1184803315 P * Solaris satan made me do it
1184803475 J * dg_guru ~rwer@cpe-65-31-0-36.insight.res.rr.com
1184803619 J * pusling_ pusling@88.212.70.38
1184803663 Q * pusling Read error: Connection reset by peer
1184803901 J * Solaris ~satan@85.138.105.7
1184804356 J * sivana ~sivana@gromit.mixdown.ca
1184804361 M * sivana hello
1184804383 M * sivana how does one increase the disk size of the guest
1184804422 Q * pusling_ Ping timeout: 480 seconds
1184804579 M * emtty sivana: using dlimits?
1184804620 M * sivana umm, I'm not sure
1184804624 M * sivana someone else set it up
1184804633 J * Ashsong ~chatzilla@orchard.laptop.org
1184804642 M * sivana critical info? :)
1184804668 M * emtty check for /etv/vserver/<name>/dlimits/blah
1184804680 M * emtty /etc/...
1184804693 M * sivana no such dir
1184804701 M * sivana in /etc/vservers/tor-ast
1184804717 M * sivana apps  cache  cpuset  fstab  interfaces  name  run  uts  vdir
1184804781 M * emtty then perhaps guest has its fs on its own lvm? check output of 'df' on host
1184804791 M * Ashsong evening folks.
1184804793 M * sivana I think so, yes
1184804804 M * sivana /dev/mapper/vservers-tor--ast
1184804806 M * sivana                      1014M  729M  286M  72% /var/lib/vservers/tor-ast
1184804810 M * Ashsong Can anyone offer me a lesson on how to configure vserver's scheduling?
1184804854 M * emtty then check out lvresize
1184804876 M * sivana thanks
1184804911 M * emtty and depending on the filesystem, resize_reiserfs or what for ext3/xfs (?)
1184804924 M * emtty i know reiserfs can be done on-the-fly
1184804936 M * emtty idk about the others, never use them
1184804961 M * blizz i think it's resize2fs
1184805005 M * sivana cool.. it's a start
1184805006 M * sivana thanks
1184805029 M * blizz but dont forget to adjust the partition table, too
1184805085 Q * bzed Quit: Leaving
1184805448 J * pusling pusling@88.212.70.38
1184805972 Q * pusling Ping timeout: 480 seconds
1184806144 J * DoberMann_ ~james@AToulouse-156-1-21-202.w86-196.abo.wanadoo.fr
1184806252 Q * DoberMann[ZZZzzz] Ping timeout: 480 seconds
1184806329 Q * ema Ping timeout: 480 seconds
1184806940 J * ema ~ema@rtfm.galliera.it
1184806947 J * pusling pusling@88.212.70.38
1184807519 Q * pusling Ping timeout: 480 seconds
1184807767 J * pusling pusling@88.212.70.38
1184808602 Q * Ashsong Quit: ChatZilla 0.9.78.1 [Firefox 2.0.0.4/2007051502]
1184811654 Q * tam Quit: Changing server
1184813972 J * Ashsong ~chatzilla@orchard.laptop.org
1184814000 M * Ashsong Hello.
1184815024 M * sivana exit
1184815026 Q * sivana Quit: leaving
1184815096 Q * ahuman Quit: Leaving
1184816533 N * Bertl_zZ Bertl
1184816542 M * Bertl morning folks!
1184816591 M * Ashsong Good morning!
1184816609 M * Bertl hey Ashsong! how's going?
1184816616 M * Ashsong Bertl: It's going well.
1184816635 M * Ashsong I'm trying to figure out how to get CPU scheduling working.
1184816645 M * Ashsong or rather, token bucketed scheduling.
1184816646 M * Bertl ah, what's the problem?
1184816655 M * Ashsong I just don't know what I'm doing yet. :)
1184816656 M * Bertl you got the docu pages?
1184816666 M * Bertl http://linux-vserver.org/CPU_Scheduler
1184816670 M * Ashsong Mmm... I've been working from the source code and some things I found on the wiki.
1184816683 M * Ashsong yeah, I got that, but I'm not sure what order the calls have to be made in.
1184816708 M * Ashsong In particular, I'm getting ENOSYS errors when I try to call vc_set_sched()
1184816725 M * Ashsong That might be because I'm trying to call it from inside the created, but not yet running context.
1184816728 M * Bertl maybe you are trying to call it from inside the guest?
1184816732 M * Ashsong Or it might be for some other reason.
1184816735 M * Ashsong Yeah.
1184816745 M * Ashsong Do I need to call it from outside, or is there a capability I can set?
1184816757 M * Bertl you don't want to be able to set it from inside
1184816768 M * Bertl would kind of defeat the purpose :)
1184816784 M * Bertl but they should work fine when still in setup state
1184816784 M * Ashsong Bertl: True enough.
1184816796 M * Ashsong ATM I was just trying to figure out how to use it. :)
1184816825 M * Bertl btw, do you have VSERVER_DEBUG enabled in the kernel?
1184816836 M * Ashsong Yeah, but I forgot to turn it on. 
1184816839 M * Ashsong I'll do so momentarily.
1184816857 M * Bertl well, you don't want to turn everything on always
1184816871 M * Bertl but the switch debugging could be turned on, I guess
1184816883 M * Ashsong That's what I meant.
1184816884 M * Bertl that should give you some hints if something goes wrong
1184816914 M * Ashsong [11259.186160] VCMD_ctx_create_v0 [v0]        2001,00000000 [0,5,0,0]
1184816916 M * Ashsong [11259.233540] VCMD_ctx_create_v0 [v0]        = 000007d1(2001) [8,1]
1184816917 M * Ashsong [11259.278450] VCMD_get_version [v0]          63,00000000 [0,0,0,0]
1184816919 M * Ashsong [11259.318981] VCMD_get_version [v0]          = 00020300(131840) [8,1]
1184816920 M * Ashsong [11259.432341] VCMD_get_vci [v0]              0,00000000 [0,0,0,0]
1184816922 M * Ashsong [11259.475724] VCMD_get_vci [v0]              = 031107b9(51447737) [8,1]
1184816923 M * Ashsong [11259.518973] VCMD_14_1[4]                   2001,bfbcc54c [0,-1,0,0]
1184816925 M * Ashsong [11259.546426] VCMD_14_1[4]                   = ffffffda(-38) [0,-1]
1184816926 M * Bertl (please use paste.linux-vserver.org for everything longer than 3 lines) 
1184816926 M * Ashsong [11259.880113] VCMD_wait_exit [v0]            2001,bfbcc778 [0,4,1,1]
1184816928 M * Ashsong [11259.887637] VCMD_wait_exit [v0]            = fffffffd(-3) [6,1]
1184816931 M * Ashsong Ah, my apologies.
1184816943 M * Bertl otherwise you will wake up too many people :)
1184817000 M * Ashsong in any case, we are definitely in the setup state during the call to vc_set_sched().
1184817010 M * Ashsong Yes?
1184817116 M * Bertl no
1184817138 M * Ashsong http://pastebin.com/d40f30b86
1184817142 M * Ashsong That's what I'm attempting to do.
1184817202 M * Bertl does pvs.c_vc_ctx_flags() guess or retrieve the values?
1184817250 M * Ashsong It's just allocates a vc_ctx_flags structure
1184817253 M * Bertl if it retrieves the flags (via syscall) I would be interested in the output
1184817267 M * Ashsong I can get that, if you want to see it.
1184817274 M * Ashsong Back in a jiffy... :)
1184817281 M * Bertl yes, please .. before you try to set the scheduler values
1184817399 M * Ashsong Will do.
1184817423 M * Bertl the call itself should suffice, you do not need to actually output the result
1184817430 M * Bertl it should show up in the switch debug
1184817455 M * Ashsong Ah, clever.
1184817459 M * Ashsong flagword is 700000000
1184817480 M * Ashsong mask is 7ffffffff
1184817493 M * Ashsong this is immediately after the call to create.
1184817505 M * Bertl that looks good
1184817523 M * Bertl could you upload the debug trace please?
1184817530 M * Bertl (for a complete run)
1184817557 M * Ashsong Sure.
1184817576 M * Ashsong Incidentally, the 7 there indicates that VXF_STATE_SETUP, VXF_STATE_INIT, and VXF_STATE_ADMIN are all set?
