1165191320 N * Bertl_oO Bertl
1165191323 M * Bertl back now ...
1165191622 Q * yarihm Quit: Leaving
1165191870 Q * Piet Quit: Piet
1165194862 J * phisi-turkey ~sebebiyat@85.101.143.76
1165194865 M * phisi-turkey  SeLamunALeykum
1165194957 Q * phisi-turkey Read error: Connection reset by peer
1165195505 J * ntrs_ ~ntrs@68-188-55-120.dhcp.stls.mo.charter.com
1165195856 Q * ntrs Ping timeout: 480 seconds
1165198236 M * ntrs_ Bertl, did you read about the netstat problem?
1165198271 M * Bertl didn't read up yet, care to give me a quick overview?
1165198286 M * ntrs_ netstat -anlp inside a guest shows all connections on all IPs on the hosts, even ones not belonging to that context.
1165198294 M * ntrs_ daniel_hozac was able to reproduce it.
1165198313 M * Bertl on which kernel version?
1165199322 M * Bertl ntrs_: 2.6.18.x with 2.1.1.2.3?
1165199332 M * ntrs_ <daniel_hozac> (on 2.6.18.3-vs2.1.1.2.3.ipv6
1165199341 M * ntrs_ and me with 2.6.18.3 with 2.1.1.3
1165199351 M * Bertl okay, tx
1165200271 J * FireEgl ~FireEgl@adsl-17-159-214.bhm.bellsouth.net
1165200289 M * Bertl wb FireEgl!
1165200308 M * FireEgl tanks =)
1165200371 M * ntrs_ Bertl, I hope this did not open guests to security breaches.
1165200396 M * Bertl not really, it is probably just some hiding mechanism which got lost
1165200410 M * Bertl checking now, we'll know in a few minutes, I thinl
1165200428 M * Bertl *think
1165200899 J * mountie ~mountie@CPE0080c6fe323f-CM000a739acaa4.cpe.net.cable.rogers.com
1165200963 M * Bertl wb mountie!
1165201601 M * Bertl ntrs_: that is only for established connections, not listening, right?
1165201643 M * ntrs_ correct
1165201674 M * ntrs_ actually it's for all except listening.
1165201731 M * Bertl so you see wait closing and such too?
1165202096 Q * RichyF Quit: ( www.nnscript.de :: NoNameScript 3.72 :: www.XLhost.de )
1165202284 M * Bertl ntrs_, daniel_hozac: hmm, yeah, I guess I know what happens here, we probably want to change the checks from vx_check() to nx_check() as we are now running in softirq context (or diversify the in_irq checks)
1165202322 M * Bertl ntrs_: can you test a patch/version somewhere?
1165202345 M * ntrs_ I can't until tomorrow, but feel free to use and reboot the test server if you want.
1165202360 M * ntrs_ Bertl, beryllium I mean.
1165202377 M * Bertl okay, that would suffice, IMHO, can we put a guest there with sockets to show/list?
1165202413 M * ntrs_ You can do anything you want. I have to go now and won't be back until tomorrow.
1165202433 M * Bertl okay, will probably prepare something and then head to bed ...
1165202439 M * ntrs_ ok
1165202455 M * Bertl thanks for spotting!
1165204394 M * ntrs_ Bertl, before I go, is there a patch to fix this problem. I could have someone here install it on a production server currently running 2.1.1.3 and try it out.
1165204604 M * ntrs_ hello Bertl?
1165205098 M * Bertl yep
1165205140 M * Bertl I#m preparing a patch atm .. should be ready in 20min with testing
1165206026 M * Bertl ntrs_: that one is _supposed_ to fix the issue, but it is not run tested yet: http://vserver.13thfloor.at/Experimental/delta-sock-fix01.diff
1165208082 J * Rusty ~rusty@a1052.adsl.pool.eol.hu
1165208087 M * Rusty hello guys
1165208088 M * Bertl welcome Rusty!
1165208091 M * Rusty i have a problem :)
1165208100 M * Bertl let's hear ...
1165208102 M * Rusty apollo:/usr/local/etc/vservers/bnc# vserver bnc start
1165208102 M * Rusty Inconsistency detected by ld.so: rtld.c: 1192: dl_main: Assertion `(void *) ph->p_vaddr == _rtld_local._dl_sysinfo_dso' failed!
