1124670632 Q * keyser_soze Quit: Abandonando 1124675806 Q * MooingLemur Ping timeout: 480 seconds 1124676057 J * MooingLemur ~troy@shells200.pinchaser.com 1124676584 Q * Hollow Remote host closed the connection 1124676600 J * Hollow ~Hollow@home.xnull.de 1124681519 P * stefani parting (is such sweet sorrow) 1124681656 Q * sid3windr arion.oftc.net kinetic.oftc.net 1124681656 Q * lilo arion.oftc.net kinetic.oftc.net 1124681656 Q * logger arion.oftc.net kinetic.oftc.net 1124681656 Q * jkl_ arion.oftc.net kinetic.oftc.net 1124681728 J * sid3windr luser@bastard-operator.from-hell.be 1124681728 J * lilo ~lilo@lilo.usercloak.oftc.net 1124681728 J * logger ~rs@vds.pas-mal.com 1124681728 J * jkl_ eric@c-67-190-142-149.hsd1.co.comcast.net 1124684874 J * keyser_soze ~cimarron@host113.201-252-24.telecom.net.ar 1124685970 M * Bertl k, off to bed now ... have a good one everyone! 1124685995 N * Bertl Bertl_zZ 1124686852 Q * keyser_soze Quit: Abandonando 1124687016 M * eyck k, off to the mountains...have a good one everyone! 1124692153 J * Aiken_ ~james@tooax8-020.dialup.optusnet.com.au 1124692482 Q * Aiken Ping timeout: 480 seconds 1124693110 J * comdata ~mertins@mx01.scheller.de 1124693140 M * comdata hello 1124693207 M * comdata anyone tested the latest patches on amd64? 1124693352 M * comdata having some problems compiling it 1124693435 M * mugwump which version of the patch/kernel release? 1124693508 A * mugwump fetches linux-2.6.12.5.tar.gz from his local mirror in 8s 1124693555 M * comdata 2.6.13-rc6 with patch-2.6.13-rc6-vs2.1.0-pre1 1124693580 M * comdata it stops with: 1124693585 M * comdata mm/built-in.o: In function `copy_hugetlb_page_range': 1124693585 M * comdata : undefined reference to `vx_rsspages_add' 1124693586 M * comdata mm/built-in.o: In function `unmap_hugepage_range': 1124693586 M * comdata : undefined reference to `vx_rsspages_add' 1124693587 M * comdata mm/built-in.o: In function `hugetlb_prefault': 1124693587 M * comdata : undefined reference to `vx_rsspages_add' 1124693587 M * comdata make: *** [.tmp_vmlinux1] Error 1 1124693802 A * mugwump prepares a build area to duplicate the fault... 1124693848 M * comdata mugwump: do u need the .config 1124694109 M * Hollow comdata: does 2.6.12.5-vs2.0 build on amd64? 1124694263 M * Hollow Bertl_zZ: we now have 2.6.12-vs2.0-gentoo-r1 in portage, so the EXTRAVERSION should be ok then.. 1124694930 A * mugwump builds 2.6.12.5-vs2.0 on amd64 as a sanity check 1124694982 M * comdata Hollow: didn't test that version, but will check that 1124695022 M * Hollow i didn't test 2.1.0_pre1 yet, but 2.0 is stable, and afaik it works on amd64 1124695039 M * mugwump sanity checks are like that 1124695108 M * Aiken_ silly question time, which kernel is the 2.1.0-pre1 against? 2.6.12.5? 1124695115 M * Aiken_ maybe time to fire up the alpha 1124695128 M * comdata 2.6.13-rc6 1124695238 M * comdata Compiling 2.6.12.5 1124695266 M * Aiken_ 15 min download followed by a 1 hour compile :( 1124695405 M * comdata Aiken_: what kind of alpha do u have 1124695660 M * Aiken_ alpha XL433 1124695661 M * comdata 2.6.12.5 with pach 2.6.12.3 compiled fine 1124695666 M * Aiken_ alcor 1124695683 M * comdata thats one of these desktop alphas? 1124695713 M * Aiken_ yes, it still had nt4 on it when I got the machine 1124695741 M * comdata have three alphastation 4/250 or so at home 1124695773 M * comdata an some ds20 here @work plus a AS2100 1124695830 M * Aiken_ this is the only alpha I am famillar with, the machine claims XL but I have to use a XLT milo and XLT kernel 1124695902 M * comdata the DS20 and alphaserver (AS) are bigger machines two times as wide as miditower you have 1124695950 M * Aiken_ and I would not say mine is small 1124695968 M * Aiken_ if I put the drive tower next it that increases the size 1124696098 M * comdata Aiken_: but then I still have the raid-array I can put next to it 1124696212 A * comdata is away: watching the machine reboot 1124696236 A * Aiken_ and his red wine are back :) 1124696254 M * Aiken_ I do not know how to get the scsi card working under linux 1124696324 M * Aiken_ if I remember correctly the controller has a i960 on it. 1124696338 M * Aiken_ I just use a 53c875 with the tower now 1124697676 J * revenger_ ~joe@bulldog.infosys.de 1124698064 Q * revenger Ping timeout: 481 seconds 1124698067 J * jsambrook ~jsambrook@aelfric.plus.com 1124698080 P * jsambrook 1124698291 J * revenger ~joe@bulldog.infosys.de 1124698307 J * fobi wht@liiifeeee.com 1124698311 M * fobi hi all 1124698343 M * fobi iv'e got sth. like this in my logs: 1124698345 M * fobi pam_limits[25363]: setrlimit limit #6 to soft=-1, hard=-1 failed: Operation not permitted; uid=0 euid=0 1124698362 M * fobi but i could'nt find the answer for this 1124698392 M * fobi probably it's vserver issue 1124698495 Q * revenger_ Read error: Operation timed out 1124699119 M * sid3windr yes, you're trying to set limits higher than the vserver itself has 1124699168 M * fobi sid3windr: soo what to do ? i don't have any limits in limits.conf so i should set some ? 1124699181 M * sid3windr I'm not sure :] 1124699193 M * sid3windr it's when you login with ssh right? 1124699207 M * sid3windr I did something to fix that in sshd_config I think 1124699211 M * sid3windr but I can't remember at all :/ 1124699278 M * fobi i think it's during login (in my case always ssh) 1124699289 M * fobi also when using su (started by cron) 1124699411 M * BWare Did you check your pam config and/or /etc/profile | /root/.bash_profile (in case of bash shell) ? 1124699468 M * BWare Wild guess, but form the log entry it seems login is trying to raise/set ulimits, which is not supported without proper CAPS 1124699842 Q * Hollow Remote host closed the connection 1124700101 M * fobi BWare: i don't have any limit's set 1124700129 M * fobi BWare: i am at vps (guest) and with basic debian installation 1124700938 M * BWare fobi: default there are some ulimits set 1124700952 M * BWare and from the looks of your log, pam is trying to change them 1124701335 Q * Greek0 Remote host closed the connection 1124701902 M * fobi BWare: ok, thanks a lot now i know what to look for 1124701913 M * BWare fobi: good luck 1124701950 M * fobi BWare: i will need it ;) 1124702297 J * sukria ~sukria@www.sukria.net 1124702460 P * Matthew-1 1124703965 J * Matthew-1 ~mlong@grm-nx.grmims.com 1124704184 A * comdata is back 1124704332 J * mlong168 ~mlong@grm-nx.grmims.com 1124704503 Q * lilo Remote host closed the connection 1124704510 J * lilo U2FsdGVkX1@lilo.usercloak.oftc.net 1124704630 Q * mlong168 Quit: Leaving 1124704666 J * mlong168 ~mlong@grm-nx.grmims.com 1124704772 J * Hollow ~Hollow@home.xnull.de 1124704810 P * Matthew-1 1124704843 J * _ag__ ag@caladan.roxor.cx 1124704853 J * mlong2031 ~mlong@grm-nx.grmims.com 1124704872 Q * _ag_ Read error: Connection reset by peer 1124704877 P * mlong2031 1124705135 Q * _ag__ Quit: 1124705368 Q * Hunger Ping timeout: 480 seconds 1124705374 J * _ag_ ag@caladan.roxor.cx 1124705809 Q * sukria Quit: leaving 1124706080 J * Hunger Hunger.hu@Hunger.hu 1124707444 Q * Aiken_ Ping timeout: 480 seconds 1124708026 Q * case Ping timeout: 480 seconds 1124711872 Q * virtuoso Ping timeout: 480 seconds 1124712898 J * virtuoso ~s0t0na@80.253.205.251 1124714038 N * Bertl_zZ Bertl 1124714054 M * Bertl morning folks! 