1258156987 Q * imcsk8 Quit: Leaving
1258158868 M * Bertl_oO smash: what's the problem ... working fine here
1258158874 N * Bertl_oO Bertl
1258160021 Q * yarihm Quit: This computer has gone to sleep
1258160903 Q * dowdle Remote host closed the connection
1258161806 Q * jrklein Remote host closed the connection
1258162531 M * smash Bertl: weird as I said above I cant get the binary to work .. grub reports back the binary is in an unknown or unsupported format .. but 'file'ing it does not help .. it seems to be correct
1258162546 M * Bertl binary means?
1258162554 M * smash the kernel binary
1258162588 M * Bertl where does that come from?
1258162594 M * smash I compiled the kernel?
1258162615 M * smash vmlinuz-2.6.31.6-vs2.3.0.36.24: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, not stripped
1258162618 M * smash looks just fine to me
1258162631 M * Bertl okay, your machine is 64bit?
1258162661 M * smash yea
1258162673 M * smash running squeeze amd64 right now (2.6.30)
1258162675 M * Bertl and grub says, unknown format?
1258162680 M * smash yea
1258162708 M * Bertl sounds to me like a problem with your grub, but can you upload the kernel somewher?
1258162713 M * Bertl *somewhere
1258162725 M * smash sure 2secs
1258162955 M * smash gave you the links in private tho
1258163077 M * Bertl yep, your kernel is definitely broken
1258163113 M * Bertl i.e. I'd change your toolchain, maybe your distro is being funny with gcc or binutils again :)
1258163144 M * Bertl try to switch to an older gcc/binutils and recompile
1258163150 M * smash yea i thought of the same
1258163153 M * smash screw squeeze
1258163154 M * smash !
1258163195 M * Bertl but the size looks really odd too
1258163207 M * Bertl are you sure you got the right one from the buildtree?
1258163222 M * Bertl i.e. what file did you copy?
1258163225 M * smash but before I'll go for gcc 3.x instead of 4 .. but thanks for reconfirming it (I had to possibility right now) and good to know its not the experimental branch that fails here
1258163264 M * smash I used kernel-package thus make-kpkg .. usually I'd go for arch/ and copy it manually and not use the one in the main src dir
1258163286 M * Bertl okay, because it looks like the unstripped debug version to me
1258163289 M * smash I also discovered a bug in kernel-package (going to slap manoj later on)
1258163309 M * Bertl compare it to the bzImage in arch/x86/
1258163459 M * smash I guess make-kpkg moves it away or something
1258163527 M * Bertl JFYI: file arch/x86/boot/bzImage
1258163530 M * Bertl arch/x86/boot/bzImage: Linux kernel x86 boot executable RO-rootFS, root_dev 0x902, swap_dev 0x3, Normal VGA
1258163533 M * smash damnit I should have compiled it by hand .. screw all that automatisms
1258163535 M * smash I checked that
1258163557 M * smash thats why I said it moved it ${somewhere}
1258163577 M * Bertl while: file vmlinux
1258163577 M * Bertl vmlinux: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), statically linked, not stripped
1258163594 M * smash damnit I should have compiled it by hand .. screw all that automatisms .. I wish we were back in good old 2.2 times ;)
1258163663 M * smash theres no kernel binary in the build tree .. gonna compile it by hand then .. should have done so in the first place
1258163685 M * Bertl yeah, it's the suggested way for Linux-VServer anyways
1258163762 M * smash yea well as you know .. time will make you lazy :P .. I am using linux since suse 6.0 which is early 1999
1258163775 M * smash with 2.0.36 afaik and later on 2.2.x
1258163792 M * smash I just wanted to spare me some time .. issueing one command, get some coffee and let the 4 xeons do their job ;)