1184817587 M * Bertl yep
1184817694 M * Bertl the mask looks suspicious though
1184817734 M * Bertl that could actually be a bug (but one which shouldn't affect you)
1184817774 M * Bertl this is on x86, right?
1184817863 M * Ashsong Yeah.
1184817876 M * Ashsong Bertl: it could also be a bug in our high-level binding, no?
1184817920 M * Bertl nah, looks more like a kernel issue
1184817937 M * Bertl do we have the trace somewhere?
1184817992 M * Ashsong okay, no immediate change in flags after the attempt to call vc_set_sched(); however, after the call to set_flags() to turn on SCHED_HARD, we have flags: 700000100 7FFFFFFFF, then, after run(),  flags: 600000100 6FFFFFFFF
1184817999 M * dg_guru where do the vdir's point to ?
1184818009 M * dg_guru i dont have a vservers dir in my / hmmm 
1184818019 M * Bertl dg_guru: usually to the guest root
1184818038 M * dg_guru yea but i dont have a vservers in my / on the host system
1184818038 M * Bertl dg_guru: if you are using a debian flavor, look in /var/lib/vservers
1184818044 M * dg_guru ah
1184818079 M * dg_guru debian makes things so much more interesting
1184818087 M * Bertl indeed
1184818105 M * dg_guru i might switch back to Gentoo but hmmm
1184818106 M * Bertl btw, may I quote that next time :)
1184818123 M * Ashsong http://pastebin.com/d22c5b5ee
1184818140 M * dg_guru Bertl, wasnt a quote it was truth ;) 
1184818158 M * dg_guru linux distros is like socks they never fit right 
1184818182 M * Bertl Ashsong: ah, you use daniels 'interpreter' script?
1184818202 M * dg_guru i dont like debian casue everything is old but yet gentoo everything is new but yet it has long compiles 
1184818224 M * Ashsong Bertl: probably yes. a bash script named vsdmesg?
1184818241 M * Bertl Ashsong: the problem is simple, you are using the V4 scheduler interface, which doesn't exist in the OLPC branch
1184818261 M * Bertl Ashsong: switching to the V5 interface should fix that :)
1184818262 M * Ashsong Beautiful.
1184818297 M * Ashsong That's the [4] marking after the VCMD_14_1 statement?
1184818304 M * Bertl exactly
1184818317 M * Bertl currently checking why that doesn't get interpreted by your vsdmesg
1184818336 M * Ashsong Hmm. 
1184818351 M * Bertl Ashsong: you might also consider switching from the V0 ctx_create to V1^
1184818354 M * Bertl -^
1184818358 M * Ashsong Ah.
1184818369 M * Ashsong What is selecting the syscall version to use?
1184818379 M * Bertl it is part of the syscall command
1184818380 M * Ashsong Is that libvserver.so.0 or is that the kernel that we have installed?
1184818396 M * Bertl so I guess the libvserver
1184818406 M * Ashsong so the version of util-lib-vserver is using older syscalls?
1184818417 M * Ashsong hmm.
1184818417 M * Bertl I presume so
1184818420 M * Ashsong alright.
1184818423 M * Ashsong That's not so bad.
1184818464 M * Ashsong I can just reuse its syscall() mechanism to update the particular calls I'm trying to use.
1184818514 M * Bertl dg_guru: I usually end up taking a distro (currently mandriva) and recompiling the important stuff to suit my needs
1184818514 M * Ashsong Hmm. Let me go experiment.
1184818546 M * dg_guru Bertl, theres always something
1184818569 M * dg_guru i guess debian is good
1184818617 M * Bertl Linux in general is good! :)
1184818693 M * Bertl Ashsong: but congrats, you found at least two (minor) bugs, I'm fixing right now! tx!
1184818709 M * Ashsong Bertl: glad I'm good for something. :)
1184818737 M * Ashsong (I had a lot of fun this afternoon reading the kernel looking for what eventually turned out to be vx_schedule() )
1184818765 M * Ashsong (One little question I had as I was reading:
1184818823 M * Ashsong Is vx_schedule() supposed to return '1' or 'ret' as it falls off the final case where CONFIG_VSERVER_HARDCPU is undefined?  (this is in kernel/sched_hard.h))
1184818910 M * Ashsong (line 307 of sched_hard.h, for me)
1184818910 M * dg_guru time ot develop my site 
1184818919 M * dg_guru i think i will do it windows 
1184818982 M * Bertl Ashsong: the '1' is correct, the ret value is discarded
1184818991 M * Ashsong okay.
1184819049 M * dg_guru i will never understand why adobe wont put photoshop into linux 
1184819059 M * Ashsong next question: when I make the syscall from userland, using util-vserver's vc_syscall() function, how do I indicate the version of the command that I want?
1184819147 M * Ashsong e.g. how do I actually cause  VCMD_set_sched(1, 5)  to be called from userland
1184819149 M * Bertl check include/linux/vserver/switch.h (the VC_CMD at the first page)
1184819158 M * Ashsong okay
1184819185 M * Ashsong Ah, I just duplicate that macro?
1184819206 M * Bertl actually you could/should/would include the header :)
1184819228 M * Bertl (which is produced by the kernel build system, for user inclusion)
1184819269 M * Ashsong Sadly, python's ctypes module doesn't parse C headers, requiring its users to duplicate constants, macros, and enums.
1184819284 M * Ashsong If we were using a C extension to python, I would certainly include it.
1184819286 M * Bertl yeah, then you ahve to duplicate it
1184819295 M * Ashsong Eh. Could be worse.
1184819336 M * neuralis afaik pyrex parses headers, doesn't it?
1184819365 M * Bertl hey neuralis! how's going?
1184819392 M * Supaplex pyrex stays crunchy in milk ;)
1184819412 M * Bertl lol
1184819423 M * neuralis Bertl: pretty well, long day, even longer day tomorrow, and a pile of work to do tonight
1184819430 M * neuralis tired
1184819431 M * neuralis Bertl: you?
1184819455 M * Bertl fine, fine, just got up (trying to switch my day/night schedule)
1184819465 M * neuralis hah, good luck with that
1184819487 M * Bertl well, works reasonably well for a few days ... (which usually suffices)
1184819490 M * neuralis it's worse for me than i imagine quitting smoking is like -- i don't even last a week before relapsing
1184819516 M * Bertl I fall back to my 'normal' cycle after a week at most
1184819630 M * Ashsong neuralis: did you follow the discussion about util-vserver-lib being out-of-date w.r.t the VServer patch we're using?
1184819669 M * neuralis haven't caught up on scollback yet, no
1184819691 M * Bertl I think daniel_hozac didn't implement all the latest versions yet
1184819714 M * Bertl as we removed the 'older' syscall commands from the OLPC branch, this now shows up
1184819736 M * Bertl so I think that should be fixed pretty soon, once he knows
1184819760 M * Ashsong Bertl: So long as my proposed workaround actually works, I'm not too worried.
1184819785 M * Bertl even if that fails, I have at least two workarounds for you
1184819793 M * Ashsong Ah, excellent.
1184819799 M * Ashsong In that case his demo tomorrow is probably safe. :)
1184819806 M * Bertl 1) you can get a patch to add V4 scheduler supprot into the kernel
1184819829 M * Bertl 2) you can probably quite easily bump the version in the lib (and change the argument structure)
1184819853 M * Ashsong Bertl: couldn't I just make the syscall directly from C?
1184819870 M * Ashsong (if it winds up being difficult to do from python)
1184819879 M * Bertl sure, but that would probably add more code on your side, which isn't necessary
1184819897 M * Bertl i.e. the v4 compat code is already there (in mainline Linux-VServer)
1184819974 M * Bertl OTOH, my original suggestion was to use the syscall directly in the first place (i.e. in the python wrapper)
1184819976 M * Ashsong Bertl: granted, but I'm only concerned about figuring out what steps I have to do.
1184819995 M * Ashsong Bertl: (which is what I'm doing. I'll move to C if that doesn't work)
1184820030 M * Bertl especially as the libvserver only makes sense with a general interface (which has to work with a lot of different ABIs) but in your case, you have a well defined interface to the kernel
1184820048 M * Ashsong Bertl: I'm just doing this to get comfortable with how to control the scheduler. If it doesn't work tonight, or next week, it's no big deal.
1184820070 M * Ashsong Further, the changes are confined to our 'pylibvserver' binding.
1184820117 M * Ashsong The important thing is just that I work out all the necessary steps.
1184820133 M * Bertl of course
1184820228 M * Ashsong :)
1184820246 M * Ashsong Prototypes: can't live with 'em, can't live without them.
1184820282 M * Bertl most prototypes (especially the really hacky ones) go straight into production :)
1184820403 M * Ashsong Sad, but true.
1184820642 M * Ashsong Bingo.
1184820646 J * agryppa ~kb2qzv@cab-dr-cas2-56.dial.airstreamcomm.net
1184820652 M * Bertl wb agryppa!
1184820655 M * Ashsong Making the syscall directly from Python appears to complete without a hitch.
1184820656 M * agryppa hi
1184820671 M * Bertl Ashsong: excellent! check the switch debug
1184820677 M * Ashsong Now to see if the options I passed were correct... :)
1184820680 M * Ashsong Bertl: will do.
1184820726 M * Ashsong Heh.
1184820733 M * Ashsong Given how long the activity is taking to start, I suspect yes...
1184820743 M * Ashsong :)
1184820792 M * Ashsong [15055.083665] VCMD_set_sched [v5]            2001,0823d260 [0,7,1,e]
1184820800 M * Bertl looks good
1184820803 M * Ashsong [15055.090242] VCMD_set_sched [v5]            = 00000000(0) [8,2]
1184820804 M * Ashsong Indeed.
1184820807 M * Ashsong To me too.
1184820822 M * Ashsong Bertl: thanks a bunch!
1184820834 M * Bertl you're welcome!
1184820925 M * Bertl Ashsong: note, the v5 interface also simplifies setting the idle time values
1184820945 M * Ashsong Ah. Could you please explain what the idle time values do?
1184820959 M * Ashsong It wasn't immediately apparent from reading either the source code or the wiki.
1184820977 M * Bertl they are controlling the fair shair behaviour
1184820986 M * agryppa my network settings without guests up are this: br0 (eth0 and tun0). With br0 as the only one to have an IP. After guests are up eth0 gets an IP. Is this OK? I rememeber creating guest with eth0 argument, Should I have given br0 instead?