1124715504 M * gregster good morning 1124715513 M * Bertl morning gregster! 1124715558 J * keyser_soze ~cimarron@host89.201-252-9.telecom.net.ar 1124715771 M * Hollow morning Bertl 1124715783 M * Bertl good morning Hollow! 1124715791 M * Bertl Doener: you're around? 1124715803 M * Hollow why is bucket probing expensive? 1124715941 M * Bertl it's not _that_ expensive now ... but in the near future, when we have per CPU buckets ... it will force the CPUs to synchronize to eachother ... 1124716008 M * Hollow ok, and i played around with your hard link approach... how do you apply patches? with -b? 1124716060 M * Bertl hmm, no, usually not, why? 1124716083 M * Hollow isn't the hardlinked file changed in every source tree if you patch it in one? 1124716096 M * Hollow it least it seemed so to me 1124716099 M * Hollow at 1124716166 M * Bertl no, patch takes care of that (by creating backup files called .orig by default) 1124716173 M * Bertl at least mine does :) 1124716236 Q * keyser_soze Quit: Abandonando 1124716262 M * Hollow well, seems like we have different patch versions than... mine has a --posix switch which creates backup files 1124716328 M * Bertl well, mine also has --unified-reject-files 1124716363 M * Bertl patch 2.5.9 1124716462 M * Hollow hm.. will look at it later.. afk for an hour or so 1124716631 M * Bertl okay, cya 1124716835 J * keyser_soze ~cimarron@host89.201-252-9.telecom.net.ar 1124717844 M * Doener Bertl: now i am 1124717920 M * Doener my patch doesn't create backups by default, but works fine with hardlinks... and i guess i got the same that Hollow has ;) 1124718015 M * Doener Bertl: hm, you answer _exactly_ the questions you get, eh? "Is there a way to limit memory for a vserver to 64MB?" "Yes." *g 1124718018 M * Doener * 1124718061 M * Bertl Doener: well, I didn't feel like explaining it over and over again :) 1124718099 M * Bertl Doener: will you be around in an hour or so? 1124718141 M * Doener yes, probably at my other desk, learning math. just cause some audible alert and i'll walk over here 1124718186 M * Bertl okay, because I would like to talk with you and Hollow (and maybe Greek0, and who else is interested :) about changes to the vshelper/reboot behaviour ... 1124718230 M * Doener ok 1124718463 M * Hollow back in about 30 mins.. 1124719624 M * wibble Bertl: something odd is going on with vdlimits... 1124719641 M * Bertl hey wibble! I'm waiting for details ... .) 1124719703 M * Hollow well, back now 1124719739 M * wibble http://www.wibblewobb1e.co.uk/vserver/vserver-issue 1124719767 M * Hollow lol 1124719792 M * wibble explain that one :) 1124719797 M * wibble hello Hollow ! 1124719801 M * wibble btw :) 1124719808 M * Hollow sorry, i can't :P 1124719813 M * wibble lol 1124719814 M * Hollow hi wibble 1124719817 M * wibble me neither 1124719822 M * wibble lets hope Bertl can :) 1124719824 M * wibble oh 1124719827 M * wibble hello Bertl !! 1124719831 M * wibble how rude was I 1124719889 M * Bertl wibble: that's expected :) 1124719896 M * wibble ? 1124719899 M * Bertl wibble: set the reserve to 0 :) 1124719904 M * wibble ah-ha 1124719945 M * wibble hmm 1124719954 M * wibble well I ran the same command, and I now have no quota! 1124719959 M * wibble vdlimit --xid 25 -s space_used=0 -s space_total=100000000 -s inodes_used=0 -s inodes_total=1000000000 -s reserved=5 /var/lib/vservers 1124719963 M * wibble backup:~# vserver test25 enter 1124719965 M * wibble test12:/# df -h 1124719968 M * wibble Filesystem Size Used Avail Use% Mounted on 1124719970 M * wibble /dev/hdv1 50G 515M 47G 2% / 1124719998 M * wibble unless 1124720003 M * wibble i am being thick 1124720004 M * wibble one sec 1124720029 M * wibble yes 1124720039 M * wibble it was me, I was setting the space_total too large 1124720051 M * wibble interestingly, it let me do it 1124720057 M * Bertl yeah .. 1124720058 M * wibble and let me do it without setting reserved 1124720068 M * wibble surely that is a bug, no? 1124720107 M * Bertl hard to say .. it is mostly based on 'you should know what you are doing' 1124720120 M * wibble I will grant you that. 1124720132 M * wibble but I can see that a few people might run into this 1124720137 M * Bertl also the tools do include but not really support vdlimits yet 1124720144 M * wibble so maybe a bigger warning sign :) 1124720153 A * wibble can write the documentation :P 1124720182 M * wibble right, so it werks on ext3 - lemme try reiser 1124720221 M * Bertl btw, I found some fix to add inode accounting for reiser :) 1124720228 M * wibble i c 1124720235 M * wibble where are the patches? :) 1124720254 M * Bertl not uploaded yet .. but I'll do so, when I'm done with my pizza :) 1124720326 M * wibble heh 1124720335 M * wibble well I have to get this server into production this week 1124720342 M * Hollow Bertl: so, you wanted to talk about the vshelper thing? 1124720344 M * wibble but have another server 1124720349 M * wibble but need these vlan fixes :( 1124720390 M * Bertl wibble: what about digging into the shell scripts, should not be too hard to fine the place where it uses a space instead of the dot? 1124720403 M * Bertl Hollow: yep, let's ring Doener! :) 1124720437 M * wibble Bertl: yes, I am going to look at - but the main issue is setting the loopback IP to the interface, instead of the actual IP! 1124720472 M * Bertl wibble: I'm not sure that _is_ an issue, did you check with 'ip addr ls' yet? 1124720483 M * wibble no, can do though 1124720537 M * Doener I heard Hell's Bells ringing? 1124720550 M * Hollow Doener: definitely ;) 1124720564 M * Bertl yep, okay, the idea/concept is simple and we already talked about that ... 1124720584 M * Bertl we should try to act on sys_reboot() as a real host does ... 1124720609 M * Hollow ah, btw: i updated http://home.xnull.de/misc/halt.png 1124720614 M * Bertl i.e. we kill each and every process in the guest (from the syscall) 1124720629 M * Bertl Hollow: excellent ... 