1258163871 M * Bertl never trust the xeons :)
1258164261 M * smash haha
1258164262 M * smash ;)
1258165315 M * Bertl off to bed now ... have a good one everyone!
1258165329 N * Bertl Bertl_zZ
1258167631 J * aj__ ~aj@e180194173.adsl.alicedsl.de
1258168061 Q * derjohn_foo Ping timeout: 480 seconds
1258169490 J * derjohn_foo ~aj@e180193060.adsl.alicedsl.de
1258169591 J * yepsen ~yepsen@pdbn-5d827bbe.pool.mediaWays.net
1258169593 Q * yepsen\rauchen Ping timeout: 480 seconds
1258169862 Q * aj__ Ping timeout: 480 seconds
1258171118 J * saulus_ ~saulus@c193029.adsl.hansenet.de
1258171527 Q * SauLus Ping timeout: 480 seconds
1258171534 N * saulus_ SauLus
1258175836 J * balbir ~balbir@122.172.27.189
1258184285 Q * hparker Quit: Read error: 104 (Peer reset by connection)
1258184870 J * grharry ~root@ppp-94-65-215-226.home.otenet.gr
1258184946 P * grharry 
1258189765 J * grharry ~root@ppp-94-65-227-51.home.otenet.gr
1258189906 M * grharry hi, I am running debian-lenny vserver 2.6.26-2 amd64, recently  when ever I enter a vhost and do a ps ax I get (  vxW: [�ps�,6020:#70|70|70] did lookup hidden ffff81007c58e050[#0,2] �/dev/pts�. ) in my logs !! Any clue ?? 
1258190467 M * zbyniu grharry: it's normal if you run vserver <some> enter
1258190500 M * zbyniu and don't use well-known-broken 2.6.26-2 from debian ;)
1258190599 J * ghislainocfs2 ~Ghislain@LPuteaux-151-41-11-129.w217-128.abo.wanadoo.fr
1258190841 M * grharry thanks, what kernel should I get ?? 
1258190885 J * yarihm ~yarihm@80-219-171-38.dclient.hispeed.ch
1258191486 Q * SauLus Quit: ...something weird happened
1258191672 J * bonbons ~bonbons@2001:960:7ab:0:2c0:9fff:fe2d:39d
1258191761 J * saulus ~saulus@c193029.adsl.hansenet.de
1258191816 J * zlb blizzard@internet-media-center.com
1258191984 Q * saulus 
1258191997 J * saulus ~saulus@c193029.adsl.hansenet.de
1258192282 P * grharry 
1258192815 Q * saulus Quit: ...something weird happened
1258192828 J * saulus ~saulus@c193029.adsl.hansenet.de
1258192887 Q * saulus 
1258192927 J * saulus ~saulus@c193029.adsl.hansenet.de
1258194902 Q * scientes Ping timeout: 480 seconds
1258195477 Q * kiorky Read error: Connection reset by peer
1258195508 J * kiorky ~kiorky@cryptelium.net
1258196090 Q * eyck Quit: leaving
1258196094 J * eyck ~eyck@nat06.nowanet.pl
1258196821 Q * ghislainocfs2 Quit: Leaving.
1258196906 J * ghislain ~Ghislain@LPuteaux-151-41-11-129.w217-128.abo.wanadoo.fr
1258198398 N * Bertl_zZ Bertl_oO
1258200224 J * AmokPaule ~amokpaule@brsg-4dbbb73e.pool.mediaWays.net
1258200812 J * saulus_ ~saulus@c193029.adsl.hansenet.de
1258200829 Q * saulus Quit: ...something weird happened
1258200837 N * saulus_ SauLus
1258203241 P * zlb 
1258203414 J * dna_ ~dna@175-195-103-86.dynamic.dsl.tng.de
1258204158 N * dna_ dna
1258205828 J * blues ~blues@cca131.neoplus.adsl.tpnet.pl
1258205947 Q * blues_ Ping timeout: 480 seconds
1258206564 J * agagag ~anton@195.169.149.112
1258206658 M * agagag hello, one of my vservers has no name anymore in vserver-stats and it seems it has become a zombie , i can not restart or kill it
1258206723 M * agagag stopping all the vserver with the util-vserver init.d script and the zombie still there
1258206924 M * agagag also tried vkill --xid 50 with no result 
1258206961 M * theocrite I had the same issue a while ago.
1258206968 M * theocrite agagag: are you using Debian ?
1258207101 M * agagag theocrite: yes. it seems that echo 50 > /var/run/vservers/bonzai fixed it though!
1258207118 M * agagag but it is a mystery why it happpened
1258207137 J * manana mayday_713@84.17.25.144
1258207367 M * theocrite That's probably a Debian related bug unfortunately. But the fix is good to know.
1258207378 Q * agagag Quit: Changing server
1258207422 M * theocrite Is there a troubleshooting page on the wiki or anywhere? I'm sure that's worth being accessible to anyone trying to find out throug a search engine.