1184821020 M * Ashsong (I assume you're referring to things like INTERVAL2 and FILLRATE2...)
1184821039 M * Bertl Ashsong: short version: the TB normally operates on rate/interval, when the CPU would get idle, the idle time enabled contexts switch to rate2/interval2 and artificially advance time
1184821062 M * Bertl agryppa: if you want br0 to carry the guest ip, yes
1184821080 M * Ashsong Ah.
1184821151 M * agryppa Can guests get their IP via dhcp? Is it configurable after guests are already defined, up and running?
1184821205 M * Bertl agryppa: yes, but as guests do not have a separate MAC (they are isolated at IP/Layer 3) you have to use some 'guest' identifier
1184821228 M * Bertl (also this is purely a host problem, neither the tools nor the guest itself is involved)
1184821324 M * agryppa when one creates two guests should they be defined with eth0 eth1 respectively? I think I have spooked everything here :-)
1184821348 M * Bertl only if you want them to use different _host_ interfaces
1184821387 M * agryppa should host have those interfaces phisically present, though?
1184821390 M * Bertl e.g. eth0 with 10.0.0.42 and eth1 with 192.168.0.42 (two different networks)
1184821400 M * Bertl agryppa: yes, that is required 
1184821528 M * agryppa my host has eth0 only but I am using tap0 with it for QEMU so I bridged it. Now with vserver guest, ubuntu and gentoo, I thing I created them with eth0 argument. Was that correct?
1184821575 M * Bertl probably better to use br0 then, but in this case eth0 should work too
1184821595 M * agryppa thank you Bertl for your answers.
1184821596 M * Bertl (although all bridge documents state that you should not assign anything to the interfaces)
1184821615 M * Bertl np
1184821640 M * agryppa I wonder whether it is possible now to change it. from eth0 to br0?
1184821671 M * Bertl sure, just shut down the guest, then edit the 'dev' file in /etc/vservers/<guest>/interfaces/0
1184821685 M * agryppa I'll check that.
1184821856 M * agryppa I found out that guest gentoo was sitting on br0 interface. And guest ubuntu on eth0. Guest gentoo has dev, ip, prefix files in that directory. Ubuntu just dev and ip. What gives?
1184822190 Q * FireEgl Ping timeout: 480 seconds
1184822256 M * Bertl you should set the prefix to your network prefix, otherwise the guest will be handled like an isolated (/32) IP
1184822441 M * agryppa its 24 in guest gentoo. I will touch prefix and put 24 in ubuntu, too.
1184822492 M * agryppa last one question tonight. When entering guest ubuntu I get this: cannot find name for group ID 11. what is it?
1184822609 M * Bertl what version of util-vserver do you use?
1184822642 M * agryppa   Installed versions:  0.30.213
1184822677 M * Bertl hmm, maybe a tool bug, basically I assume it is caused by having different uid/gid between host and guest
1184822695 M * Bertl and for whatever reason, you enter with group 11
1184822709 M * agryppa one guest has a 1011 and the other 1012 context value.
1184822709 M * Bertl (which is not defined in /etc/group)
1184822722 M * Bertl I think that is unrelated
1184822741 M * Bertl what does grep :11: /etc/group give you on the host?
1184822758 M * agryppa when I enter guest gentoo it doesn't complain. 
1184822794 M * agryppa localhost ubuntu # grep :11: /etc/group
1184822794 M * agryppa floppy::11:root,haldaemon
1184822815 M * agryppa sorry, that was a guest. wait.
1184822842 M * agryppa not, it is correct. I was just in the dir called ubuntu :-)
1184822864 M * Bertl so root is in that group, as you can see
1184822881 M * Bertl I assume the guest where it complains, doesn't know that group
1184822926 M * Bertl please (if you got the time) file a bug report against util-vserver (I guess it will be fixed in the next release)
1184823006 M * agryppa I will gladly do that but wait. In fact guest_ubuntu doesn't have :11: in its group file. Guest_gentoo does have it. So maybe it is a distro problem?
1184823016 M * Bertl Ashsong: http://vserver.13thfloor.at/Experimental/TOOLS/vcmd-0.10.tar.bz2 (should interpret the 'old' sched call correctly, and also supports v5 interfaces for testing)
1184823028 M * Bertl agryppa: no, a distro is not required to have group 11 :)
1184823038 M * agryppa ok.
1184823045 M * Bertl agryppa: AAMOF, each distro/guest is completely independant
1184823068 M * Bertl agryppa: just that 'enter' brings something from the host into the guest (in this case the group 11)
1184823092 M * Bertl (which should not happen)
1184823116 M * agryppa can I fix it temporarily by just creating a group on the guest?
1184823128 M * Bertl you can actually ignore it
1184823138 M * agryppa right :-)
1184823147 M * Bertl or if you really like to fix it, remove root from the floppy group on the host
1184823157 M * Bertl (which is kind of moot anyway)
1184823219 M * agryppa so this is unrelated to the context value, right. by the way. Do you just make those up, the context values? 
1184823223 M * Bertl (note: to have effect, you need to log out and logon again, check with groups)
1184823243 M * Bertl agryppa: basically you can make them up as you go (2-49151)
1184823263 M * Bertl agryppa: they should be unique, and the tools just choose one for you if you do not specify it
1184823309 M * Bertl (they are in /etc/vservers/<name>/context)
1184823447 M * agryppa I've fought for two days to get xdm running on guests. Still nothing. that's why I asked the question about networking (br0, eth0 e.t.c.). I will have to check what is going on with that xdm. Basically I do X -query ubuntu, and ubuntu has xdm running. No login though.
1184823492 M * Bertl hmm, by default (recent) X/xdm have tcp disabled, sure you enabled it?
1184823502 M * agryppa yes I did.
1184823512 M * Bertl does it show up in lsof -ni ?
1184823557 M * agryppa please explain what lsof -ni is?
1184823573 M * Bertl 'man lsof' :)
1184823697 M * agryppa I don't have that command available on the guests. the host has it. what line should I be looking for?
1184823723 M * Bertl you want to look inside the guest (or at least inside the guest's network space)
1184823758 M * Bertl and you are looking for xdmcp/177
1184823797 M * Bertl you have xdmcp enabled too, yes? (if unsure, check out http://tldp.org/HOWTO/XDMCP-HOWTO/index.html)
1184823817 M * agryppa I think lsof from host wouldn't show anything about guests, would it? xdmcp/177 exists after I start xdm on a guest.
1184823861 M * agryppa netstat -plnu: udp        0      0 192.168.0.3:177         0.0.0.0:*                           2380/xdm
1184823882 M * Bertl 192.168.0.3 is the guest ip?
1184823891 M * agryppa yes it is.
1184823902 M * Bertl only udp shows up? no tcp?
1184823908 M * agryppa yes
1184823973 M * agryppa no tcp. Should it be there?
1184824115 M * Bertl what does 'X -query 192.168.0.3' produce in the logs?
1184824201 M * Bertl (do the access controls in Xaccess allow your client?)
1184824418 M * agryppa xdm's Xaccess allow my client. Which logs do you mean, the host's or the guest's logs?
1184824534 M * Bertl the logs for the xdm
1184824586 M * agryppa here is a snippet of the logs:
1184824588 M * agryppa xdm error (pid 2380): XDMCP socket creation failed, errno 97
1184824588 M * agryppa xdm error (pid 2380): XDMCP socket multicast join to 0.0.0.0 failed, errno 22
1184824588 M * agryppa xdm error (pid 2382): server open failed for gentoo:1, giving up
1184824588 M * agryppa xdm error (pid 2380): Display gentoo:1 cannot be opened
1184824588 M * agryppa xdm error (pid 2402): server open failed for gentoo:1, giving up
1184824680 M * Bertl hum, maybe you should first describe your setup (and what you want to achieve) first?
1184824686 M * Bertl -first
1184824706 N * DoberMann_ DoberMann
1184824760 M * agryppa I think its kinda late for me now. I will read on this topic a little more and see what I can do. Its 00:55 here at my place. If you are in Austria then is 7:55, is it?