1165208133 M * Bertl sounds interesting ... please share some details, like kernel version/patches/tools
1165208208 M * Rusty  2.6.18.2-grsec2.1.9-vs2.1.1
1165208231 M * Rusty testme.sh is succeed :)
1165208244 M * Bertl that's a good start, tools would be? and arch?
1165208257 M * Rusty c
1165208278 M * Rusty apollo:~# vserver --version
1165208278 M * Rusty vserver 0.30.211 -- manages the state of vservers
1165208289 M * Rusty this is a debian stable operating system
1165208303 M * Bertl okay, arch is x86? or x86_64?
1165208307 M * Rusty x86
1165208320 M * Bertl the guest is debian too?
1165208334 M * Rusty yes
1165208345 M * Bertl did you compile the tools yourself?
1165208349 M * Rusty but sebian testing
1165208352 M * Rusty yes
1165208378 M * Bertl okay, could you upload the output of 'vserver-info - SYSINFO' to paste.linux-vserver.org please?
1165208441 M * Rusty http://paste.linux-vserver.org/727
1165208445 M * Bertl tx
1165208463 M * Bertl Use dietlibc: no (you have been warned)
1165208470 M * Bertl that is most likely your problem
1165208483 M * Bertl get a recent (and decent :) dietlibc and recompile the tools
1165208511 M * Bertl best get the dietlibc version from testing, as it fixes some issues
1165208538 M * Rusty hmm, thx :)
1165208549 M * Bertl np, let me know how it goest ...
1165208553 M * Bertl -t
1165208693 M * Rusty checking whether to enable dietlibc... no (detected)
1165208694 M * Rusty configure: WARNING: ***                                                         ***
1165208694 M * Rusty configure: WARNING: *** it is strongly recommended to link util-vserver against ***
1165208694 M * Rusty configure: WARNING: *** dietlibc;  glibc's NSS functions are very unreliable in ***
1165208694 M * Rusty configure: WARNING: *** chroot() environments,  so you have to  expect problems ***
1165208695 M * Rusty configure: WARNING: *** there.                                                  ***
1165208695 M * Rusty configure: WARNING: ***                                                         ***
1165208710 M * Rusty but i installed the dietlibc deb :)
1165208724 M * Bertl still missing the dietlibc, it's the one labeled dietlibc-dev IIRC
1165209005 Q * FireEgl Remote host closed the connection
1165209171 M * Rusty oh it works!! thx :)
1165209193 M * Bertl excellent!
1165209211 M * Bertl feel free to hang around if you like ... otherwise, have a good day :)
1165209211 M * Rusty u'r the god :)
1165209245 M * Bertl nah, I'm just a little comfortable with the stuff :)
1165209650 T * Bertl http://linux-vserver.org/ | latest stable 2.02.1, exp 2.02.2-rc8, devel 2.1.1.3, 2.2.0-pre5, 2.1.x-t4(2.6.19), stable+grsec 2.0.2.1, devel+grsec 2.1.1 | util-vserver-0.30.211 | libvserver-1.0.2 & vserver-utils-1.0.3 | He who asks a question is a fool for a minute; he who doesn't ask is a fool for a lifetime -- share the gained knowledge on the Wiki, and we'll forget about the minute ;)
1165209676 M * Bertl *2.1.1.3, 2.2.0-pre5 and 2.1.x-t4(2.6.19)
1165209781 M * Bertl Rusty: if you don't mind the question, what do you use Linux-VServer for?
1165209955 M * Bertl daniel_hozac: just for the record, the ulimit-exec issue goes away with the cvs version of dietlibc
1165210684 M * Rusty Bertl: i use vserver for separate services on my server's... throught security :)
1165210698 M * Bertl ah, excellent choice!
1165210758 M * Rusty yeah, i think, vserver+grsec is a hardened tool :)
1165214041 M * Bertl daniel_hozac: sarge64 guest (amd64) will get the wrong sources.list entry, is that a build or a debian (debootstrap) bug?
1165214102 M * Bertl daniel_hozac, micah: in general, it would be cool if the 'install source' would get entered there
1165214157 M * Hollow morning
1165214174 M * Bertl good morning Hollow!
1165214204 M * Eyck- morning Bertl, Hollow
1165214225 M * Bertl hey Eyck-! new nick?
1165214233 M * Hollow Bertl: did you take a look at the scheduler page? i'm not sure all the formulas are correct
1165214243 M * Eyck- nice, isn't it?
1165214259 M * Bertl Hollow: not yet, but I can look at it now ... sec
1165214379 M * Bertl Hollow: is all that text from you?
1165214419 M * Hollow thew first paragraph about processes is copied from a gnu fdl page.. have to link to the source probably
1165214431 M * Hollow the rest is my own
1165214463 M * Bertl the linux scheduler, strictly speaking, does not run in a thread, it is a small piece of code invoked when either a process gives up the cpu (when rescheduling) or when the timer irq decides (with some even smaller piece of code) that the task has to give up the cpu :)
1165214487 M * Hollow it was copied from http://josh.trancesoftware.com/linux/linux_cpu_scheduler.pdf
1165214492 M * Bertl except for that, it is very nice
1165214539 M * Hollow are the examples correct? i'm not sure if the S and M are correct, and i'm also not sure about the fair share example ..
1165214674 M * Bertl making T small has some advantages but also some disadvantages
1165214700 M * Bertl i.e. that probably needs a deeper discussion, please remind me to add a few words there
1165214713 M * Hollow ok
1165214731 M * Hollow i have that info from the old wiki page ..
1165214734 M * Bertl basically the problem is with oscillation with too little tokens
1165214769 M * Bertl e.g. if a minfill of 10 is set, the tokens are added quickly, you get small intervalls of 10 ticks for each context