1258208491 Q * DLange Quit: upgrading that crappy client we all use
1258208559 J * DLange ~DLange@dlange.user.oftc.net
1258209273 J * hparker ~hparker@208.4.188.201
1258210575 J * pmenier ~pmenier@ACaen-152-1-5-128.w83-115.abo.wanadoo.fr
1258212235 J * matthew-_ ~ms@ns2.wellquite.org
1258212907 N * Bertl_oO Bertl
1258212911 M * Bertl back now ...
1258213167 J * elops ~lhabs@77.42.225.169
1258213490 M * Bertl welcome elops!
1258213505 M * elops thnx
1258213511 M * elops do i kniw u sir?
1258213600 M * Bertl maybe, do you? :)
1258213656 M * elops erm no :)
1258213809 M * Bertl well, obviously I'm Bertl, so now you know me, right? :)
1258213842 Q * yarihm Quit: This computer has gone to sleep
1258213845 M * elops yea
1258213848 M * elops nice to meet youu
1258213850 M * elops =))))s
1258215024 M * theocrite So I guess there is no troubleshooting page as is. Maybe the "ghost guest" fix should be mentionned on the faq or anywhere else ?
1258215502 M * Bertl there is a debian specific page with 'debian specific problems' maybe put it there?
1258215974 J * Sirenia ~sirenia@93.186.164.51
1258215984 M * biz Bertl: https://bugs.launchpad.net/bugs/482292
1258216048 M * biz Bertl: obviously I showed him the linux-vserver wiki page with some of the needed fixes... (but there is more to be done)
1258216128 M * biz I'm not sure how to continue from here, I'd like to keep the "hardcoded" fixes to a minimum to do upgrades easily for a large number of guests... but it seems Ubuntu past "upstart" is not well suited anymore :(
1258216154 M * biz (... and not really willing to help either)
1258216218 M * biz *sigh*
1258216499 J * fLoo fLoo@188-193-148-129-dynip.superkabel.de
1258216595 M * Bertl well, we had the upstart developer here, and IIRC, I still have a guest running upstart (as we figured exactly what needs to be done)
1258216627 M * Bertl it's all on the IRC log, if nobody related to ubuntu cares about it ... that's fine for me :)
1258216636 Q * opuk Ping timeout: 480 seconds
1258216644 P * elops 
1258217618 Q * AmokPaule Quit: Nettalk6 - www.ntalk.de
1258217625 J * AmokPaule ~amokpaule@brsg-4dbbb73e.pool.mediaWays.net
1258218343 M * Bertl nap attack ... bbl
1258218348 N * Bertl Bertl_zZ
1258218509 M * theocrite Bertl_zZ: I'm not sure it's 2.6.26 related, not am I sure that what I wrote is in a correct english, but at least it's here : http://linux-vserver.org/Installation_on_Debian#.22Ghosts.22_guests
1258218522 M * theocrite Sleep well.
1258223768 N * DoberMann[ZZZzzz] DoberMann[PullA]
1258224370 J * opuk ~kupo@80.67.6.156
1258225660 Q * yepsen 
1258227306 Q * infowolfe Ping timeout: 480 seconds
1258227346 M * biz Bertl_zZ: I've got around my problem by patching the upstart source, but to understand the problem better, could you give me a hint how (and when exactly) you fake "pid 1" for the init process within a guest?