1184824777 M * Bertl yep, in the morning :)
1184824784 M * agryppa sunny?
1184824791 M * Bertl yep, and quite hot
1184824806 M * agryppa not humid though I hope.
1184824813 J * FireEgl FireEgl@Sebastian.Atlantica.CJB.Net
1184824849 M * Bertl well ... :)
1184824857 M * agryppa I will be in Poland in September. A lot closer to Austria than from here :-)
1184824875 M * Bertl okay, so we continue in september *G* :)
1184824887 M * agryppa or tomorrow :-)
1184824907 M * agryppa bye now. Thank you for your time.
1184824916 M * Bertl you're welcome!
1184824922 Q * agryppa Quit: Leaving
1184825476 Q * Hollow Remote host closed the connection
1184825489 J * Hollow ~hollow@proteus.croup.de
1184825838 J * DavidS ~david@85.125.165.34
1184825854 M * Bertl morning DavidS!
1184825856 N * DavidS DavidS|Wien
1184825899 M * DavidS|Wien hi Bertl!
1184826005 J * HeinMueck ~Miranda@dslb-088-064-030-094.pools.arcor-ip.net
1184826012 M * Bertl wb HeinMueck!
1184826036 M * HeinMueck Howdy Bertl :)
1184826640 N * DoberMann DoberMann[PullA]
1184826770 Q * jkl Ping timeout: 480 seconds
1184826998 J * jkl jkl@c-67-173-253-237.hsd1.co.comcast.net
1184827626 M * waldi hmm, what is the difference between 2.2.0 and 2.2.0.2?
1184827639 M * Bertl bugfixes and cleanups
1184827654 M * Bertl there will be more shortly (2.2.0.3)
1184827705 J * dna ~naucki@1-217-dsl.kielnet.net
1184828223 Q * LaZaR 
1184828440 J * markus__ ~chatzilla@mail.netcare.at
1184828505 M * markus__ Hi.damn. I deleted a vserver files (except in dev/pts, proc/ and tmp/ because Device or resource busy
1184828521 M * Bertl hey markus__!
1184828533 M * markus__ ) and now I can't stop the vserver: "Can not find a vserver-setup at '/etc/vservers/testing01/" . Yeah, i deleted it. I thought the vserver was down. Which he wasn't. How to fix it?
1184828536 M * markus__ Hi bertl :)
1184828577 M * Bertl vkill is your friend
1184828589 A * markus__ think Bertl is my friend :)
1184828619 M * markus__ great!!! thanks!!!
1184828623 M * Bertl you're welcome! 
1184829013 M * markus__ Btw, I finally upgraded my kernel incl. vserver patch to 2.6.21 / 2.2.01 :-)
1184829105 J * cluk ~cluk@p5B17F208.dip.t-dialin.net
1184829120 M * markus__ goodbye debian kernel
1184829275 M * cluk Hi
1184829528 Q * DavidS|Wien Quit: Leaving.
1184829681 M * Bertl wb cluk!
1184829716 M * cluk Hi Bertl
1184829748 M * cluk I am still fighting with the Debian kernel and it's cpu scheduling. :)
1184829827 M * Bertl did you try a mainline kernel yet?
1184829991 M * markus__ cluk: I don't know your specific problem and I'm a complete greenhorn in this case, but all troubles went away once I grabbed an official kernel and installed that under debian
1184830005 M * markus__ (did I say I had CPU scheduling problems too)
1184830040 M * cluk No, did not try mainline yet as this machine is running in production already. I usually prefer debian kernels so I dont have to think about security updates.
1184830066 M * markus__ same for me, but ...
1184830082 M * Bertl well, you might talk to waldi, maybe he can help :)
1184830105 M * markus__ cluk: just in case, I used this guide http://www.howtoforge.com/kernel_compilation_debian_etch and had no problems at all
1184830135 M * markus__ Bertl: is it just my perception or are there more debian related than other distribution question coming up here?
1184830184 M * cluk markus__: thanks. What kind of scheduler probs have you been facing?
1184830192 M * Bertl well, debian is quite popular, but unfortunately often outdated and broken
1184830343 M * cluk Bertl: waldi is Bastian Blank?
1184830599 M * Bertl last time I checked, whois said so :)
1184830655 N * Bertl Bertl_xX
1184830659 N * Bertl_xX Bertl
1184830702 M * cluk I will try submitting a bug report. :)
1184830721 M * cluk on the debian kernel of course. :)
1184830763 M * markus__ cluk: I had two vserver with just the same scheduler fillrate/interval and starting the second vserver resulting (from my subjective feeling) on starting a 10Hz machine .. after upgrading, no problem
1184830776 M * Bertl cluk: well, would be interesting to know if the issues go away with mainline
1184830796 M * Bertl cluk: maybe you _are_ observing Linux-VServer issues after all :)
1184830838 M * cluk Bertl: you mean same kernel and vserver version just mainline sources?
1184830866 M * cluk or latest mainline?
1184830875 M * Bertl well, what version are you currently using?
1184830891 M * Bertl (i.e. do you know what vserver version this kernel includes?)
1184831228 M * cluk according to the changelog it is 2.0.2.2-rc9
1184831278 M * Bertl on what baseline kernel?
1184831467 M * cluk baseline is 2.6.18. but I do not know which bugfix version
1184831521 M * cluk i have just noticed that there is an updated kernel available in etch-proposed-updates which contains two vserver fixes.
1184831551 M * Bertl okay, then I would suggest the http://vserver.13thfloor.at/Experimental/patch-2.6.18.5-vs2.0.3-rc1.diff
1184831735 M * cluk Applied to the debian or mainline kernel source?
1184831823 Q * virtuoso Ping timeout: 480 seconds
1184832106 J * virtuoso ~s0t0na@80.253.205.251
1184832838 M * Bertl cluk: the mainline of course
1184833365 Q * AStorm Quit: Bye
1184834111 J * AStorm ~astralsto@host-81-190-179-124.gorzow.mm.pl
1184834140 M * cluk markus__: the behaviour here is mostly what you describe. 
1184834166 M * markus__ :|
1184834206 M * cluk starting a vserver with sched_hard turned off and turning it on afterwards leaves a working but very very very slow vserver.
1184834261 M * markus__ I gave up my holy religion to only use debian kernels when my quad-cpu server wasn't able to run individual vservers on more then perceptive 10Hz :)
1184834270 M * cluk i would bet this is an overflow issue, because watch -n1 cat /proc/virtual/xid/sched show large negative values quiet often
1184834299 M * cluk this can not be observed on our second system where sched_hard is working correctly
1184834327 M * cluk this second system is also running a debian kernel. just an older version.
1184834516 M * Bertl well, if you observe it with mainline, it is considered a bug (here) and will be fixed
1184834865 J * cedric ~cedric@80.70.39.67
1184835402 Q * HeinMueck Quit: Aah!
1184835420 M * Bertl wb cedric!
1184835525 J * Piet hiddenserv@tor.noreply.org
1184835681 M * Guy- CTX   PROC    VSZ    RSS  userTIME   sysTIME    UPTIME NAME
1184835686 M * Guy- 24     243 183.1M   1.2G   8h49m59  41m26s26   3d12h43 av3notes
1184835693 M * Guy- can RSS be bigger than VSZ? :)
1184835718 M * Bertl unlikely, looks more like an overflow
1184835722 M * Bertl 0.30.213?
1184835723 M * Guy- 2.6.20.6-vs2.2.0, 0.30.213
1184835758 M * Bertl I would file a bug report or contact daniel_hozac ...
1184835779 M * Guy- OK
1184835804 M * blizz hmm, could be an overflow
1184835815 M * blizz (just to verify that idea :-P)
1184835884 J * meandtheshell ~markus@85.127.105.7
1184835997 M * blizz hmm, debian 4.0 util-vserver still at 0.30.212, right?
1184836063 M * Bertl I'm sure they will update to 0.30.213 when we release 0.30.214 :)
1184836071 M * blizz *lol*
1184836079 M * blizz yeah, yeah.. always on the safe side :-P
1184836180 M * blizz hey, they have 0.30.213 in testing and unstable
1184836187 M * cluk Bertl: We have one amd64 machine where the sched_hard is working correctly, one where it does not.
1184836201 M * Bertl with identical kernels?
1184836222 M * cluk the working one has a kernel running which is build from debian sources with one single additional patch applied.
1184836243 M * cluk this is a 2.6.16 kernel
1184836262 M * cluk the one which is not working has an original etch 2.6.18 debian kerne.
1184836290 M * Bertl well, you are comparing apples and oranges :)
1184836303 M * cluk well, wait a moment.
1184836306 M * cluk :)
1184836349 M * cluk i had to apply this one additional patch on the working system to get the sched_hard working as far as i remember
1184836361 M * Bertl hehe :)
1184836372 M * cluk this patch would still apply to the original debian kernel source in etch
1184836393 M * cluk i have pasted the patch on the usual place. would you take a look?
1184836456 M * Bertl could be to blame, definitely nothing which affects mainline
1184836516 M * Bertl maybe you should have filed a bug report to debian when you patched 2.6.16 :)
1184836557 M * cluk yes. i should. :)
1184836583 M * Bertl would have saved you quite some worries (as would have using a mainline kernel)
1184836815 Q * UukGoblin Remote host closed the connection
1184836923 J * UukGoblin ~jaaa@sr-fw1.router.uk.clara.net
1184837118 Q * AStorm Quit: Bye
1184837358 M * markus__ Can I raise the virtual memory (rss.hard) limit while the vserver is running? Currently I've 65536 but I need 131072
1184837448 M * Bertl sure
1184837622 M * markus__ Which comand does this? vdlimit?
1184837770 J * Pazzo ~ugelt@195.254.225.136
1184837889 M * Bertl nope, vlimit
1184837896 M * Bertl welcome Pazzo! LTNS!
1184837905 M * Pazzo hi Bertl!
1184837911 M * Bertl markus__: vdlimit is for 'disk limits'
1184837922 M * markus__ vrlimit ? :)
1184837924 M * Pazzo really long time, once again :)
1184838193 M * Pazzo Bertl: how are you? everything fine?
1184838211 M * Bertl everything fine, ad you?
1184838213 M * Bertl *and
1184838522 M * harry vd limit, isn't that viral disease? ;)
1184838544 M * markus__ Bertl: hmm, I used vlimit but when checking inside the vserver it added the additional ram to the swap space. Is this the way this is handled?