1124720666 M * Bertl so the remaining question is, what shall become of the vshelper (as reboot helper) 1124720697 M * Bertl and the more I think about it, I'd say we remove it (again) 1124720736 M * Hollow imo we should add (or at least prepare) a system for all kinds of state changes 1124720746 M * Bertl what we actually want, is a helper which is called _after_ the context is destroyed, but with the knowledge 'how' it was rebooted ... 1124720769 A * Hollow nods 1124720794 M * Doener hm, i'd still consider that a reboot helper ;) 1124720805 M * Bertl yes, basically ... 1124720834 M * Bertl so the idea is to store the info from sys_reboot() somewhere (context) 1124720851 M * Bertl and then kill everything off, and return ... 1124720878 M * Bertl the context will go away shortly after that, and invoke the 'reboot' helper 1124720895 M * Doener hm, wouldn't that mean that you won't get clean shutdowns for non plain init style vservers? 1124720927 M * Bertl not if you do reboot -f (which would be realistic, no?) 1124720935 M * Doener unless you do "/etc/init.d/rc 6" instead of "reboot -f" 1124720964 M * Doener hm, i considered the "reboot -f" a required hack, which essentially does what "reboot" does on a normal host 1124720971 M * Hollow if sys_reboot is called before you shutdown your processes on the host, your lost too ;) 1124720973 M * Doener you loose that 'init-emulation' 1124720990 A * Hollow doesn't want hacks 1124721004 M * Hollow we need fixes, not hacks 1124721023 M * Bertl yes, but the 'init-emulation' style setup is a special hack we did ... 1124721038 M * Bertl and I think we can still keep that, without adding another hack 1124721055 M * Hollow let's consider that a feature ;) 1124721056 M * Bertl the first issue here is that 'reboot' will not work at all 1124721072 M * Hollow in fakeinit case? 1124721092 M * Bertl for some time now, we go and tell the folks: "yeah, stupid, you have to use reboot -f of course" 1124721106 M * Bertl but let's look at the 'of course' part ... 1124721134 M * Bertl because to me it's not really intuitive to use a 'force immediate reset' command to do a clean shutdown, no? 1124721144 M * Doener right 1124721270 M * Bertl so, IMHO we can adress this in several ways ... 1124721282 M * Bertl a) tell the folks to use /etc/init.d/rc 6 1124721296 M * Bertl b) replace the reboot/halt binary by something else 1124721312 M * Bertl c) let folks shutdown sysv style guests from outside 1124721338 J * renihs ~renihs___@193.170.52.70 1124721350 M * Hollow why do you want non-plain init style? it doesn't make that much sense to me.. 1124721351 M * Bertl welcome renihs! 1124721367 M * Bertl Hollow: because it is more resource friendly 1124721384 M * Hollow mhm 1124721385 M * Bertl if you do not need init lurking around, why start it at all? 1124721411 M * Hollow well, if you want the guest to behave like a real system you definitely want init 1124721422 M * Hollow but in the jail scenario, of course not 1124721427 M * Bertl that's something the 'customer' can decide ... 1124721448 M * Bertl but I have no problem with c) (only) in this case :) 1124721499 M * Bertl so, IMHO the next step could be: 1124721515 M * Bertl - add a flag to specify 'delayed vshelper' 1124721537 M * Bertl - based on that flag, disable the vshelper from sys_reboot() 1124721554 M * Bertl - also based on it, enable the vshelper on context disposal 1124721576 M * Hollow this would be the short time fix, right? 1124721581 M * Bertl - and finally make the sys_reboot() cause the kill 1124721582 M * Hollow s/time/term/ 1124721595 M * renihs hi Bertl! 1124721610 M * Bertl well, we can then start to move some stuff into 'legacy mode' (of course, slowly) 1124721661 M * Doener hm, sounds nice... less trouble with init-based vservers, while keeping a not-so-strange reboot for sysv vservrs 1124721661 M * Hollow i meant "long term" wrt a state change system 1124721684 M * Bertl we can also add some special sys_reboot() code which calls a helper in the sysv-based case (option b, from above) 1124721704 M * Bertl yes, long term, if we once get a control daemon ... 1124721709 M * Hollow ok 1124721723 M * Hollow i tend to use b) 1124721727 M * Bertl then we add all the different state changes and whatnot to that, test it and make it default ... 1124721752 M * renihs btw, i am not sure if its vserver related, but anyone happens to have tried to compile net-fs/samba-3.0.14a-r2 on a gentoo vserver box? 1124721776 M * Hollow but how does the kernel know if we have plain or sysv? 1124721781 M * Bertl renihs: does it fail? if so, why/what error? 1124721797 Q * comdata Quit: using sirc version 2.211+KSIRC/1.3.12 1124721823 J * Greek0 ~greek0@81.189.246.175 1124721823 M * Bertl Hollow: the kernel does not, we just add the flag for the new behaviour ... and the b-case can call out in both scenarios ... 1124721829 M * wibble Bertl: i think you are right 1124721831 M * renihs Bertl, wait i ll try to compile again, yes it fails 1124721834 M * wibble 8: eth0.2000: mtu 1500 qdisc noqueue 1124721834 M * wibble link/ether 00:11:d8:ba:54:c6 brd ff:ff:ff:ff:ff:ff 1124721834 M * wibble inet 127.0.0.1/8 brd 127.255.255.255 scope host eth0.2000 1124721834 M * wibble inet 10.0.20.254/24 brd 10.0.20.255 scope global eth0.2000 1124721836 M * Bertl ah, welcome Greek0! (right on time) 1124721850 M * Greek0 re 1124721851 M * wibble so, all I need to do is find that space... 1124721868 M * Bertl Greek0: please read up on the log of the last 15-20 minutes ... 1124721878 M * Bertl (if you've got some time, that is :) 1124721884 M * Hollow so, if sys_reboot is called, we call a helper which decides whether to shutdown via sysv or plain, and afterwards call the reboot helper 1124721885 M * Greek0 is an up-to-date log available online? 1124721889 M * Greek0 i was disconnected some time ago 1124721898 M * renihs btw, i DaPhreak on holiday or alike? :) 1124721904 M * Bertl Greek0: http://irc.13thfloor.at/LOG/2005-08/ 1124721911 M * Bertl (it's realtime :) 1124721921 M * Hollow renihs: no, he just did dd if=/dev/zero of=/dev/hda ;) 1124721931 M * renihs hehe 1124721960 M * Hollow and he's busy with his final test at university afaik 1124721969 M * Bertl Hollow: I'd say, we do it this way: 1124721970 M * renihs i did a a dd if=/dev/sda of=/dev/sdb yesterday without adding bs....it was a 200gb hd (takes hours) 1124721974 M * renihs Hollow, ah ok 1124721999 M * Bertl - add delayed_helper flag 1124722032 M * Bertl - call the vshelper with special option when sys_reboot() is called with special magic (i.e. not reboot/halt) unconditionally 1124722055 M * Bertl - delay/store the reboot/halt syscalls 1124722067 M * Bertl - call back unconditionally on context disposal 1124722090 M * Bertl note: the context disposal callback is different from the current shutdown callback 1124722149 J * Neubix ~brian@p54B0793D.dip.t-dialin.net 1124722158 M * Neubix Hi ll 1124722161 M * Bertl renihs: could you upload the error/log? 1124722161 M * Hollow yeah, sounds reasonable 1124722166 M * Bertl welcome Neubix! 