1165214789 M * Bertl which actually makes it _very_ smooth, but also has a lot of scheduling overhead
1165214817 M * Bertl raising the min fill can help here as well as increasing the T with constant R/T ratio
1165214900 M * Hollow so you basically need to find the golden mean between overhead and "smotthness"
1165214907 M * Bertl the formular for the burst time is wrong :)
1165214916 M * Hollow yeah, i thought so... ;)
1165214930 M * Bertl basically you have a second level term which goes with the time
1165214941 M * Hollow i took it from vschedcalc and thought, uhm no.. that must be wrong
1165214945 M * Bertl (as you add tokens while you burst too)
1165214979 M * Bertl if you remind me next week, I'll do the full math and write a few words how it can be calculated
1165215016 M * Hollow ok, that would be great... a scheduler math reference is really needed ;)
1165215060 M * Bertl the fair share part looks interesting .... :)
1165215079 M * Hollow heh, i looked up that formula in one of our scheduler irc talks
1165215083 M * Hollow so it should be right ;)
1165215120 M * Bertl okay, I take it the wiki can do the formulars now (as I actually see them :)
1165215131 M * Hollow yep, i enabled that a few days ago
1165215148 M * Bertl so there should be no problem adding more math ...
1165215372 M * Bertl daniel_hozac, ntrs_: looks good, recent patch versions fix the socket visibility
1165215533 M * Hollow off now to university, cu later!
1165215600 J * Aiken_ ~james@tooax8-039.dialup.optusnet.com.au
1165215643 M * Bertl Hollow: cya! have fun!
1165215649 M * Bertl morning Aiken_!
1165215732 Q * ricola Ping timeout: 480 seconds
1165215926 Q * Aiken Ping timeout: 480 seconds
1165216275 M * Aiken_ hi Bertl
1165216305 M * Aiken_ killed my guest setup this morning, alpha + 2.6.19 + aoe does not work :(
1165216415 M * Bertl hmm, so some breakage there?
1165216424 M * Aiken_ yes
1165216428 M * Aiken_ and it is mainline
1165216436 M * Bertl thought so ... 
1165216449 M * Bertl have you reported it upstream?
1165216462 M * Aiken_ not good when a filesystem ends up a handfull of numbered files in lost&found
1165216464 M * Aiken_ not yet
1165216482 M * Bertl ah, so it led to fs corruption?
1165216500 M * Bertl i.e. it looked like it was working, but wasn't 
1165216560 M * Aiken_ led the massive filesystem corruption, trying to use a guest, untarring a file, even mkfs.ext{2,3} fails