1184839066 M * Bertl markus__: depends on the kernel version and what you actually changed
1184839130 M * markus__ vlimit -c 40 --rss 131072
1184839192 M * Bertl okay, so you changed the hard limit, which is shown as swap space (over the soft limit)
1184839217 M * markus__ okay ...
1184839229 M * markus__ the swap space was 0
1184839242 M * markus__ probably not a good idea ...
1184839348 M * markus__ ?
1184839408 M * Bertl well, doesn't matter, it's just what the guest sees
1184839600 M * markus__ ah, ok
1184839628 M * Bertl i.e. right now it has no bad effect on the guest to have no swap :)
1184839693 J * AStorm ~astralsto@host-81-190-179-124.gorzow.mm.pl
1184839989 M * Bertl wb AStorm!
1184840065 M * AStorm Hmm, I'm building a weird RAID array right now ;-)
1184840082 M * Bertl out of floppy drives?
1184840095 M * blizz lol
1184840098 Q * emtty Read error: Operation timed out
1184840112 M * Bertl well, that would be weird, wouldn't it?
1184840143 M * blizz yeah, pretty weird
1184840210 M * AStorm Bertl, pendrives
1184840217 M * AStorm close, yes?
1184840242 M * AStorm it's raid0 for non-critical parts, raid 1 for more critical ones
1184840249 M * Bertl ah, striping? to push the wear leveling over the edge?
1184840250 J * lilalinux ~plasma@dslb-084-058-199-178.pools.arcor-ip.net
1184840268 M * AStorm Well, it shouldn't do that really
1184840379 M * AStorm actually, if the stripe size is right, it should help wear leveling
1184840387 M * Pazzo Bertl: (sorry, been busy) fine too, thanks!
1184840447 M * Bertl AStorm: you can test that theory with QEMU, if you like ...
1184840460 J * HeinMueck ~Miranda@ppp-62-245-208-116.dynamic.mnet-online.de
1184840519 M * AStorm Bertl, it can emulate wear-leveled drives?
1184840546 M * Bertl I did write an emulation for the cafe nand flash controller
1184840552 M * blizz talking about flash memory.. anyone heard about/tried "intel turbomemory"?
1184840587 M * AStorm blizz, no, it's still in the labs
1184840626 M * blizz are you really sure? i found it in some product descriptions (they could very well be wrong)
1184840651 M * AStorm how fast is it supposed to be?
1184840675 M * blizz that's what i'd love to know... i'll google
1184840701 M * Bertl blizz: 'Works on Windows Vista only' (tm)(R)(C)
1184840734 M * Bertl http://www.intel.com/design/flash/nand/turbomemory/316979.pdf
1184840742 M * blizz yeah, officially :-)
1184840743 M * AStorm ah, it's still NAND
1184840757 J * fishingshrimp ~staenzer@static-87-79-237-223.netcologne.de
1184840768 M * Bertl blizz: I would not see how it could help Linux for example?
1184840775 M * Bertl welcome fishingshrimp!
1184840780 M * fishingshrimp hello bertl :)
1184840788 M * blizz well, does it read faster than a harddisk?
1184840796 M * blizz (without caching)
1184840816 M * Bertl given that recent laptop harddisks reach 25MB/s++
1184840878 M * Bertl I really doubt it :)
1184840893 M * fishingshrimp just found out about this IRC channel. I'm not a frequent IRC user but I'm playing around with vserver for a while now and still don't get the concept of disk limiting, especially what it is with inodes_total and reserved... have googled for an hour and read about inodes in general (and understood) - but how does the inode-setting relate to the space_total?
1184840902 M * AStorm Bertl, actually, my pendrive is faster than that
1184840910 M * AStorm than these 25 MB/s
1184840923 M * AStorm it does 32 MB/s 
1184840928 M * Bertl fishingshrimp: logon to your host, and do 'df' then 'df -i'
1184840934 M * AStorm and is faster than the hard disc
1184840962 M * Bertl AStorm: yeah, but 1GB vs 160GB?
1184840975 M * AStorm Bertl, 4GB vs 80 GB, but anyway :>
1184840990 M * blizz Bertl: maybe it could be used for swapping (hooray, that would work for a week. it should be toasted by then :-P)
1184840990 M * Bertl the Turbo thingy is 1GB (max)
1184840994 Q * HeinMueck Quit: Aah!
1184840999 M * AStorm Bertl, uhm? what's the speed?
1184841003 M * fishingshrimp Bertl: So the total space in GB for example is "divided" in Inodes? Do I get this right?
1184841017 M * AStorm Maybe it's a single-chip speed and can be sped up using raid (hint)
1184841028 M * Bertl fishingshrimp: well, it is not that simple, but basically yes
1184841043 M * Bertl fishingshrimp: every 'element' in the filesystem consumes an inode
1184841071 M * Bertl depending on the cluster and block size, the number of inodes relates to your filesystem/disk size
1184841089 M * fishingshrimp Bertl: I just read about those basics. So there can't be more inodes than files?
1184841091 M * Bertl (most filesystems allow you to tune the number somewhat)
1184841121 M * Bertl fishingshrimp: no, usually there will be more inodes than files, as directories consume inodes too
1184841131 M * blizz Bertl: IF the flash reads faster than the hdd it could be used for faster booting and suspend/hibernation
1184841132 M * AStorm some filesystems don't even have inodes
1184841141 M * blizz but i wouldnt use it as a hdd cache, i bet it won't last long enough
1184841144 M * Bertl blizz: 1GB?
1184841150 M * AStorm blizz, actually, totally wrong
1184841156 M * AStorm it will last longer than the HDD :>
1184841156 M * blizz AStorm: tell me more, please
1184841168 M * AStorm and will decrease seek times dramatically
1184841168 M * fishingshrimp Bertl: Ah okay, but I still don't understand how many Inodes I have to set a guest to related to the total space. Do I have to set the inodes?
1184841181 M * AStorm (for cached data, of course)
1184841191 M * AStorm it's mostly useless for non-seeky loads
1184841204 M * AStorm (As hdd is most likely faster)
1184841219 M * Bertl fishingshrimp: well, you probably want to protect that resource too, otherwise an evil guest could DoS you by using up all the inodes
1184841271 M * blizz AStorm: nand flash survives more write cycles than a harddisk?
1184841277 M * fishingshrimp Bertl: So there is no "easy script" where I say "this guest should have 10 gb" and the script calculates the other value based on my harddisk etc?
1184841281 M * AStorm blizz, some kinds do, yes
1184841291 M * AStorm especially newest tech
1184841298 M * Bertl blizz: especially for broken harddisks :)
1184841312 M * AStorm new hard disks are much less resilient...
1184841325 M * AStorm density comes with a price (not only $$$)
1184841336 M * Bertl typical nand flash can survive a month of continuous writes
1184841341 M * blizz AStorm: sorry then, i only had the old tech in mind (nand flash == very few write cycles)
1184841393 M * blizz Bertl: i bet it's possible to do _something_ usefull with 1gb of free nand flash :-)
1184841394 M * Pazzo (a little bit off-topic, but as you're talking about broken disks, filesystems etc...: is someone here using a scalable, fast, reliable and fault-safe cluster filesystem? (and satisfied!)