1124722172 M * Neubix Hi Bertl 1124722191 M * renihs Bertl, once i have it, still compiling :) 1124722202 M * Hollow after call back we decide upon the stored syscall whether to call reboot helper or not? 1124722235 M * Neubix question .. my virtual interface is named eth0:something. can I change the "eth0" ?? confixx need this .. 1124722258 M * Bertl Neubix: change it to what? 1124722259 M * Greek0 thanks bertl 1124722284 M * Neubix change in venet0 ... venet0: error fetching interface information: Device not found 1124722291 M * Bertl Hollow: hmm, please rephrase .... 1124722346 M * Neubix problem is, confixx need a device called eth0 or venet0 1124722366 M * Hollow well, i don't see where the distinction between halt (we just kill remaining processes and are done) or reboot (we need to additionally call some helper which starts the vserver) is made... 1124722392 M * Bertl Neubix: well, broken confixx, but you can use nameif (at least until you get name-clashes) 1124722422 M * Bertl Hollow: ah, no, I don't want to do the distrinction at all ... 1124722449 M * Bertl Hollow: we call the context disposal helper in any case, it just gets some info if it was halt or reboot 1124722461 M * Hollow aah ;) 1124722490 M * Bertl but we make the original reboot helper depend on a special magic != halt _or_ reboot 1124722492 M * Hollow and if the helper returns, what do we do? 1124722503 M * Neubix hmm ... nameif .. ? .. 1124722521 M * Bertl Hollow: the context disposal helper? nothing, it can be called async ... 1124722537 M * Hollow hm, then it seem i still get the procedure wrong 1124722546 Q * monrad Quit: Leaving 1124722563 M * Bertl Hollow: what did you use for your fancy diagram? 1124722568 M * Hollow Dia 1124722576 M * Bertl okay, I remember that one ... sec 1124722642 M * Bertl Neubix: it's a very old command, used to name interfaces 1124722644 M * Hollow am i right, that the problem of the whole thing is the "shutdown" function of init? 1124722651 M * Hollow Bertl: ^ 1124722660 M * Neubix I fond it and read then man ... 1124722666 M * Neubix thanx 1124722683 M * renihs Bertl, http://www.doebling.sth.ac.at/vserver/samba.log 1124722695 M * Neubix I have write a howto .. bevor I public this, you like to read ? 1124722710 M * Hollow renihs: emerge info? 1124722774 M * Bertl Neubix: if you plan to put it on the wiki, just do that (folks will hopefully fix it up, if necessary), if you plan to publish it elsewhere, just paste an url here/on the wiki, folks will look at it ... 1124722793 M * renihs Bertl, http://www.doebling.sth.ac.at/vserver/emerge.log 1124722807 M * Bertl renihs: LC_ALL=C LANG=C :) 1124722818 M * Hollow hm, looks sane 1124722818 M * Neubix ok , I public it in my website and send you the link .. 1124722863 M * renihs Bertl, http://www.doebling.sth.ac.at/vserver/LC.log 1124722893 M * Bertl renihs: hehe, funny guy you are ... 1124722899 M * Hollow *gg* 1124722900 M * renihs thx :) 1124722934 M * Bertl looks like a gcc/binutils issue ... 1124722960 M * Bertl renihs: could you try that without the vserver startup, just with chroot? 1124722981 M * renihs Bertl, i am quite sure that works but hold on 1124723003 J * nox ~nox@noxlux.de 1124723136 M * Bertl welcome nox! 1124723419 M * nox thx Bertl 1124723423 M * Bertl Hollow: nah, dia would take too long (for me) you can do that if you like (just tried, got one decision and already unhappy :) 1124723439 M * Bertl I'll describe the process verbaly ... 1124723472 M * Bertl let's assume the context is running, we do not differentiate between sysv or init based guests ... 1124723491 M * Bertl now the following cases can happen: 1124723506 Q * zobel Ping timeout: 480 seconds 1124723520 M * Bertl A) the last process exits, context is disposed (natural death) 1124723542 M * Bertl B) the admin decides to shutdown/restart the guest (admin control) 1124723566 M * Hollow is B supposed to be used from guest or host or both? 1124723585 M * Bertl C) the guest admin decides to shutdown from inside, either by calling the rc script or reboot/halt or a modified reboot/halt 1124723595 M * Bertl (B is host admin) 1124723601 M * Hollow ok 1124723603 M * Greek0 so the main problem with reboot is that /dev/initctl doesn't exist, and thus reboot(1) can't tell init what to do.. 1124723612 M * Hollow yup 1124723623 M * Bertl Greek0: no, that's only in some cases the problem 1124723634 M * Greek0 what are the other problems? 1124723649 M * Bertl Greek0: precisely in non init style guests with unmodified reboot/halt :) 1124723680 M * Bertl (because for init style guests, init can be contacted, and if reboot/halt is modified, it doesn't matter) 1124723699 M * Greek0 ok, what are the other problems that can arise? 1124723709 M * Bertl Greek0: the problems are basically all kinds of issues arising from the non-host like behaviour 1124723734 M * Bertl reboot -f is known to do a system reboot, not a clean shutdown 1124723746 M * Bertl we changed that and try to work around the resulting issues 1124723757 M * Hollow i'd say if the guest has no init, it should be responsible for work-around that, not the kernel 1124723769 M * Bertl yep, that's my opinion too ... 1124723771 M * Hollow and not util-vserver 1124723789 M * Bertl okay, let's continue with the scenarios from above .. that we know what we are talking about 1124723795 M * Hollow and if sys_reboot is called, kill it, no matter what 1124723799 M * Hollow k 1124723802 M * wibble Bertl: ok, I think I found the vlan issue 1124723807 J * zobel zobel@zobel.irc.ftbfs.de 1124723809 M * Greek0 ok, the root of all evil just seems to be that /dev/initctl is missing in some cases, so reboot -f is needed, and it currently includes a hack that makes it do a clean reboot instead of a sudden kill. 1124723812 M * wibble its not putting a . in 1124723832 M * Greek0 so if we provide /dev/initctl in all cases unmodified reboot/halt binaries could be used. the reboot -f hack could be removed. 1124723853 M * Greek0 then we'd have reboot/halt in the guest just work, and reboot -f would just kill the context 1124723876 M * Bertl basically that's where we are heading to ... 