1165216591 M * Bertl nice
1165216624 M * Aiken_ was very pleased when I found a usable backup of that filesystem
1165216636 M * Bertl I can imagine
1165216765 M * Aiken_ was going to try it on x86 and a couple of the 19-rcX before I report to what/when it broke
1165217144 M * Bertl yep, can't hurt ...
1165217659 Q * m4z Remote host closed the connection
1165217670 J * sFLc^CR[e ~hollow@styx.xnull.de
1165217684 Q * Hollow Read error: No route to host
1165217693 J * m4z m4z@bastard-operator.from-hell.net
1165217729 N * sFLc^CR[e Hollow
1165219169 M * Bertl off to bed now ... have a good one everyone! cya!
1165219176 N * Bertl Bertl_zZ
1165219863 J * dreamind ~dreamind@C2107.campino.wh.tu-darmstadt.de
1165219891 M * dreamind Hi folks :D
1165220356 J * Torsti76 tkurbad@gate.iwm-kmrc.de
1165221414 Q * h01ger Quit: h01ger
1165221440 J * h01ger ~holger@socket.layer-acht.org
1165221939 J * praeR ~Benjamin@host.187.57.23.62.rev.coltfrance.com
1165222103 Q * dreamind Quit: dreamind
1165223879 J * Piet hiddenserv@tor.noreply.org
1165224773 J * dna ~naucki@110-216-dsl.kielnet.net
1165225064 J * renihs ~penguin@83-65-34-34.arsenal.xdsl-line.inode.at
1165225201 M * renihs hmm i seem to experience random crashes on a collegues vserver (services within guests)
1165225238 M * renihs syslog shows loads of stuff like agetty[9204]: /dev/tty4: No such file or directory
1165225240 M * renihs init: Id "c1" respawning too fast: disabled for 5 minutes
1165225315 M * DavidS renihs: guests don't have access to the vts by default 0> no ttyX -> disable agetty in inittab
1165225322 M * renihs DavidS, mkay
1165225328 M * DavidS but that most likely has nothing to do with your crashes
1165225341 M * renihs that was what i wanted todo, i am just trying to figure out why they are not disabled
1165225348 M * renihs baseloayout change maybe?
1165225366 M * renihs DavidS, ya, but it looks like they are related somehow, at least for ftp/apache 
1165225369 M * renihs dunno
1165225405 M * renihs however i noticed a few days ago a amok running vcontext process eating up all cpu
1165225408 M * renihs also :)
1165225447 M * DavidS dunno
1165225515 J * Milf ~Miranda@146.140.252.55
1165225525 M * Milf Moin
1165225556 M * renihs morning Milf 
1165226502 J * lilalinux ~plasma@dslb-084-058-200-151.pools.arcor-ip.net
1165226521 Q * Piet Ping timeout: 480 seconds
1165226625 J * Piet hiddenserv@tor.noreply.org
1165227600 P * Serge Leaving
1165227733 Q * DavidS Quit: Leaving.
1165228061 Q * lilalinux Ping timeout: 480 seconds
1165228120 J * lilalinux ~plasma@dslb-084-058-200-151.pools.arcor-ip.net
1165228545 Q * michal` Quit: kernel update
1165229113 J * ntrs__ ~ntrs@68-188-55-120.dhcp.stls.mo.charter.com
1165229132 J * dna_ ~naucki@110-216-dsl.kielnet.net
1165229147 Q * ntrs_ Read error: Connection reset by peer
1165229380 Q * ms_ Quit: [BX] Yo quiero BitchX
1165229515 Q * dna Ping timeout: 480 seconds
1165230192 Q * lilalinux Ping timeout: 480 seconds
1165230302 J * lilalinux ~plasma@dslb-084-058-200-151.pools.arcor-ip.net
1165230469 J * Johnsie ~jdlewis@jdlewis.org
1165230541 Q * Johnnie Ping timeout: 480 seconds
1165231541 Q * Piet Ping timeout: 480 seconds
1165231629 J * Piet hiddenserv@tor.noreply.org
1165232072 M * brcc w
1165232286 M * renihs v
1165232581 Q * Aiken_ Ping timeout: 480 seconds
1165233634 Q * ensc Killed (NickServ (GHOST command used by ensc_))
1165233643 J * ensc ~irc-ensc@p54B4FCEE.dip.t-dialin.net
1165234662 J * michal ~michal@www.rsbac.org
1165234665 N * michal Guest263
1165234680 N * Guest263 michal`
1165235065 J * dreamind ~dreamind@p54A7A35A.dip0.t-ipconnect.de
1165235073 M * dreamind Hi folks :)
1165235113 M * renihs zod
1165235130 J * rgl ~Rui@84.90.8.72
1165235133 M * rgl hello
1165235144 M * renihs narf
1165236846 Q * praeR Ping timeout: 480 seconds
1165239048 Q * virtuoso Ping timeout: 480 seconds
1165239178 Q * Johnsie Ping timeout: 480 seconds
1165239189 J * virtuoso ~s0t0na@shisha.spb.ru
1165239477 J * Johnsie ~jdlewis@jdlewis.org
1165239882 Q * Eyck- Remote host closed the connection
1165240618 Q * Johnsie Read error: Connection reset by peer
1165240671 J * Johnsie ~jdlewis@jdlewis.org
1165240770 J * eyck ~eyck@nat-old.nowanet.pl
1165241071 J * pmenier10et ~pmenier@ACaen-152-1-14-177.w83-115.abo.wanadoo.fr
1165241105 Q * pmenier10et 
1165241136 Q * Johnsie Read error: Connection reset by peer
1165241185 J * Johnsie ~jdlewis@jdlewis.org
1165242356 Q * Johnsie Ping timeout: 480 seconds
1165242529 J * Johnsie ~jdlewis@jdlewis.org
1165245025 M * daniel_hozac Bertl_zZ: 2.1.x-t4 works a lot better, thanks!
1165245183 Q * Johnsie Ping timeout: 480 seconds
1165245227 Q * Torsti76 Quit: Download Gaim: http://gaim.sourceforge.net/
1165245228 J * Johnsie ~jdlewis@jdlewis.org
1165245996 Q * Johnsie Quit: G'bye!
1165246296 J * sebastian ~info@p54A95E72.dip.t-dialin.net
1165246494 Q * borgfish 
1165248826 Q * Milf Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org
1165249477 J * crypt ice@p54A3BA0B.dip0.t-ipconnect.de
1165249478 Q * Vudumen Read error: Connection reset by peer
1165249487 M * crypt Hi all
1165249504 M * daniel_hozac hello
1165249523 J * Vudumen 6fd8f3dfd0@perverz.hu
1165249743 M * crypt daniel_hozac, I have a short question: is there a protocol change in uv 0.211 concerning as rlimit?
1165249756 M * daniel_hozac hmm?
1165249853 M * crypt i have file as and rss in the rlimit dir. rss with 256000 and as with 1024000 content, but the vserver only have 102400 Ram in top
1165249969 M * crypt kernel is  2.6.18.1-vs2.1.1-rc47-amd64-squash-drbd-256ip-ipv6
1165250134 M * doener isn't the displayed memory controlled by rss only? hard for mem, soft for swap?
1165250140 Q * dreamind Quit: dreamind
1165250141 M * daniel_hozac yep.
1165250197 M * daniel_hozac or well, other way around, no?
1165250207 M * daniel_hozac soft for mem, hard - soft for swap.
1165250215 M * doener uhm... yep :)
1165250225 J * stefani ~stefani@tsipoor.banerian.org
1165250229 M * crypt hmm but actually i should get the rss value in mem or?
1165250236 M * daniel_hozac which you do.
1165250250 M * crypt i get the as value as mem and the rss value isn't shown anywhere
1165250265 M * daniel_hozac no, 1024000 pages is 4 GiB.
1165250275 M * daniel_hozac roughly.
1165250294 M * daniel_hozac 256000 pages is roughly 1 GiB.
1165250306 M * crypt hmm ok for a short understanding: i want to give the guest 256 MB ram and 1024 MB Swap so which values do i have to set?
1165250332 M * daniel_hozac well, you know you can't _actually_ assign swap, right?
1165250357 M * crypt yes
1165250372 M * crypt but on oterh installation i got the as value as Swap shown in guests
1165250390 M * daniel_hozac but to get that appearance, rss.hard should be 327680, rss.soft 65536.
1165250400 M * daniel_hozac on a stable kernel, right?
1165250428 M * crypt now a 2.6.17 kernel of derjohn
1165250438 M * daniel_hozac 2.0 doesn't support soft limits for, well, anything.
1165250439 M * crypt on the working installation
1165250474 M * crypt so i have to rename my files to rss.hard and as.hard ?
1165250493 M * daniel_hozac no.
1165250508 M * daniel_hozac 17:39 < daniel_hozac> but to get that appearance, rss.hard should be 327680, rss.soft 65536.
1165250522 M * daniel_hozac btw, is there any particular reason you have such an old kernel from derjohn?
1165250542 M * daniel_hozac he has 2.1.1.2 available...
1165250562 M * crypt yes ;) I can't reboot the testsystem atm because i have running several tests
1165250592 M * crypt ok so generally i shoudn't use rss only instead i should use rss.hard and rss.soft
1165250608 M * daniel_hozac rss only sets both the hard and soft limits.
1165250623 M * daniel_hozac so you end up with no visible swap.
1165250662 M * crypt hmm i thought rss stands for virtual mem and as for virtual swap
1165250692 Q * Vudumen Ping timeout: 480 seconds
1165250711 M * daniel_hozac for 2.0, yes.
1165250718 M * daniel_hozac not so for 2.1.
1165250735 M * crypt ahh there is the change :)
1165250771 M * crypt but for what is standing as in 2.1?
1165250786 M * daniel_hozac as is limiting address space, just like it does in 2.0.
1165250801 M * daniel_hozac it's just not a factor for VXF_VIRT_MEM.
1165250837 J * comfrey ~comfrey@201.243.176.219
1165250838 M * daniel_hozac (you can have tons of AS, but practically no memory or swap allocated, IIUC)
1165250845 M * crypt hmm, so I never will get a number in swap inside a guest
1165250847 Q * Rusty 
1165250860 N * Bertl_zZ Bertl
1165250862 M * daniel_hozac not with your current configuration, no.
1165250865 M * daniel_hozac morning Bertl!
1165250866 M * Bertl morning folks!
1165250877 M * crypt hi Bertl 
1165250881 M * Bertl crypt: update to 2.1.1.3 and you will get nice numbers
1165250898 M * daniel_hozac not if you configure rss and as though ;)
1165250917 M * crypt ok that sounds pretty nice, so which config files do i have to set to get these nice numbers?
1165250931 M * daniel_hozac 17:41 < daniel_hozac> 17:39 < daniel_hozac> but to get that appearance, rss.hard should be 327680, rss.soft 65536.