1258227478 M * biz upstart's /sbin/init checks if it is pid 1, if not, it _replaces_ itself with telinit which throws an error because there is no dbus daemon / no init process that could handle the request
1258227542 M * biz I've commented out the check for pid 1 with an additional commandline option, so now I start the vserver like "vserver foo start --rescue /sbin/init --fixme" to not execv() telinit but stay in place
1258227565 M * biz this works fine, init is there, and (I don't know when or how) actually has pid 1 after startup
1258227602 M * biz then I can enter the vserver and issue "telinit <myrunlevel>" and everything works fine because init listens for dbus events
1258227835 N * Bertl_zZ Bertl
1258227838 M * Bertl back now ...
1258227914 M * biz welcome back :D
1258227919 M * Bertl biz: assuming you are using the proper flags (and/or init syle), the process starting init becomes the init pid process, execing (i.e. overloading it's executable) the specified init (which then has pid 1) IIRC
1258227961 M * biz hrmm
1258227989 M * biz upstart's /sbin/init does   if (getpid () > 1)
1258228006 M * biz at this time it seems it is not pid 1 yet?
1258228016 M * Bertl that's how the 'normal' plain init style works, as almost any init assumes that it is started as pid == 1 nad if not it acts as telinit
1258228028 M * Bertl *and
1258228052 M * Bertl i.e. IIRC, upstart just mimics that for compatibility or so
1258228085 M * biz yeah I guess it's because of initscript or whatever to stay compatible with "init <runlevel>" calls
1258228132 M * Bertl what kernel/util-vserver version do you use?
1258228179 M * biz util-vserver 0.30.216~r2772-6 on 2.6.22.19-grsec2.1.11-vs2.2.0.7
1258228248 M * biz ew, I've just put a wrapper in place of /sbin/init and you're right, pid is 1
1258228274 M * biz let me debug this a bit more... :/
1258228410 M * Bertl not the best choice kernel/util wise but should work
1258228597 M * biz http://bazaar.launchpad.net/~scott/upstart/trunk/annotate/head%3A/init/main.c
1258228624 M * biz have at the block starting at line 132
1258228631 M * biz I'm not sure why, but it is executed
1258228794 M * Bertl how do you know?
1258228875 M * biz $ vserver foo start
1258228879 M * biz init called, arg = 3, ppid = 25657, pid = 1
1258228879 M * biz telinit called, ppid = 1, pid = 1
1258228879 M * biz telinit-: Failed to connect to socket /com/ubuntu/upstart: Connection refused
1258228910 M * Bertl and the pid/ppid output comes from?
1258228912 M * biz but you're right, it could also be some job called by the startup event
1258228935 M * biz #!/bin/sh
1258228940 M * biz echo "init called, arg = $@, ppid = $PPID, pid = $$"
1258228940 M * biz exec /sbin/init- "$@"
1258228945 M * biz (the same for telinit)
1258229423 J * ciprian_craciun ~ciprian@79.113.192.31
1258229448 M * Bertl could be your shell chaning pid on the exec
1258229525 M * ciprian_craciun Hello all!
1258229543 M * ciprian_craciun I have a question about the internals of vserver...
1258229604 M * Bertl let's hear ...
1258229619 M * ciprian_craciun Ok. What is the difference between vnamespace and vspace?
1258229642 M * ciprian_craciun From what I can deduce from the --help of both, vspace has more fine-grained control over the new namespace.
1258229646 M * ciprian_craciun Is that right?
1258229714 M * ciprian_craciun (Or better said what is the difference between `spaces` and `namespaces`, because in the `Paper` I couldn't find anything relating to `spaces`.)
1258229753 M * ciprian_craciun (I think I've botherd you Bertl also last week. Sorry :) )
1258229787 M * Bertl np, well, nowadays I think vspace and vnamespace are one and the same
1258229814 M * ciprian_craciun So I could just use vspace without vnamespace?
1258229817 M * Bertl historically we had a tool to create the filesystem namespace (long before any other 'spaces' were avail)
1258229829 M * Bertl *available
1258229847 M * Bertl nowadays, i.e. in recent kernels, there are many spaces
1258229873 M * Bertl ipc, uts, user, pid ... and of course the filesystem and mnt namespace
1258229929 M * ciprian_craciun Aha. Ok.
1258229976 M * ciprian_craciun Another one: in the kernel configuration there is an option for PID namespaces, but vserver does not select it. Is this on purpose? Will something bad happen if I enable it also? (Just out of curiosity.)