1124723877 M * Greek0 i.e. the helper would never cause a clean shutdown of the guest, it would just be responsible for restarting it if needed 1124723905 M * Bertl okay, so let's look closer at the simple/easy case A 1124723955 M * Bertl the context dies, and currently nothing happens at all 1124724020 M * Greek0 I imagine it could work to turn the helper into some sort of post-death notification thingie. ie. it gets called whenever a context exits. (i.e. in case A too), probably with appropriate arguments 1124724029 M * Bertl I'd like to see that call a helper (once the context is gone) to report natural death 1124724038 M * Greek0 yep 1124724049 M * Bertl the tools might then decide what to do, i.e. restart it after a delay or whatever 1124724075 M * Bertl now the case B, where the host admin decides to shutdown ... 1124724097 M * Bertl depending on the guest config/init-style the tools would do: 1124724123 M * Bertl B.1 (sysv) simply call the shutdown scripts 1124724144 M * Bertl B.2 (init) contact guest init, to initiate a shutdown/reboot 1124724165 M * Bertl B.3 (unknown) kill out all processes? 1124724184 M * Hollow killall5 e.g. 1124724220 M * Bertl of course, in B.1 and B.2 the guest might somehow survive the shutdown, so a timeout is required, until a real vkill (all) happens 1124724264 M * Bertl if B.1/2 goes as expected, it will ultimately lead to calling reboot/shutdown no? 1124724296 M * Hollow yep 1124724298 M * Bertl in the B.1 case, the last command _after_ the script could be a guest side reboot/halt too ... 1124724333 M * Bertl in this setup, we would end up with a sys_reboot() in _any_ case, except for the timeout/killall 1124724378 M * Bertl as the tools _know_ that they are in a shutdown state (for the context) they could handle _all_ callbacks accordingly 1124724383 M * Greek0 hmm. if we're talking about details, how to implement the timeout/killall? 1124724406 M * Hollow Bertl: in the init case too.. 1124724408 M * Bertl Greek0: we have a vwait() and the util-vserver tools actually do that right now ... 1124724410 M * Greek0 I'm not convinced a sleep 5/10/whatever; ; is a very good idea 1124724444 M * Greek0 the change _you_ made to the tools to enable a clean shutdown? 1124724455 M * Greek0 when I was pretty new here? 1124724468 M * Bertl was an example ... but some parts of it are already in 0.30.208 1124724502 A * Greek0 looks up the timeout value 1124724509 M * Bertl okay, so what happens in the B.* cases is, that ultimately the disposal helper is called and takes whatever action necessary/reuested 1124724512 M * Bertl *requested 1124724558 M * Bertl now the C case ... 1124724588 M * Bertl if the admin calles the shutdown scripts (from inside) they will also end up in a halt/reboot 1124724592 M * Bertl *calls 1124724618 M * Bertl or, if not, end as a 'natural death' of the guest 1124724653 M * Bertl in the cases where some init is involved, that one will call the reboot/halt 1124724666 M * Bertl so we can identify 3 cases for C too ... 1124724678 M * Bertl C.1 natural like death (no reboot/halt called) 1124724687 M * Bertl C.2 reboot/halt called from scripts 1124724695 M * Bertl C.3 reboot/halt called from init 1124724704 M * Bertl we now can also add another case 1124724719 M * renihs Bertl: in chroot it fails with same error 1124724720 M * Bertl C.4 special modified reboot/halt called from inside 1124724732 M * Bertl renihs: so it's a gentoo issue I guess ... 1124724759 M * renihs yah seems like 1124724768 M * Greek0 Bertl: to be honest I don't think requiring a patched reboot/halt within the guests is such a great idea.. 1124724768 M * renihs gruml, thats the *only* ebuild there is :) 1124724776 M * Bertl the C.4 case would allow to provide the current reboot -f behaviour, without any init running (so for sysv/gentoo? guests) 1124724794 M * Hollow please forget gentoo init style 1124724802 M * Hollow or well 1124724803 M * Greek0 I'd much rather just create a /dev/initctl, so the reboot/halt work fine. 1124724806 M * Hollow don't forget it 1124724818 M * Greek0 that would be as close to a real system as we can get IMHO 1124724825 M * Hollow but it definitely needs to be reviewed :P 1124724827 M * Bertl Greek0: that's not such a good idea ... why? simple: 1124724864 M * Bertl the initctl (although placed in /dev) is created by ... init, and the telinit (and friends) know what information to send there ... 1124724880 M * Bertl different inits .. different messages passed ... 1124724910 M * Bertl so we do not want to parse/understand each and every init somebody might come up with ... 1124724923 M * Bertl that's something we leave for the real init ... 1124724929 M * Greek0 ok, seems reasonable 1124724948 Q * lilo Ping timeout: 480 seconds 1124724951 M * Bertl so folks who want /dev/initctl, have to use an init. period. 1124724983 M * Bertl but I can figure that a 'slightly' modified halt/reboot command, could be interesting for those who do not want to do so ... 1124724999 M * Bertl and I don't see that this would be very complicated either ... 1124725009 M * Greek0 I haven't looked at the init sourcecode a lot, but how complex are those messages usually? 1124725012 M * Bertl we just use the existing framework, but with a different magic/argument 1124725032 M * Bertl Greek0: for init, I have no idea ... 1124725054 M * Bertl Greek0: but there is no rule that init has to communicate via initctl at all 1124725056 M * Greek0 I understand that it sucks to parse what whichever telinit wants to write to that device, but aside from that it would be the cleanest solution IMHO 1124725178 J * lilo U2FsdGVkX1@lilo.usercloak.oftc.net 1124725234 M * Greek0 I mean replacing restart/halt seems to defeat the purpose of virutalisation a bit. guest users can't just recompile them and expect it to work, unless they know about the patch. and they'd also be gone if you just install a new package version 1124725296 M * Bertl correct, so those will be using a real init anyway ... because replacing the init package will have strange results, no? 1124725300 M * Greek0 and when users come and complain why reboot doesn't work from the guest, telling them "oh, you have to use a patched reboot/halt" is probably even stranger then telling them to use reboot -f 1124725342 M * Bertl but it could be easy to tell them, use vreboot/vhalt instead? 1124725388 M * Greek0 Bertl: do you think this init-style could be made the default? then users who want to save the overhead of an extra init process per guest would have to deal with this, while the broad majority of users doesn't have to care 1124725413 M * Bertl yes, that's where I think we should head to ... 