1184841413 M * Bertl blizz: it's free? please send me one then :)
1184841444 M * AStorm I could use that too.
1184841456 M * blizz Bertl: after comparing prices of recent notebooks it seems like it comes for free *lol*
1184841472 M * AStorm blizz, uhm, CompactFlash isn't free at all
1184841487 M * blizz i understand your point
1184841520 M * blizz but i could not see a price-hike between comparable non-turbomemory and turbomemory notebooks
1184841526 M * AStorm should the card reader be faster, it could be a very viable medium
1184841555 M * AStorm blizz, right, and I still prefer my RAID 0 to some turbomemory hacks :P
1184841565 M * AStorm though I lose the seek time advantage
1184841567 M * blizz afaik the intel stuff is 1gb of embedded nand flash
1184841670 M * Bertl most of those 'new' technologies are papering over software issues ... which is kind of funny, as originally the software was used to fix/work around hardware issues :)
1184841676 M * blizz "20% faster boot up times in windows vista" hooray for worthless benchmarks
1184841724 M * AStorm :>
1184841729 M * Bertl recently I installed a Core2 Duo workstation, and it was really hard to get both CPUs working at 100% (with workstation kind of loads)
1184841744 M * AStorm I could boot linux faster too if I dumped the cache on disk
1184841760 M * AStorm and e.g. ondemand-readahead
1184841762 M * Bertl I had to run Xaos 3 times, glxgears and some other things ...
1184841769 M * blizz *lol*
1184841772 M * AStorm unfortunately, fscache is not there yet
1184841784 M * AStorm Bertl, uhm, it's quite simple
1184841790 M * AStorm get drive encryption
1184841794 M * Bertl other folks simply install vista(tm/R/C) and the cpu is gone
1184841796 M * AStorm play some audio, compile something too
1184841799 M * AStorm :>
1184841820 M * AStorm esp. compilation is nicely multiprocessed
1184841830 M * blizz -j1337 [tm]
1184841835 M * AStorm blizz, too much :>
1184841839 M * AStorm 2*cores + 1
1184841841 M * AStorm at most
1184841859 M * AStorm otherwise you're just overloading the scheduler
1184841872 M * Bertl doesn't matter, -j99 (max) works as well as -j 5
1184841882 M * AStorm Bertl, max is -j
1184841888 M * AStorm :>
1184841900 M * blizz (most of the time i'm kidding with what im saying)
1184842100 Q * Aiken Quit: Leaving
1184842112 M * AStorm Bertl, well, RAID 0 does almost add the performance of the drives
1184842119 M * AStorm even more so, because flash is seekless
1184842139 M * AStorm (~30MB/s + ~30 MB/s -> ~52 MB/s)
1184842163 M * AStorm for small files it's even better
1184842196 M * AStorm I expect the hard drive to fail earlier than that flash :P
1184842198 M * blizz how much do you pay for a 2G usb stick?
1184842199 M * Bertl well, 190MB/s here on raid5 (disks)
1184842210 M * AStorm Bertl, uhm, laptop, different scale :P
1184842232 M * AStorm blizz, it depends, fast ones are still quite expensive
1184842240 M * Bertl okay, let's wait for the server USB sticks then :)
1184842260 M * AStorm Bertl, there are flash disk drives already
1184842264 M * blizz i payed 20 euros for a cheapo stick (BASF i think)
1184842266 M * AStorm cost bit $$$
1184842270 M * AStorm *big
1184842287 M * blizz ahh, i saw a company manufactering these
1184842289 M * AStorm like 1200$/40 GB
1184842299 M * AStorm are about 40 MB/s
1184842308 M * AStorm though 0 seek time
1184842312 M * AStorm and no moving parts
1184842323 M * AStorm I'm actually baffled they cost so much
1184842342 M * AStorm should be less expensive than individual pendrives of the same capacity
1184842347 M * AStorm and it's almost the same
1184842350 M * Bertl together with a bunch of pci bridges, they will almost get the performance of a real storage appliance :)
1184842382 M * AStorm Bertl, heh, USB fwit
1184842409 M * Bertl hmm, that would be limiting the bandwidth ...
1184842447 M * AStorm not really, usb controllers are cheap
1184842471 M * AStorm though in a custom solution, you can skip that and add sata instead
1184842497 M * Bertl why would you use a mediation protocol for memory ...
1184842502 M * AStorm even do hardware raid on the chips, if you have the cash
1184842507 M * AStorm Bertl, I wouldn't :>
1184842835 J * HeinMueck ~Miranda@dslb-088-064-021-019.pools.arcor-ip.net
1184842846 Q * Baby Quit: bye!
1184842965 M * AStorm Hmm, does any of you know whether linux can use LVM2 root w/o initramfs?
1184843168 M * AStorm Well, I'll need the initramfs anyway
1184843231 M * Bertl no, it cannot, as there is no LVM2 kernel side setup
1184843634 M * AStorm Unfortunately. And doing LVM2 would mean the initramfs has to support hibernation resume too
1184844178 J * lilalinux_ ~plasma@dslb-084-058-248-099.pools.arcor-ip.net
1184844536 Q * lilalinux Ping timeout: 480 seconds
1184845051 Q * FireEgl Read error: Connection reset by peer
1184845871 J * pmenier ~pmenier@LNeuilly-152-22-72-5.w193-251.abo.wanadoo.fr
1184846036 J * FireEgl FireEgl@Atlantica.Inet6.Info
1184846061 M * Bertl dilinger: http://vserver.13thfloor.at/Stuff/OLPC/delta-OLPC.0.4.2-0.4.3.diff
1184846089 M * Bertl dilinger: together with the 4.1-4.2 update, you should be fine now
1184846179 M * Roey hey all!
1184847145 Q * ex Ping timeout: 480 seconds
1184847189 M * cluk Bertl: I have applied the patch I posted earlier to the debian kernel source. 
1184847214 M * cluk sched_hard is working as expected now
1184847320 J * ex ex@valis.net.pl
1184847508 M * Bertl welcome ex!
1184847529 M * Bertl cluk: good :) but I would suggest (for debian) to update to more recent versions
1184847596 M * cluk ok. but this patch would apply to the latest vserver release also or did i miss something?
1184847612 M * Bertl no, that patch won't apply there
1184847624 M * Bertl because the issues doesn't exist there at all
1184847744 M * cluk you think it is still debian specific?
1184847759 M * Bertl yes, this _is_ a debian issue
1184847776 M * Bertl and it would not exist, if they had updated their patches with newer mainline releases
1184848160 M * sid3windr Bertl: vserver on OLPC? :)
1184848168 M * cluk ok, I will file a debian bug report. thanks very much for your support.
1184848174 M * Bertl sid3windr: indeed ...
1184848176 M * sid3windr cool
1184848290 Q * kaner_ Remote host closed the connection
1184848646 J * cruser ~chatzilla@72.242.194.162
1184848993 M * Roey Bertl:  is it just XFS that has issues with VServer then, or is it LVM2 as well?
1184849041 M * AStorm Roey, XFS has no issues yet
1184849049 M * AStorm it will have them in 2.6.23
1184849068 M * Bertl xfs already _has_ issues
1184849068 M * AStorm you'll have to migrate IUNLINK and BARRIER flags
1184849072 M * AStorm Bertl, ?
1184849182 M * AStorm Bertl, what are them?
1184849281 M * Roey Bertl:  hmm.  I hope yesterday's talk on #xfs bore some fruit
1184849291 M * Roey that I can eat
1184849293 M * Roey as a user
1184849394 Q * Piet Read error: Connection reset by peer
1184849443 J * Piet hiddenserv@tor.noreply.org
1184849493 M * Bertl AStorm: http://vserver.13thfloor.at/Stuff/BUGHUNT/bertl-0004/
1184849524 M * AStorm ugh
1184849530 M * AStorm smells of a deadlock
1184849568 M * Bertl no, it is a circular lock :)
1184849595 M * AStorm which is the same, but detected by the kernel? :>
1184849652 M * Bertl the interesting part is, that it doesn't happen where we would expect it :)
1184850182 M * cruser Hi. Does anyone know if the samba "smb" daemon can run on more than one vguest? I found out that both vhost and vguest can not run a smb daemon when the vguest smb daemon stopped working. 
1184850286 M * Bertl cruser: hmm, I would say you are jumping to conclusions here
1184850299 M * cruser :)
1184850316 M * Bertl cruser: most likely, your guest smb daemon stops working because the host daemon already binds all IPs
1184850335 M * Bertl cruser: restricting the host daemon probably would solve that issues
1184850359 M * Bertl cruser: also running the daemon in two guests (not sharing a common IP) will work
1184850385 M * Bertl (you do not need to restrict it there, as this is done by the IP isolation)
1184850387 M * cruser Sounds good. The error was saying that 0.0.0.0 was already taken
1184850400 M * Bertl see, you just confirmed my theory :)
1184850505 M * cruser Thanks. Where do I need to look about restricting the host daemon to binding to an IP.
1184850552 M * Bertl you have two options there: 1) in the smb config (if that is supported) and 2) by using the network isolation (either via chbind or via the v_* wrappers)