1258230022 M * Bertl by default, Linux-VServer uses pid isolation, which has some advantages over pid namespaces
1258230048 M * ciprian_craciun Like?
1258230057 M * ciprian_craciun (the advantages)
1258230071 M * Bertl like the ability to enter a guest, or run it without an actual init
1258230094 M * Bertl (sysv init syle)
1258230125 M * ciprian_craciun Aha. So if I don't need init, or have an init that doesn't care about beeing PID 1, I could use also PID namespaces.
1258230140 M * Bertl nah, the other way round
1258230158 M * Bertl i.e. if you want to run a guest without init, you need pid isolation not pid namespaces
1258230212 M * ciprian_craciun Ok. Got it.
1258230296 M * biz Bertl: wtf -- I wrote a small wrapper in C that shows me all pids, and guess what -- the pid of the /sbin/init process is 1. Now I've added this line to the upstart source, directly after if (getpid() > 1):
1258230300 M * biz printf("XXX pid is %d\n", getpid());
1258230305 M * biz q
1258230316 M * biz $ vserver foo start
1258230318 M * biz XXX pid is 5369
1258230318 M * biz init: Failed to connect to socket /com/ubuntu/upstart: Connection refused
1258230346 M * Bertl so getpid() returns different in your case?
1258230367 Q * ciprian_craciun Quit: leaving
1258230382 M * biz yes, getpid() in upstart returns not 1
1258230384 M * Bertl how does your C wrapper check the pid?
1258230426 M * biz # ... include stdio, sys/types.h, unistd.h
1258230427 M * biz int main(void) { printf("pid = %d\n", getpid()); return 0;
1258230427 M * biz }
1258230442 M * Bertl and that returns correclty, yes?
1258230448 M * biz yes, it returns 1
1258230521 M * biz too bad I can't ptrace /sbin/init, I'll add some debug prints to upstart
1258230527 M * biz it does something weird...
1258230544 M * Bertl very interesting ... in any case, try with a different kernel and recent util-vserver if possible (just to make sure)
1258230849 M * biz getpid() _directly_ after "int main(...) { " returns some high pid like 14912, not 1
1258230906 M * biz I start wondering what the fsck I'm missing :D
1258232814 M * biz HAH!
1258232926 M * Bertl enlighten us! :)
1258233010 M * biz I'm not 100% sure yet, it's quite hard to debug (I can't work with gdb in the vserver with grsec, and I'm not sure how to strace the creation of the context)
1258233026 M * biz But I guess.... this has something to do with it, from man 2 getpid:
1258233028 M * biz Since glibc version 2.3.4, the glibc wrapper function for getpid() caches PIDs, so as to avoid  additional  system  calls  when  a process calls getpid() repeatedly.  Normally this caching is invisible, but its correct operation relies on support in the wrapper functions for fork(2), vfork(2), and clone(2): if an application bypasses the glibc wrappers  for  these  system  calls  by  using syscall(2),  then a call to getpid() in the child will ...
1258233034 M * biz ... return the wrong value (to be precise: it will return the PID of the parent process).  See also clone(2) for discussion of a case where getpid() may return the wrong value even when  invoking  clone(2)  via the glibc wrapper function.
1258233158 M * Bertl well, util-vserver is built against dietlibc, so that shouldn't apply at that level
1258233557 M * Mr_Smoke Evening :)
1258233580 M * Mr_Smoke I'd like to ask a common/best practice question
1258233607 M * biz Bertl: I've got it! :)))
1258233610 M * Mr_Smoke Imagine this scenario where one physical host is used as www server + MX for two domains
1258233628 M * Mr_Smoke What's more sensible: use a vserver container per domain, or per task, or both ?
1258233662 M * Mr_Smoke So far I settled for vserver per task (ie. I have an MX vserver and a www vserver)
1258233692 M * Mr_Smoke But then I allocated only 1 IP for two domains, meaning I can't do SSL in apache
1258233743 M * Mr_Smoke Of course I could split the IP over the vservers, meaning mx.foo.com = www.foo.com over 2 vservers, and mx.bar.com = www.bar.com over the same 2 vservers
1258233747 M * Mr_Smoke Any ideas ?
1258233855 M * biz so you have two IPs and want to use both for apache (two different ssl certs with different CNAMEs?!), and also one of them for your mx?