1124725425 M * Greek0 yep, fine then 1124725454 M * Hollow yup, i understand and agree ;) 1124725473 J * stefani ~stefani@superquan.apl.washington.edu 1124725486 M * Bertl welcome lilo, stefani! 1124725491 M * Hollow i'll do this fancy dia stuff for this scenario too, but dinner time first.. cu later 1124725501 M * Bertl cya, and tia! 1124725502 M * stefani Bertl: hola. 1124725782 M * Neubix Medive: I found in IRC logs, that you have fix the confixx eth0 problem with "ifconfig wrapper" 1124725798 M * Neubix sorry, Medivh 1124725910 M * Bertl yeah, now I remember that one ... 1124725960 M * Neubix with nameif I think it's not running 1124725983 M * Bertl I don't know the confixx source .. so I can't tell :) 1124726006 M * Neubix but the ifconfig wrapper source *smile* 1124726133 M * Bertl not yet, but I remember that I suggested to use a wrapper some time ago ... 1124726191 M * Neubix ok, medivh tell me it's easy .. now I must understand :) 1124726223 M * Bertl Medivh: is the wrapper script available somewhere? 1124727551 M * wibble Bertl: I have fixed the issue... :) 1124727564 M * wibble will create a patch me thinks :) 1124727573 M * Bertl excellent! 1124727607 M * wibble hehe, technically do i count as a vserver developer then? :) 1124727630 M * wibble right 1124727631 M * wibble bbl 1124727632 M * Bertl a contributor .. yes, developer .. hmm, after 20+ fixes, I guess 1124727633 M * wibble pub time 1124727637 M * wibble lol 1124727644 M * wibble well somehow I don't think so 1124727656 M * wibble one patch is good enough for me 1124727712 M * Bertl we leave that decision to the contributors :) 1124727789 M * Neubix ok, I must go ... CU@all 1124727796 M * Bertl cya Neubix! 1124727807 Q * Neubix Quit: Verlassend 1124727817 M * Bertl (working on a strictly schedule, 18:00, off home) 1124728209 J * menomc ~amery@200.75.27.2 1124728237 M * Bertl welcome menomc! 1124728313 M * menomc thx Bertl :) 1124728344 N * menomc mnemoc 1124730162 Q * renihs Remote host closed the connection 1124730526 M * mnemoc Bertl: sorry to ask but, what happened with the job offer for virtuozzo? 1124730674 M * Bertl they did not get back to me and I did not get back either ... 1124730705 M * mnemoc oh 1124730824 M * Bertl would you prefer virtuozzo over linux-vserver? 1124730899 A * BWare prefers freedom of choice ;) 1124731146 M * Nicoli Bertl i would but licenecing is expensive 1124731161 M * Nicoli they ahve a really nice gui for it 1124731181 M * Nicoli altho it can fail to create VDS's quite often 1124731197 A * Nicoli uses virtuozzo @ work with 28 servers 1124731224 M * Nicoli and sometimes it wont start VDSs for an unspecified reason 1124731240 M * Nicoli and other times it will randomly leave some stuff out when creating vds's 1124731250 M * Nicoli but thast only 10% of the time 1124731290 M * Bertl Nicoli: well, everybody is free to make that decision ... I definitely prefer linux-vserver (with source code, for free) over virtuozzo (without source, for big cash) :) 1124731333 M * Nicoli yea liek i said, liceincing is too expensive 1124731349 M * Nicoli im more of an endusers and not a coder so having the code wouldnt make too much of a dif to me 1124731366 M * Nicoli altho virtuozzo does have a windows vds ver 1124731372 M * Nicoli altho its beta @ best 1124731385 M * Nicoli we had lotsa networing trouble with it 1124731391 M * Nicoli networking 1124731423 M * Hollow Bertl: i'm still not sure about the B.1/2 step... e.g.: we're tell init to shutdown, it stops all processes, but some maybe left... so should we simply kill the rest, and the main problem: when to kill them? how to choose the timwout? 1124731432 M * Bertl well, if you find/have features in virtuozzo, you are missing on linux-vserver, feel free to make feature requests ... (or hire somebody to implement them) 1124731484 M * Bertl Hollow: init, results in sys_reboot() being called, which in turn kills all processes, no timeout required here ... 1124731503 M * Nicoli hehe well my request would be maybe a web-based thing for starting \ stopping \ creating VDS's 1124731505 M * Nicoli thast about it 1124731508 M * Hollow init definitely calls sys_reboot, no matter what? 1124731513 M * Nicoli vserv is good about everythign else 1124731525 M * Bertl Nicoli: did you look at the OpenVPS and similar stuff? 1124731549 M * Bertl (IIRC there is even a project to admin guests via PDA) 1124731550 M * Hollow (or at least say it should do) 1124731564 M * Nicoli Bertl nope 1124731590 M * Nicoli i actually heard about vserver thru xen 1124731623 M * Nicoli a chance post on the mailing list 1124731711 M * Nicoli which is quite luky cause i liek it more, much less pain in the ass to setup and can be done by pasteing a few (long) lines of code into terminal 1124731757 M * Nicoli but i actually have a set of lines that will intsall it and get it running on centOS 4.1 1124731786 M * Nicoli its compelatly automatic aside from the kernel config 1124731792 M * Nicoli which is easy 1124732093 M * Hollow Bertl: the init process itself never calls sys_reboot 1124732115 M * Hollow at least i can't find it in the source 1124732128 M * Hollow only halt calls it 1124732143 M * Hollow also init never kills itself 1124732164 M * Hollow even if you change to the reboot/shutdown runlevel 1124732168 M * Bertl but the tools/scripts do so, no? 1124732186 M * Hollow yeah, they call halt at the end 1124732191 M * Hollow or halt -f 1124732216 M * Hollow so, basically, we don't have to wait for any timeout... 1124732241 M * Bertl if everything works out as expected, no ... but to make sure, we have to 1124732294 M * Hollow well, if there are processes left when halt -f is called, just kill them, if there is some error before the halt -f call, just leave as-is, no? 1124732961 M * Hollow Bertl: http://home.xnull.de/misc/disposal.png 1124733006 M * Bertl IMHO, sys_reboot() doesn't have to check for processes 1124733017 M * Bertl it can just go wild and kill everything left, no? 1124733037 M * Hollow yeah.. but if nothing's left, it will kill nothing ;) 1124733037 M * Bertl (of course, that will not do anything if there is none left) 1124733042 M * Hollow ;) 1124733076 M * Bertl yeah, the difference between host and guest is just the invocation 1124733082 M * Hollow so, the sysv style needs some script inside to kill all processes safely right? 1124733109 M * Hollow vhalt or sth 1124733208 M * Bertl if any remain there ... 