1184850629 M * cruser Thanks!
1184850639 M * Bertl you're welcome!
1184851181 M * harry top - 15:19:33 up 105 days, 21:33,  1 user,  load average: 1.04, 0.85, 0.66
1184851181 M * harry Tasks: 461 total,   3 running, 457 sleeping,   0 stopped,   1 zombie
1184851184 M * harry pretty neat! :)
1184851205 M * harry the only vserver with considerable load :)
1184851310 Q * AStorm Quit: Bye
1184851502 Q * cluk Quit: Ex-Chat
1184851538 M * Bertl okay, nap attack ... back later ...
1184851545 N * Bertl Bertl_zZ
1184851576 M * harry hehe
1184851576 M * harry hf
1184851894 M * matti Hi Bertl_zZ :)
1184851918 J * chand ~chand@212.99.51.254
1184852287 P * cruser 
1184853546 J * lilalinux__ ~plasma@dslb-084-058-233-161.pools.arcor-ip.net
1184853896 J * Erik123 root@dudesofknowledge.nl
1184853941 Q * lilalinux_ Ping timeout: 480 seconds
1184853959 M * Erik123 Hello
1184854021 M * Erik123 i'm currently having the following problem with vserver: can't reach the net from any vserver :(
1184854026 M * Erik123 tips and idea's are welcome ;)
1184854096 M * Erik123 running debian/etch (2.6.22) and  vs2.2.0.2-rc1 patch
1184854199 Q * Piet Quit: Piet
1184854242 M * Erik123 oh i'm lying its vs2.2.0.1-rc1
1184854243 M * Erik123 hmm
1184854262 J * lilalinux_ ~plasma@dslb-084-058-247-084.pools.arcor-ip.net
1184854514 Q * Erik123 Quit: trying ! :)
1184854635 Q * lilalinux__ Ping timeout: 480 seconds
1184855220 Q * fishingshrimp Quit:  HydraIRC -> http://www.hydrairc.com <- *I* use it, so it must be good!
1184855978 Q * Solaris Read error: Connection reset by peer
1184858219 Q * Pazzo Quit: ...
1184859568 J * Solaris ~satan@85.138.105.7
1184860587 M * coderanger Question, what is fill_rate vs. fill_rate2 in the CPU scheduler struct?
1184860801 M * daniel_hozac regular value and idle-time value.
1184860833 M * daniel_hozac whenever the CPU goes idle, if idle-time advancing is enabled, the scheduler switches to fill_rate2/interval2
1184860851 M * coderanger thanky
1184860893 M * coderanger Are the default schedule params listed anywhere so we can use them as a starting point?
1184860940 M * daniel_hozac if you want the defaults, why set it at all?
1184860949 N * ensc Guest761
1184860949 M * daniel_hozac but yes, just start up a context, cat /proc/virtual/<xid>/sched.
1184860959 J * ensc ~irc-ensc@p54B4E4EA.dip.t-dialin.net
1184860974 M * coderanger daniel_hozac: I just want to see what "reasonable" values look like
1184860981 M * daniel_hozac kernel/vserver/sched_init.h
1184860993 M * daniel_hozac "reasonable" values depend on what you want to achieve.
1184861070 Q * Guest761 Ping timeout: 480 seconds
1184861070 M * coderanger Toggle between the default, and 10% of the default, so thats exactly what I needed :)
1184861110 M * daniel_hozac hmm, you're aware the default is 25%, right?
1184861129 M * coderanger Yes, this is just for a demo
1184861146 M * daniel_hozac okay then.
1184861279 M * daniel_hozac note that if you want to show that the limiting works, you'll not want to enable idle-time (unless you'll simultaneously be running something CPU-intensive in a non-10% guest).
1184861332 M * coderanger Indeed, it doesn't make sense for us in general
1184861354 J * stefani ~stefani@flute.radonc.washington.edu
1184861356 M * daniel_hozac well, it's primarily meant for fair sharing.
1184861381 M * coderanger We don't want to play fair though, once an app is backgrounded it doesn't get much CPU
1184861412 M * daniel_hozac but does it make sense to limit the foreground app to 25%?
1184861439 M * daniel_hozac (yes, i realize that'll change)
1184861440 M * coderanger No, like I said this is being thrown together for a demo tomorrow morning
1184861474 M * daniel_hozac but what's the harm of letting something in the background use CPU if it's not being used for anything else?
1184861494 M * coderanger Because CPU usage == heat and power
1184861509 M * daniel_hozac okay, good point.
1184861548 M * daniel_hozac does spreading it out over a longer period of time decrease the total amount of heat generated and power consumed though?
1184861618 M * coderanger No, this isn't for delaying something CPU intensive, more like a speed-limit on unused stuff
1184861630 M * coderanger So they can't sit in a tight loop and kill the machine
1184861765 Q * fosco Ping timeout: 480 seconds
1184861821 M * daniel_hozac yeah, okay.
1184861841 M * coderanger Any particular reason that libvserver doesn't wrap the getter functions for the sched data?
1184861856 J * fosco fosco@konoha.devnullteam.org
1184861878 M * daniel_hozac humm. i thought i added all of those.
1184861981 M * coderanger I just see get_shed and get_shed_info
1184861988 M * coderanger er, sched
1184862035 M * coderanger No mapping in libvserver though
1184862060 M * coderanger Ahh, sched_info is wrapper, but not get_sched
1184862128 M * daniel_hozac yeah, sched_info is used by vserver-stat.
1184862133 M * coderanger I really can't type this morning
1184862136 M * daniel_hozac i thought i added get_sched at the same time.
1184862216 M * daniel_hozac ah, i see. get_sched is V5.
1184862271 M * daniel_hozac i didn't use v5 in util-vserver because 2.1.1 only has v4.
1184862276 M * daniel_hozac (IIRC)
1184862472 J * bonbons ~bonbons@2001:5c0:85e2:0:20b:5dff:fec7:6b33
1184862514 M * daniel_hozac bonbons: ping
1184862535 M * bonbons daniel_hozac: pong
1184862566 M * daniel_hozac ::1 isn't handled in current IPv6 patches?
1184862605 M * bonbons there is no special handling for
1184862633 M * bonbons it's just like any address - so no spoofed ::1 for guests
1184862720 M * daniel_hozac and no rewriting.
1184862745 M * daniel_hozac how is it supposed to be handled?
1184862809 M * daniel_hozac i was somewhat surprised when nc -6 ::1 6000 returned "Cannot assign requested address" :)
1184863035 M * bonbons for now it's handled like any other address, either allowed or not. A way to handle it would be to map ::1 to something in ::/64 area or tag everything on ::1 per-context
1184863078 M * bonbons did not really follow what has finally been to for IPv4 127.0.0.1 ...
1184863120 M * daniel_hozac in 2.3, each guest gets its own 127.<xid>.1 which is rewritten to 127.0.0.1 and the other way around.
1184863123 M * daniel_hozac (or at least, it's supposed to)
1184863136 M * daniel_hozac so the mapping thing sounds good.
1184863235 M * daniel_hozac is the entire ::/64 reserved for host-local things though?
1184863244 M * bonbons does the socket know the rewritten part?
1184863259 M * daniel_hozac in 2.3, ideally no.
1184863270 M * daniel_hozac i don't know how far along it is yet though.
1184863304 Q * pmenier Quit: pmenier
1184863321 M * bonbons if it was known it would make ip address listings on the host... overfilled
1184863335 M * daniel_hozac what?
1184863371 M * bonbons in IPv6 /64 is the smalles subnet permitted
1184863381 M * daniel_hozac yeah.
1184863421 M * bonbons well, if for each guest you see 127.<xid>.1 in addition to all the others and the host can communicate with that address it would be problematic
1184863459 M * daniel_hozac i guess it might make sense to hide those from the host, assuming the kernel is the one who adds them.
1184863472 M * daniel_hozac but i don't think that's how it's going to work.
1184863632 M * bonbons too many places in kernel that work with ip addresses... whole iptables stack and of curse IP stack
1184863724 M * bonbons the cleaner would probably be to add context-tags to all packets (and allow collisions only for a few pre-defined addresses)
1184863762 J * lilalinux__ ~plasma@dslb-084-058-216-059.pools.arcor-ip.net
1184863956 J * lilalinux ~plasma@dslb-084-058-247-130.pools.arcor-ip.net
1184864131 Q * lilalinux_ Ping timeout: 480 seconds
1184864281 Q * lilalinux__ Ping timeout: 480 seconds
1184864417 N * Bertl_zZ Bertl_oO
1184864740 J * DavidS ~david@217.19.46.18
1184864837 M * sannes hm, when I use vtop I see two processes called vcontext with a pid of other processes, anyone experienced that?
1184864897 J * lilalinux_ ~plasma@dslb-084-058-223-013.pools.arcor-ip.net
1184864900 M * Bertl_oO sannes: yes, IIRC, folks using older util-vserver experienced that
1184864993 M * sannes hm, using 0.30.212
1184865167 M * daniel_hozac with a pid of other processes?
1184865176 M * daniel_hozac you mean you have duplicate pids on your system?
1184865274 Q * lilalinux Ping timeout: 480 seconds
1184865376 M * Bertl_oO daniel_hozac: btw, I upload a testfs.sh-0.16 version with some minor changes to your patches
1184865391 M * Bertl_oO daniel_hozac: and fixed the 'known' issues including the jfs one :)
1184865407 M * Bertl_oO the following remains:
1184865426 J * lilalinux__ ~plasma@dslb-084-058-238-075.pools.arcor-ip.net
1184865453 M * Bertl_oO - xfs circular lock (comment on the xfs channel: most of those traces are false positives) -> ignored for now
1184865506 M * Bertl_oO - dlimit deviations (can be shown with -vv) for reiser/jfs (xfs doesn't implement dlimits)
1184865571 M * Bertl_oO not sure those are actually reporting bugs or just the filesystem specific way to (pre)allocate structures (i.e. we have to check that in detail)