1165250935 M * daniel_hozac :)
1165250939 M * crypt ah ok
1165250961 M * crypt and for as the same scheme?
1165251018 M * daniel_hozac as isn't used at all for the VXF_VIRT_MEM calculations, as i said.
1165251049 M * crypt so i could use my as file as i used before
1165251072 M * daniel_hozac sure, you just won't see it from inside the guest until you hit it.
1165251086 M * crypt ah ok
1165251097 M * crypt thank's for your infos!
1165251704 M * Bertl daniel_hozac: reg -t4, any issues you encountered with that?
1165251920 J * Vudumen c91f40942f@perverz.hu
1165252099 J * bonbons ~bonbons@83.222.39.117
1165252109 M * Bertl wb Vudumen! hey bonbons!
1165252125 M * bonbons hey Bertl!
1165252132 M * crypt hum: vsched: non-numeric value specified for '--priority_bias' I didn't get this error before some hints
1165252140 M * crypt ?
1165252218 M * daniel_hozac Bertl: not yet.
1165252226 M * daniel_hozac crypt: yes, i fixed the parsing.
1165252231 Q * lilalinux Ping timeout: 480 seconds
1165252241 M * daniel_hozac you have to fix your dummy lines to 0, or upgrade to the new scheduler configuration.
1165252274 M * daniel_hozac see http://people.linux-vserver.org/~dhozac/p/uv/experimental/configuration.html /etc/vservers/vserver-name/sched
1165252279 M * crypt thx
1165252346 M * crypt and when i doesn't using this dir? Should i create all the files and setting them to 0 ?
1165252376 M * daniel_hozac if you continue to use the schedule file (it's still supported for now), you'll just have to fix the dummy line to say 0 or whatever priority bias you want.
1165252396 M * daniel_hozac note that if the sched directory exists, the schedule file won't be used at all.
1165252422 M * crypt ah k thx so in future releases the sched dir is used to set scheduling params?
1165252431 M * daniel_hozac preferably, yes.
1165252442 M * daniel_hozac as it supports all of the 2.1 scheduler options.
1165252479 M * daniel_hozac it is only available in 0.30.212+ though.
1165252590 M * crypt hmm i'm just upgraded to debian etch's util-vserver 211-1 and there the error occurs
1165252655 M * daniel_hozac hmm, you sure it's -1?
1165252665 M * daniel_hozac -3 or -4 upgraded to 0.30.212-rc1.
1165252719 M * crypt ah sry it's -4
1165252727 M * crypt i missed that
1165252735 J * Piet_ hiddenserv@tor.noreply.org
1165252771 J * lilalinux ~plasma@dslb-084-058-200-151.pools.arcor-ip.net
1165252815 M * Bertl daniel_hozac: while adding the 'hack' for chcontext
1165252832 Q * Piet Remote host closed the connection
1165252840 M * Bertl I found that 212-rc2 doesn't use the recent interfaces for context creation
1165252844 M * daniel_hozac right.
1165252868 M * Bertl that is why I had to disable the EPERM in set_namespace :)
1165252887 M * daniel_hozac hmm?
1165252900 M * Bertl I don't see a problem with that, but you should consider/update that
1165252956 M * Bertl i.e. the namespace is currently set twice, once as legacy support on context creation, and a second time as separate syscall
1165252992 M * daniel_hozac right, i saw that in the interdiff. but what EPERM are you talking about?
1165252993 M * Bertl maybe the VCI isn't set properly for the tools? could be, no idea ...
1165253000 M * daniel_hozac i.e. did i miss it?
1165253007 M * daniel_hozac no, _v1 isn't used at all currently.
1165253011 M * Bertl we had a check in set_namespace
1165253027 M * Bertl which bails out when a namespace is already set
1165253043 Q * transacid Quit: Lost terminal
1165253046 M * Bertl with that enabled, the tools were not able to start a guest
1165253048 M * daniel_hozac ah, i see it now.
1165253069 Q * Vudumen Remote host closed the connection
1165253079 M * Bertl you might want to add that back when you test the new interfaces
1165253104 M * Bertl (we can probably make it a config time option too, although I'm not sure it is worth the trouble)
1165253135 M * daniel_hozac hmm, how long has VCMD_ctx_create_v1 been around? seems to be all the way since 2.0, at least.
1165253160 M * daniel_hozac i was somehow under the mistaken impression that it was a recent addition.
1165253185 M * crypt daniel_hozac, i just set rss.hard and rss.soft and now i get a cannot allocate mem when trying to exec top in the guest, is this a kernel bug?
1165253202 M * daniel_hozac crypt: what did you set them to, and how much is currently used?
1165253217 M * daniel_hozac crypt: i.e. what does cat /proc/virtual/<xid>/limit show?
1165253235 J * besonen__ ~besonen@dsl-db.pacinfo.com
1165253235 J * ntrs_ ~ntrs@68-188-55-120.dhcp.stls.mo.charter.com
1165253314 J * thunder11 ~thu@tor-irc.dnsbl.oftc.net
1165253317 Q * thunder1 Read error: Connection reset by peer
1165253317 Q * ntrs__ Read error: Connection reset by peer
1165253317 Q * besonen_ Read error: Connection reset by peer
1165253326 M * daniel_hozac ah right, it was VCMD_ctx_migrate that i changed to v0... i suppose ctx_create_v1 shouldn't be a problem, other than me having to break API :)
1165253329 M * crypt http://paste.linux-vserver.org/729
1165253381 N * thunder11 thunder1
1165253385 M * daniel_hozac crypt: well, you're hitting the AS limit.
1165253401 J * transacid ~transacid@transacid.de
1165253407 M * crypt hmm i set as to 1024000
1165253422 M * daniel_hozac it's set to 262144 there :)
1165253431 M * daniel_hozac but either way, that's not a whole lot.
1165253465 M * daniel_hozac at least not if you want it to be able to use that much RAM+"swap".
1165253474 M * crypt what would be a better value?
1165253532 M * daniel_hozac well, run the guest unlimited for a while, do the things it's supposed to do, take the maximum from that file and add some for safety, and use that?
