibm.com
1184283397 Q * gerrit Quit: Client exiting
1184284688 N * Bertl_oO Bertl
1184284694 Q * slack101 Ping timeout: 480 seconds
1184284697 M * Bertl back now ... 
1184284715 J * slack101 ~default@cpe-65-31-15-111.insight.res.rr.com
1184284717 M * Bertl (this time successful :)