1184865604 M * daniel_hozac yeah...
1184865634 Q * lilalinux_ Ping timeout: 480 seconds
1184865646 M * Bertl_oO ah, and yes, 2.6.22.1 was triggering an oops (incorrect refcounting for nsproxy :)
1184865684 M * Bertl_oO daniel_hozac: ah, and one more ... libvserver doesn't support latest ABI it seems
1184865701 M * daniel_hozac yeah, it's intentional (for vc_set_sched).
1184865728 M * Bertl_oO okay, that struck OLPC folks yesterday, as they do not have the V4 anymore :)
1184865735 M * daniel_hozac yeah, i saw.
1184865749 M * daniel_hozac re: testfs, i actually fixed some more things today.
1184865774 M * Bertl_oO sounds good
1184865779 M * daniel_hozac or rather, extended the do_xattr_cowbreak to use cmp.
1184865881 Q * dg_guru Remote host closed the connection
1184865904 M * Bertl_oO daniel_hozac: princeton should be up-to-date for tests :)
1184866096 M * daniel_hozac are the sleep 10's in do_dlimit_remove really necessary?
1184866110 M * Bertl_oO nah, those are accidential :)
1184866118 M * daniel_hozac hehe, okay.
1184866119 M * Bertl_oO i.e. debug left-overs :)
1184866124 M * Bertl_oO let me fix them in place
1184866175 M * Bertl_oO okay, done
1184866190 M * daniel_hozac btw, what do you think about my conversation with bonbons?
1184866207 M * Bertl_oO didn't read up on that yet (too hot here today :)
1184866213 M * daniel_hozac heh, okay, np.
1184866228 M * daniel_hozac http://people.linux-vserver.org/~dhozac/p/m/delta-cowcmp-feat01.diff is the patch for cmp.
1184866251 M * daniel_hozac /dev/sda3 is ready to be trashed?
1184866259 M * Bertl_oO yes, should be fine
1184866262 M * daniel_hozac okay.
1184866442 M * daniel_hozac hmm, looks like i added the cowbreak test to the wrong section. that should be 135, no?
1184866481 Q * DavidS Ping timeout: 480 seconds
1184866521 M * daniel_hozac (rather than 125)
1184866548 M * daniel_hozac it seems to work fine though, xfs fails that test.
1184866699 J * christianna` WatchMe@87-97-209-13.2072807291.ddns.evrocom.net
1184866707 M * Bertl_oO welcome christianna`!
1184867260 J * kaner kaner@strace.org
1184867284 Q * chand Quit: chand
1184867295 T * Bertl_oO http://linux-vserver.org/ | latest stable 2.2.0.3, 2.0.3-rc3, devel 2.3.0.12, stable+grsec 2.0.2.1, 2.2.0.2 | util-vserver-0.30.213 | 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 ;)
1184867321 M * Bertl_oO 2.2.0.3 (19/20/21) and rc1 for 22
1184867352 M * Bertl_oO let me know what I forgot to add :)
1184867360 M * Bertl_oO off for now .. back later I guess
1184867419 Q * zLinux Ping timeout: 480 seconds
1184867499 Q * HeinMueck Ping timeout: 480 seconds
1184867791 J * Ganneff ~joerg@ganneff.noc.oftc.net
1184867820 Q * christianna` autokilled: This host violated network policy. If you feel an error has been made, please contact support@oftc.net, thanks. (2007-07-19 17:57:00)
1184867826 P * Ganneff 
1184868062 Q * ema Quit: leaving
1184868068 J * DavidS ~david@217.19.46.18
1184868880 J * AStorm ~astralsto@host-81-190-179-124.gorzow.mm.pl
1184868958 J * zLinux ~zLinux@88.213.33.64
1184869752 Q * AStorm Quit: Bye
1184869850 Q * cedric Quit: cedric
1184870255 Q * DavidS Quit: Leaving.
1184870374 Q * lilalinux__ Remote host closed the connection
1184873044 J * AStorm ~astralsto@host-81-190-179-124.gorzow.mm.pl
1184873314 J * WorkingGeier ~richter@mnch-4db1e172.pool.einsundeins.de
1184873319 M * WorkingGeier hi
1184873326 M * daniel_hozac hello
1184873335 M * WorkingGeier I'm trying to use vserver as "chroot on steroids"
1184873385 M * WorkingGeier i.e. I have an installation that does not have a proper /sbin/init and all the daemons
1184873394 M * WorkingGeier so normally no processes are left running
1184873408 M * daniel_hozac so what's it used for?
1184873420 M * WorkingGeier "vserver enter" complains that because there is no process, the vserver is not running, hence I cannot enter
1184873429 M * daniel_hozac right.
1184873431 M * WorkingGeier building software packages on demand
1184873433 M * daniel_hozac as it should.
1184873478 M * WorkingGeier i.e. when the outside system decides something should be done, it should spawn a new process chrooted into the appropriate directory and with the right context
1184873484 M * WorkingGeier how would I do that?
1184873487 M * daniel_hozac you'll want to use the persistent flags.
1184873524 M * daniel_hozac i.e. put persistent in /etc/vservers/<guest>/flags and /etc/vservers/<guest>/nflags
1184873547 M * WorkingGeier then "start" would mark it as running, even if it does nothing?
1184873569 M * daniel_hozac yep.
1184873586 M * WorkingGeier (so all I'd have to do was to make sure that "start" does nothing, and "stop" kills all processes with that context
1184873611 M * WorkingGeier since there is no proper init inside the vservers)
1184873620 M * daniel_hozac there usually isn't.
1184873637 M * daniel_hozac the default initstyle, sysv, doesn't run an init.
1184873659 M * WorkingGeier but rather runs the init scripts directly?
1184873665 M * daniel_hozac yep.
1184873668 M * WorkingGeier hmm
1184873698 M * WorkingGeier there are init scripts, since the chroot contains unpacked Debian packages
1184873725 M * harry another update??????
1184873747 M * daniel_hozac yep.
1184873753 M * WorkingGeier but I don't want them to run
1184873806 M * daniel_hozac set apps/init/cmd.start to /bin/true then.
1184873870 M * WorkingGeier that is per-vserver?
1184873894 M * daniel_hozac yep.
1184873896 M * WorkingGeier cmd.stop similar?
1184873903 M * daniel_hozac yep.
1184873961 M * WorkingGeier or should that be "vkill -c 12345 -s TERM -1"?
1184873962 J * marcfiu ~mef@aegis.CS.Princeton.EDU
1184874074 Q * AStorm Quit: Bye
1184875738 J * AStorm ~astralsto@host-81-190-179-124.gorzow.mm.pl
1184877006 Q * FireEgl Quit: Bye...
1184877211 Q * meandtheshell Quit: Leaving.
1184878002 Q * bonbons Quit: Leaving
1184879232 J * Aiken ~james@ppp121-45-220-241.lns2.bne1.internode.on.net
1184879698 J * DK20563 ~DK20563@c-a7c7e455.43-40-64736c10.cust.bredbandsbolaget.se
1184879843 J * bzed ~bzed@dslb-084-059-102-103.pools.arcor-ip.net
1184880478 N * DK20563 atterdag
1184880872 J * rgl ~Rui@84.90.10.107
1184880878 A * rgl waves
1184880949 M * AStorm The weird raid works
1184880959 M * AStorm I had to add usb-storage.delay_use=0 option
1184880966 M * AStorm so as to avoid creating an initrd
1184881050 M * AStorm *initramfs
1184881159 M * rgl usb raid?
1184881334 M * AStorm rgl, yep, with pendrive
1184881345 M * AStorm http://wklej.org/id/fd5d3ee727
1184881360 M * AStorm stupid slow laptop hdd :>
1184881391 M * AStorm in real usage that RAID is even faster - less seeks and better random access
1184881550 M * rgl flash pendrive?
1184881584 M * AStorm Yep.
1184881596 M * AStorm one of the best available currently
1184881757 Q * AStorm Quit: Bye
1184881778 J * meandtheshell ~markus@85.127.116.18
1184881860 Q * meandtheshell 
1184882095 J * AStorm ~astralsto@host-81-190-179-124.gorzow.mm.pl
1184882175 Q * AStorm 
1184882200 J * AStorm ~astralsto@host-81-190-179-124.gorzow.mm.pl
1184882204 M * rgl which brand/model AStorm ?  is that better than mechanical SATA drives?
1184882235 M * AStorm Probably not
1184882242 M * AStorm it's a pendrive, not USB flash disk
1184882248 M * AStorm Patriot Xporter XT 4GB
1184882289 J * DK20563_ ~DK20563@c-a7c7e455.43-40-64736c10.cust.bredbandsbolaget.se
1184882324 Q * atterdag Read error: Connection reset by peer
1184882411 Q * dna Quit: Verlassend
1184882418 N * DK20563_ atterdag
1184882493 Q * AStorm 
1184882541 Q * WorkingGeier Quit: leaving
1184882795 Q * marcfiu Quit: Leaving.
1184882812 J * AStorm ~astralsto@host-81-190-179-124.gorzow.mm.pl
1184883813 P * stefani I'm Parting (the water)
1184885604 Q * rgl Ping timeout: 480 seconds
1184886808 Q * arachnist Read error: Connection reset by peer
1184887086 N * ensc Guest780
1184887096 J * ensc ~irc-ensc@p54B4E4EA.dip.t-dialin.net
1184887189 Q * Guest780 Ping timeout: 480 seconds
1184887198 Q * atterdag Quit: Leaving
1184887508 Q * fosco Remote host closed the connection
1184887651 J * fosco fosco@konoha.devnullteam.org