1165253591 J * Vudumen 53e70b1a95@perverz.hu
1165253597 M * crypt so rss limit's can i use but have to set the the as limit to a value i tested out
1165253616 M * daniel_hozac well, ideally you'd test for all the limits.
1165253665 M * crypt ok right thank's for the hints again
1165253691 M * crypt so i know which thing we have to put into openvcp for the next release
1165253710 M * daniel_hozac hehe.
1165253870 M * Bertl daniel_hozac: do we need some VCI tagging for the tools? to mark *spaces or something?
1165253946 M * daniel_hozac well, if the interfaces are changing from what we have now (i.e. clone(CLONE_*)+vc_set_namespace), that'd be very appreciated.
1165253947 M * Bertl when we actually decide to burry 2.0 with 2.6.18, we need to get 2.2 out soon
1165253973 Q * Vudumen Remote host closed the connection
1165253976 M * Bertl daniel_hozac: in what way, please elaborate
1165254008 M * daniel_hozac i mean, if there will be more things required, i'd love to have a way for the kernel to tell me so.
1165254014 M * daniel_hozac but as is, i don't think it's really required.
1165254021 M * daniel_hozac though i guess it would make sense.
1165254039 M * daniel_hozac (especially if we want to go in the vspace direction)
1165254042 M * Bertl I'm expecting a few 'new' *spaces in the near future
1165254058 M * daniel_hozac oh, which ones?
1165254067 M * daniel_hozac pid i guess is already in progress, i've seen traces of it 2.6.19.
1165254068 M * Bertl so the tools should be able to 'select' them properly in the set/clone
1165254075 N * dna_ dna
1165254093 M * Bertl yes, pid will be there soon, and probably will require special workarounds for spectator
1165254098 M * daniel_hozac yeah.
1165254111 M * Bertl (or maybe a completely different approach, we'll see)
1165254124 M * Bertl I guess socket namespaces will be next
1165254158 M * Bertl I moved most of the socket stuff into network contexts
1165254172 M * daniel_hozac yeah, i saw that, makes sense.
1165254173 M * Bertl (except for accounting, for now)
1165254199 M * daniel_hozac i guess that might break for people using ancient utils though.
1165254204 M * Bertl so that will require special 'unsharing' on network context creation
1165254213 M * daniel_hozac (i.e. ones that use vc_set_ipv4root)
1165254235 M * Bertl I don't have a problem to drop those on 2.2 too
1165254260 M * daniel_hozac i guess having the spaces available from vc_get_vci would make sense.
1165254281 M * daniel_hozac or maybe a special vc_get_spaces?
1165254282 J * Vudumen ~vudumen@perverz.hu
1165254287 M * Bertl IMHO it's much better to get something like 'not supported' instead of acting like it works, and silently fail (as early 2.6.19 patches did
1165254304 M * daniel_hozac yeah, i agree.
1165254325 M * daniel_hozac aren't vc_set_ipv4root and vc_new_s_context removed from 2.2.0 already?
1165254326 M * Bertl we can do something like that, but I guess vci should have enough free bits, let me check
1165254345 M * daniel_hozac well, i was thinking more like the mapping would be easier.
1165254351 M * Bertl they should be, but they are still in the 2.6.19 branch
1165254356 M * daniel_hozac we could just return the CLONE_* flags directly.
1165254361 M * daniel_hozac okay.
1165254374 M * daniel_hozac (if there was a separate vc_get_spaces)
1165254389 M * Bertl if you prefer that, I have no problem with it
1165254437 M * daniel_hozac well, doesn't really matter to me.
1165254484 M * Bertl why not, probably easier to map, just a mask for clone/unshare
1165254807 J * bronson ~bronson@adsl-75-36-144-172.dsl.pltn13.sbcglobal.net
1165254826 J * borgfish ~bla@141.12.9.118
1165254838 J * derjohn ~derjohn@dslb-084-058-200-151.pools.arcor-ip.net
1165254902 M * crypt is there actually the possibility to get virtual 127.0.0.1 lo interfaces inside a guest? Because as i know you can simply add one guest this interface.
1165254957 M * daniel_hozac crypt: there have been several implementations of a per-guest isolated 127.0.0.1.
1165254985 M * crypt so i can use that in the newest dev versions?
1165254992 M * Bertl crypt: if you feel like testing that, you can get a patch from me
1165255238 Q * Vudumen Read error: Connection reset by peer
1165255260 M * crypt Bertl, i think i have to test quota more before could testing the next thing ;)
1165255361 M * Bertl well, that'd be nice, before we remove it :)
1165255376 M * Bertl daniel_hozac: okay, what about this:
1165255389 M * Bertl - we add a single VCI bit for *space support
1165255430 M * Bertl - at the same time, we add the new vc_namespace_mask() command
1165255437 J * DavidS ~david@217.19.46.10
1165255467 M * Bertl tools can/will check for the VCI bit and/or query the mask, sounds reasonable?
1165255473 M * crypt Bertl, please do not remove that yet ;) atm i have much trouble but after christmas i'll make the next tests