1258233883 M * Mr_Smoke Well at the moment I haven't implemented SSL yet
1258233896 M * biz that's not really vserver specific, but you can do that with DNAT, regardless of how you setup your vservers
1258233903 M * Mr_Smoke Very true
1258234013 M * Mr_Smoke One possible scenario, as I described, would be to use 1 IP per domain, but have it "split" over the vservers, so that the MX vserver is lisetning on ports 25/587/etc while the www server is using the same IP but on ports 80/443
1258234052 M * Mr_Smoke I'm just wondering whether it sounds absolutely stupid or not :)
1258234082 M * biz actually I don't assign any "external" IPs to the guests, so I can handle them all/centrally at the host system and give all guests internal addresses, then I use DNAT/SNAT and have great flexibility this way :)
1258234152 M * Mr_Smoke Right, I did that on the previous setup too
1258234204 M * Mr_Smoke I could still do that, too
1258234297 M * Mr_Smoke Actually I'm not even sure postfix/dovecot know how to assign a different certificate based on the IP 
1258234303 M * Mr_Smoke address*
1258234311 M * Mr_Smoke So my point might be moot after all
1258234328 M * biz Bertl: a question in regards to this, can loaded shared libraries be re-used in different contexts/guests?
1258234514 M * biz .... imagine you go for one vserver per domain, run the same webserver in 100 guests and have to load the same shared libraries 100 times, that would be quite bad
1258234515 M * Mr_Smoke Moot indeed. There is no support for that ; I'd need two separate instances for dovecot, so let's forget about that
1258234905 M * Bertl biz: that's what unification (unify/hashify) is for
1258235455 J * scientes ~scientes@174-21-207-238.tukw.qwest.net
1258235580 M * biz Bertl: http://biz.baze.de/files/init.c
1258235592 M * biz without nptl:
1258235596 M * biz $ vserver foo start
1258235601 M * biz ppid = 20278, pid = 1
1258235601 M * biz syscall getpid = 1
1258235608 M * biz with nptl:
1258235612 M * biz $ vserver foo start
1258235616 M * biz ppid = 20523, pid = 20524
1258235616 M * biz syscall getpid = 1
1258235632 M * biz and since upstart links with it.... finally I found it ;)
1258235814 M * biz but ... _where_ is the pid cached exactly? upstart's init doesn't clone() prior to getpid(), so I guess it's vserver?
1258235850 P * ghislain 
1258235946 M * Bertl biz: please try with a recent kernel/util-vserver and make sure util-vserver is built with dietlibc
1258236781 M * biz humm, the debian/rules doesn't disable dietlibc, and dietlibc-dev is in the build-depends
1258236820 M * biz q
1258236858 M * biz aye.. crappy xinerama, focus doesn't follow my clicks ;)
1258237290 A * biz is building a 2.6.31.6-vs2.3.0.36.24
1258237313 Q * dna Quit: Verlassend
1258239179 A * biz realizes building the kernel on an old laptop was not the best idea... :D
1258239217 M * biz reminds me of the old times... 1h+ compile times for the kernel
1258240564 M * Bertl hmm, sounds like your kernel is not cleaned up properly :)
1258241118 M * biz well, it's an old HP omnibook with 133mhz and 64M :)
1258241662 M * biz ok, I'll stop it here.. I would need to get/build a guest first, too
1258241698 M * biz Bertl: I'll report back tomorrow or monday with current linux-vserver (... on some "usable" test machine :D)
1258241706 M * Bertl okay, thanks!
1258241739 M * biz thanks, bye
1258241887 M * fback time to get some sleep before morning birds, see you!
1258242819 M * Bertl cya