1124733244 M * Bertl in theory, the runlevel scripts should stop all services, no? 1124733263 M * Hollow yeah 1124733288 M * Hollow so, basically, we don't care.. if sys_reboot is called we kill the processes 1124733327 M * Bertl right 1124733360 M * Hollow that sounds pretty straight :) 1124733411 M * Hollow and the disposal helper has to be called somewhere between ctx_kill and ctx unhashing, right? 1124733447 M * Bertl ad straight: yes, doesn't save us from putting the emergency trip wire in place (the timeout) 1124733463 M * Hollow where would we need a timeout? 1124733488 M * Bertl ad helper: no, the disposal helper will be called after the context structures are disposed (i.e. when the context doesn't exist any longer) 1124733506 M * Bertl Hollow: in the 'host-admin' shuts down guest scenario 1124733522 M * Hollow wouldn't we have the problem of "no such context id" then...? 1124733534 M * Bertl you do not want to have a vserver stop to hang around forever, because some funny guy put a sleep 1000 into his scripts ... 1124733535 M * Hollow if the disposal helper calls tools with --xid 1124735024 Q * keyser_soze Quit: Abandonando 1124735373 M * wibble right 1124735375 M * wibble here is the vlan fix 1124735377 M * wibble http://www.wibblewobb1e.co.uk/vserver/util-vserver-0.30-208-vlanfix 1124735383 M * wibble minimal testing 1124735600 M * Bertl hmm ... the $@ suggests that this better go upwards no? 1124735622 M * Bertl I suspect if you have two IPs they will be all lsited there ... 1124736028 M * wibble yes, i think so 1124736032 M * wibble but im no hacker :) 1124736059 M * wibble i will investgate further 1124736086 M * Bertl k, TIA! 1124736562 J * dchouinard ~dchouinar@142-217-213-125.telebecinternet.net 1124736566 M * wibble hmm 1124736680 M * Bertl welcome dchouinard! 1124736681 M * wibble problem is the code is used for both the construction of the interface and deconstruction of the interface 1124736736 M * dchouinard Hi Hebert :) 1124736781 M * dchouinard The guy to talk to when you have been working for days looking like a fool ;) 1124736833 M * Bertl yes? 1124736881 M * dchouinard I'm trying to do something simple (i guess). I have a vserver that need to mount the /home of the host server 1124736902 M * dchouinard Everything else can mount the /home but the share 1124736945 M * dchouinard Then I start to play with the v_portmap (on the host) and now my nfs service refuse to start. 1124736996 M * dchouinard So I guess I have to say something to the nfs service too but I don't how what or how. 1124737013 M * Bertl so you have a /home directory on the host ... and you want to share that with some guests, right? 1124737030 M * dchouinard right. ANd the guess is within the vserver 1124737035 M * Bertl dchouinard: my question here is, what are you doing with nfs? 1124737085 M * dchouinard I'm currently using on big fat machine and I will seperate the servers later on two seperate machine 1124737150 M * Bertl well, nfs to the localhost is a tricky thing ... you have to be careful about that 1124737176 M * Bertl I would suggest to go for --bind now, and use nfs later, once you have the second machine, and only for the second one ... 1124737227 M * dchouinard Is that from a technical perpective or a security point of view? 1124737475 M * Bertl mainly from an administrative/resource overhead POV :) 1124737503 M * wibble hmmm 1124737518 M * wibble samba is having issues with the new interface stuff 1124737529 M * wibble as i think it uses "ifconfig" 1124737538 M * wibble which doesn't actually display anything 1124737542 M * wibble but ip addr list 1124737543 M * wibble does 1124737562 M * Bertl you can always assign an alias name ... 1124737603 M * wibble ? 1124737637 M * Bertl if you specify (in /interfaces//name) a name, an alias will be created 1124737656 M * Bertl e.g. eth0.135:hansi (you put hansi into the name file) 1124737673 M * Bertl this will show it in ifconfig ... 1124737684 J * mess-mate ~mess-mate@lns-vlq-7-lil-82-254-207-254.adsl.proxad.net 1124737703 M * Bertl welcome mess-mate! 1124737735 M * mess-mate I'm there with more questions :) 1124737759 M * Bertl really? any questions left? :) 1124737790 M * mess-mate no, several :) 1124737809 M * wibble ok it works now, anyway to create a name in vserver buid ? 1124737811 M * wibble build* 1124737828 M * Bertl --interface hansi=eth0.135:... 1124737879 J * yarihm ~yarihm@vpn-global-dhcp1-18.ethz.ch 1124737889 M * Bertl wb yarihm! 1124737902 M * wibble ah i see, another undocumented feature!" 1124737925 M * yarihm hi Bertl 1124737986 M * Bertl wibble: depends how you define 'undocumented' 1124738005 M * Bertl vserver - build --help says: 1124738007 M * Bertl --interface [=][:][/] 1124738007 M * Bertl ... declares an network-interface; this option can be specifie 1124738008 M * mess-mate what's the difference between using or not the interfaces of the host ? 1124738015 M * wibble ah ok :) 1124738061 M * Bertl mess-mate: what do you mean? 1124738181 M * mess-mate Bertl: well, if a vserver is created without declaration on the installation of say IP, hostnamE etc..; automatically they are taken those from the host. 1124738223 M * Bertl I _assume_ (read: I haven't verified it yet) that it will bind to 0.0.0.0 (the guest) or not use chbind at all 1124738228 M * mess-mate Bertl:doc: Set the hostname (FQDN) of the vserver. If empty, the hostname of the host server is used. 1124738269 M * Bertl regarding your latest email questions: 1124738270 M * Bertl 1. how can i run whitin that veserver as a normal usre, not as roo T 1124738286 M * Bertl just use 'su - user' once inside the guest 1124738305 M * Bertl 2. what other than a legacy ? vitual ? how can i setup a vserver other than a legacy ? 1124738337 M * Bertl if you use recent tools (read: util-vserver 0.30.208 or so), then they will create 'normal' (also often called new-style) configs 1124738378 M * Bertl if you use old/outdated tools (or some distro specific scripts) then they might create legacy guests (old-style config, the single file) 1124738402 M * Bertl the legacy config is deprecated since almost a year, and it will not support recent features ... 1124738421 J * monrad ~monrad@213083190134.sonofon.dk 1124738425 M * Bertl welcome monrad! 1124738488 M * mess-mate .conf and .sh files 1124740090 M * Bertl okay, I'm off for now .. but will be back later ... 