1165255481 M * daniel_hozac vc_set_namespace_mask?
1165255494 M * Bertl hmm, no set there :)
1165255495 M * daniel_hozac or, oh, i get it now.
1165255500 M * daniel_hozac yeah.
1165255519 M * daniel_hozac but it would still require a single vc_set_namespace call?
1165255534 M * daniel_hozac i.e. the mask vx_set_namespace has isn't exposed to userspace?
1165255645 M * Bertl the set namespace interface can be changed to allow replacing specific spaces if that helps
1165255666 M * Bertl actually I didn't see a point in doing so, only for the enter, where it is already implemented
1165255699 M * daniel_hozac hmm, how else would e.g. chhbind --ip 127.0.0.1 true work?
1165255722 M * daniel_hozac (if a socket namespace is introduced)
1165255789 M * daniel_hozac although that's problematic on another level as well, as vc_set_namespace operates on vxis.
1165255833 M * Bertl I think the purpose of chbind is to do an unshare, and that's it
1165255846 M * Bertl same would go for chcontext (or vcontext)
1165255858 M * daniel_hozac and how would you enter it?
1165255874 M * Bertl that depends on how the tools are 'defined'
1165255881 M * daniel_hozac or would net_migrate handle that?
1165255884 M * Bertl I could imagine the following cases:
1165255910 M * Bertl - have tools vcontext and vnet* do the unshare, but no set_namespace()
1165255930 M * Bertl - have the chcontext and chbind toos do the unshare and set the spaces
1165255939 M * Bertl *tools
1165255967 M * daniel_hozac chcontext and chbind are just wrapper scripts.
1165255972 M * Bertl we probably want to identify the other spaces too (like nid/xid)
1165255987 M * Bertl daniel_hozac: so the set namespace would go into vnamespace, no?
1165256002 M * daniel_hozac right.
1165256104 M * daniel_hozac but then we'd need a way to specify which namespaces should be set.
1165256122 M * daniel_hozac and only fail if that namespace hasn't been set before.
1165256129 M * daniel_hozac s/hasn't/has/
1165256187 M * daniel_hozac so chbind would just set the socket namespace, chcontext would set IPC, pid and uts, etc.
1165256209 J * Vudumen a36f8182c0@perverz.hu
1165256434 M * Bertl okay, we add a similar mask to the set_namespace and make that a new version
1165256452 M * Bertl we also allow to override the settings
1165256471 Q * Piet_ Quit: Piet_
1165256632 J * Piet hiddenserv@tor.noreply.org
1165256862 Q * Vudumen Ping timeout: 480 seconds
1165257183 Q * Piet Remote host closed the connection
1165257856 Q * Loki|muh Ping timeout: 480 seconds
1165258080 J * RichyF ~RIchy@82-32-120-219.cable.ubr04.hawk.blueyonder.co.uk
1165258291 M * Bertl daniel_hozac: does that sound feasable to you?
1165258355 M * Bertl wb RichyF!
1165258365 M * RichyF :) hi
1165258396 M * daniel_hozac Bertl: yeah, sounds good.
1165258497 J * Vudumen ~vudumen@perverz.hu
1165258631 J * Loki|muh loki@satanix.de
1165259182 Q * Vudumen Ping timeout: 480 seconds
1165259386 J * Vudumen 2ef0a11bf3@perverz.hu
1165259427 Q * Loki|muh Ping timeout: 480 seconds
1165259625 J * Loki|muh loki@satanix.de
1165260141 Q * Vudumen Read error: Connection reset by peer
1165261415 M * Bertl okay, off for now .. back later ...
1165261419 N * Bertl Bertl_oO
1165262111 J * Vudumen c28212698f@perverz.hu
1165263250 Q * lilalinux Remote host closed the connection
1165263391 Q * DavidS Ping timeout: 480 seconds
1165263745 J * DavidS ~david@217.19.46.10
1165264248 Q * mnemoc Ping timeout: 480 seconds
1165264715 J * mnemoc ~amery@kilo105.server4you.de
1165264846 Q * sid3windr Ping timeout: 480 seconds
1165265042 J * undefined ~undefined@adsl-65-71-67-133.dsl.rcsntx.swbell.net
1165265047 J * sid3windr luser@bastard-operator.from-hell.be
1165265226 M * undefined does anybody know what's required to get libusb working in a guest (so cups can access a usb printer)?
1165265242 M * undefined there's this nonconclusive thread: http://www.mail-archive.com/vserver@list.linux-vserver.org/msg09632.html
1165265338 M * undefined and a generic comment about usb devices at http://linux-vserver.org/Capabilities_and_Flags ("SYS_RAWIO: Allow sending USB messages to any device via /proc/bus/usb")
1165265392 M * daniel_hozac and what's the problem, exactly?
1165265459 M * undefined daniel_hozac: my apologies, i remembered that email thread and thought i would do 10 minutes of research before getting myself ignorantly into an hour of pain
1165265467 J * DreamerC_ ~dreamerc@59-115-48-84.dynamic.hinet.net
1165265471 M * undefined daniel_hozac: so i haven't tried anything yet
1165265509 M * daniel_hozac http://www.mail-archive.com/vserver@list.linux-vserver.org/msg09633.html works for me.
1165265591 M * undefined daniel_hozac: saw the post and didn't mean to doubt your reply, but the guy never said that worked for him
1165265592 M * daniel_hozac i.e. i can see the files just fine.
1165265736 M * undefined daniel_hozac: thanks, i'll try that for cups w/ usb printer and maybe send my findings to the mailing list (since i can't find anything affirmative anywhere else for this scenario)
1165265773 M * daniel_hozac sounds like a good idea.
1165265821 Q * DreamerC Ping timeout: 480 seconds
1165265943 M * daniel_hozac feel free to put something on the wiki as well.
1165266641 J * Aiken ~james@tooax7-056.dialup.optusnet.com.au
1165267040 M * daniel_hozac so which of the boring and weedpage styles do people prefer for the great flower page?
1165267281 Q * bonbons Quit: Leaving
1165267431 M * undefined daniel_hozac: "boring"
1165268238 M * daniel_hozac i guess it'd be hard to keep calling it the great flower page then though... hmm.
1165268705 Q * sebastian 
1165268861 M * sid3windr weedpage!
1165268871 M * sid3windr weedpage is teh rox
1165268872 M * sid3windr ;)
1165269220 J * David1 ~david@217.19.46.18
1165269263 M * undefined daniel_hozac: i don't know why it's called "the great flower page" because by default it doesn't display a flower, but instead a leaf
1165269566 M * daniel_hozac i guess most people don't care to make that distinction ;)
1165269573 N * Bertl_oO Bertl
1165269578 M * Bertl back now ...
1165269581 Q * DavidS Ping timeout: 480 seconds
1165269607 M * Bertl daniel_hozac: well, we can keep the current default for IE :)
1165269624 M * undefined daniel_hozac: agreed
1165269686 M * daniel_hozac is anyone using IE? :)
1165269708 M * daniel_hozac but i'm not sure how we'd even accomplish that.
1165269717 M * Bertl we'll find out soon enough ...
1165269757 M * Bertl that's easy, you can check for that in a redirect rule or script and reorder the stylesheets
1165269792 M * Bertl probably it is even sufficient to put the IE default style into a comment :)
1165269932 M * daniel_hozac heh, the first sounds like a lot of work for (hopefully) nothing ;)
1165270212 M * Bertl well, I'm perfectly fine with the current default style, but the message is definitely wrong
1165270245 M * Bertl it says, if you don't like it .. send different one .. and so folks did send them, but nothing changed
1165270281 M * daniel_hozac i think the current default is rather difficult to read.
1165270296 J * meandtheshell ~markus@85-125-231-195.dynamic.xdsl-line.inode.at
1165270514 J * FireEgl ~FireEgl@adsl-61-147-76.bhm.bellsouth.net
1165270636 Q * mire Ping timeout: 480 seconds
1165270864 J * lilalinux ~plasma@dslb-084-058-200-151.pools.arcor-ip.net
1165271005 Q * dna Quit: Verlassend
1165271109 Q * David1 Quit: Leaving.
1165271429 J * mire ~mire@141-166-222-85.adsl.verat.net
1165271562 M * rgl is there a SNMP MIB for vservers?
1165271609 M * Bertl hmm, not that I know of .. what would you expect from that?
1165271704 Q * lilalinux Remote host closed the connection
1165271714 M * rgl actually, I'm just curious about this snmp thing that someone mentioned on the mailinlist :D
1165271743 M * hardwire sup puinks
1165271755 M * rgl but I guess one could get everything by having an snmp agent inside the guest hehe
1165271940 M * rgl gtg.  g'night!
1165271954 Q * rgl Quit: Fui embora
1165272619 Q * crypt Quit: Leaving
1165273745 M * sid3windr yeah
1165273748 M * sid3windr that snmp thing was me
1165273761 M * sid3windr but I haven't had time to check the new interface stuff
1165273775 M * sid3windr I have snmp agents inside the guests but would like to monitor running vservers memory usage etc
1165273828 M * daniel_hozac have you seen all the plugins to various monitoring daemons in the wild?
1165273836 M * daniel_hozac collectd, munin, etc.?
1165273978 M * Bertl sid3windr: IMHO _inside_ the guest is a bad choice
1165274058 M * Bertl you can get all guest specific info easily from the host, and even provided combined information
1165274076 M * Bertl and you can save the snmp daemon thingy per guest
1165274136 P * stefani I'm Parting (the water)
1165274143 M * sid3windr daniel_hozac: what do I need those for? :)
1165274165 M * daniel_hozac well, what do you need SNMP for? :)
1165274167 M * sid3windr Bertl: I probably could. but I'd have to change a lot of stuff if I move vservers to another hose
1165274185 M * sid3windr daniel_hozac: to be able to do centralized polling/stats/monitoring with a single program
1165274223 M * daniel_hozac and you can't do that with another daemon?
1165274242 M * sid3windr no, because then it's no longer the one centralized box/software, is it ? :)
1165274252 M * daniel_hozac why not?
1165274274 M * sid3windr because my cisco/dell/hp/juniper/alliedtelesis stuff doesn't run various daemons, for example ;)
1165274305 M * daniel_hozac you didn't mention those ;)
1165274315 M * sid3windr it's pretty much implied if you use snmp ;)
1165274323 M * sid3windr otherwise I could just use ssh keys and get some stuff from hosts anyway
1165274329 M * sid3windr but snmp is nice++ ;)
1165274340 M * sid3windr (... to think I hated it a year ago..)
1165274355 M * daniel_hozac well, if you write an SNMP plugin to export information about guests, i'm sure people would be interested.
1165274362 M * sid3windr that was my intention
1165274371 M * sid3windr I just need a way to get the data from a perl script
1165274379 M * sid3windr and parsing vserver-stat and /proc wasn't a good idea
1165274388 M * sid3windr so something else was
1165274404 M * sid3windr I have Bertl and other's messages marked up in my mailclient but simply haven't had the time to really look at it
1165274429 M * daniel_hozac recent devel kernels should export everything available in /proc through some syscall.
1165274448 M * sid3windr is it easy to do syscalls from perl? :)
1165274457 M * daniel_hozac i have no idea.
1165274461 M * Bertl yep, and vcd should allow to gather that ...
1165274465 M * daniel_hozac it is from Python :)
1165274471 M * daniel_hozac vstatd, i thought?
1165274480 M * Bertl ah, yes
1165274533 M * sid3windr :)
1165274546 M * sid3windr I'll have another look soonish
1165274548 A * sid3windr off now
1165274548 M * sid3windr thanks :)
1165275531 Q * _cob Ping timeout: 480 seconds
1165275641 J * _cob ~cob@pc-csa01.science.uva.nl