1124740095 N * Bertl Bertl_oO 1124740378 Q * yarihm Quit: Leaving 1124740741 Q * mess-mate Quit: leaving 1124741804 Q * skceb arion.oftc.net unununium.oftc.net 1124741804 Q * BobR_oO arion.oftc.net unununium.oftc.net 1124741804 Q * chand arion.oftc.net unununium.oftc.net 1124741804 Q * locksy arion.oftc.net unununium.oftc.net 1124741804 Q * derbien arion.oftc.net unununium.oftc.net 1124741804 Q * ray6 arion.oftc.net unununium.oftc.net 1124741804 Q * Bertl_oO arion.oftc.net unununium.oftc.net 1124741804 J * chand ~chand@staff.lycos.fr 1124741810 J * ray6 ~ray@klon3.gcsc2.ray.net 1124741813 J * derbien ~derbien@whiterabbit.nbmc.de 1124741818 J * locksy ~locksy@mrtg.sisgroup.com.au 1124741819 J * BobR_oO ~georg@212.16.62.52 1124741827 J * skceb skceb@bang-bang.feuer-frei.com 1124742106 J * Bertl_oO ~herbert@212.16.62.52 1124742149 Q * albeiro Ping timeout: 480 seconds 1124742203 J * albeiro ~albeiro@graffias.estrefa.pl 1124742581 Q * dchouinard Quit: Leaving 1124743762 Q * keyser_soze Quit: Abandonando 1124744980 Q * Snow-Man Read error: Connection reset by peer 1124745107 J * mess-mate ~mess-mate@lns-vlq-7-lil-82-254-207-254.adsl.proxad.net 1124745188 J * Snow-Man_ ~sfrost@snowman.net 1124745249 Q * Snow-Man_ Remote host closed the connection 1124746143 J * Snow-Man ~sfrost@snowman.net 1124747328 N * Bertl_oO Bertl 1124747339 M * Bertl back now ... 1124747357 M * mess-mate was a long time :) 1124747394 M * Bertl yeah? 1124747411 M * Bertl mess-mate: so you already missed me? 1124747428 M * mess-mate Bertl: YES ! 1124747500 M * mess-mate Bert: i'm trying that newserver command. The originally installed from a debian package won't work. Let you know what happen. 1124747563 M * Bertl well, you have to talk to the debian maintainer about that ... 1124747656 M * mess-mate Bertl: yes i do. But i think the source won't work with debian also. 1124747688 M * mess-mate Bertl: that d.. resolve-deps thing. 1124747699 M * Bertl hmm? 1124747733 M * Bertl you mean that etch and what was it sid are broken, and cannot be debootstrapped? 1124747987 M * mess-mate Bertl: don't know about sid. It's always that resolve-deps option. Like to run now when added --resolve-deps after debootstrap at line 395, see #Fire off 'debootstrap' .... 1124747995 Q * sid3windr Ping timeout: 480 seconds 1124748005 J * Aiken ~james@tooax6-089.dialup.optusnet.com.au 1124748075 M * Bertl mess-mate: well, at least with woody and maybe sid (don't remember) we checked it some time ago ... but yeah, maybe it needs some fixing ... not sure ... 1124748092 M * Bertl welcome Aiken! 1124748239 M * Aiken hi 1124748248 M * Aiken trying to compile 2.1.0-pre1 1124748259 M * Bertl brave! 1124748263 M * Aiken running into my kernel size problem again 1124748290 M * Aiken on my 4th compile trying to get the new kernel small enough 1124748309 M * Bertl hmm .. isn't there any good bootloader? 1124748358 M * Aiken the milo I have been able to run was the xlt milo from my alpha redhat 4.1 cd 1124748359 M * Bertl I mean, maybe we shoulöd add some decompression stuff for alpha? 1124748369 M * Aiken newer version of milo won't even work for me 1124748390 M * Aiken I am trying with no ksyms to see what the size will be like 1124748405 M * Aiken though I imagine that won't help if a backtrace is needed 1124748435 M * mess-mate After installing a vserver with 'newvserver' (from debian) a est2.conf file is installed in /etc and not a test2 directory ! 1124748458 M * Aiken 2.6.12.4-vs2.0 was just under 4 meg in size, but 2.6.13 with the same config was 4.3 meg in size :( 1124748469 M * Bertl mess-mate: yes, that's because the debian newvserver tool is hopelessly outdated ... 1124748490 M * mess-mate I'll wipte it out !!! 1124748507 M * Bertl better complain to the maintainer first, then wipe it out! 1124748613 M * Aiken when you said brave, how brave should a person be? fullsystem backup first or just be prepared for a fsck next boot? 1124748796 Q * obi Ping timeout: 480 seconds 1124748998 M * Bertl Aiken: no, actually I suspect 2.1.0 to work quite fine ... 1124749021 M * Bertl Aiken: but testing is always brave, no? 1124749077 M * Aiken this is on a system I have a full back of 1124749098 J * obi ~obi@asus.saftware.de 1124749108 M * Aiken I am not overly worried if I kill that image 1124749129 M * Aiken I only test on 2 system and I have full backups of each just in case something goes wrong 1124749177 M * Bertl that's actually a good thing in any case ... 1124749237 M * Aiken the alpha also has 2 system images on it, one I use as a rescue system if needed and the main image 1124749246 M * Aiken backups are always good 1124749358 M * Aiken the room that alpha is in is cold enough I have until at least spring before it will be put into use so I have time as well 1124749426 M * Bertl what's the temperatur outside at day? 1124749464 M * Aiken outside has just climbed to 13C, it is 8:24am local 1124749483 M * Aiken sun has been up 2 hours 1124749506 M * Bertl so it will get to 15-18°C ? 1124749627 M * Aiken I think the predicted max is 18 - 20 1124749645 M * Aiken Linux pebbles.bedrock 2.6.13-rc6-vs2.1.0-pre1 #3 Tue Aug 23 08:21:07 EST 2005 alpha Unknown Alcor GNU/Linux 1124749653 M * Aiken testme passed 1124749662 M * Bertl excellent ... you now have COW and BME :) 1124749673 M * Aiken bme? 1124749685 M * Aiken blind mount extensions? 1124749686 M * Bertl the bind mount extensions, ro and such for --bind 1124749696 M * Aiken but still not sure what that means 1124749710 M * Aiken ok 1124749734 M * Bertl so if you do 'mount --bind -o ro /tmp /mnt 1124749749 M * Bertl then /mnt will not be able to change the /tmp contents 1124749784 M * Aiken so this would allow ro access to parts of the host? 1124749816 M * Bertl yeah, and other flags atime/noatime work too 1124749817 M * Aiken if host /usr and all guest /usr where the same just to a ro bind to the host /usr 1124749831 M * Bertl you could do that ... yes 1124749849 M * Aiken that could save quite a lot of space 1124749870 M * Bertl well, unification does that too, just without the issues of being ro 1124749888 M * Bertl and with the COW link breaking, even in a writeable manner 1124750011 M * Aiken does the COW have to be off the host or can it be off a master vserver image 1124750043 M * Bertl usually you use either a master vserver, or unification between guests 1124750046 M * mess-mate Bertl: COW ? 1124750052 M * Bertl Copy on Write 1124750062 M * mess-mate ok 1124750083 M * Bertl Aiken: so you do not need a master/template, but you can have one 1124750111 M * Aiken I have been using a master image 1124750183 M * mess-mate do you mean a 'base-system' with master/template ? 1124750210 M * Bertl define base-system :) 1124750253 M * mess-mate Bertl: a base-system is a set of only the absolute necessary packages to run a system. 1124750277 M * Aiken I have a guest image I made up called master. when I want a new vserver I copy master across 1124750297 M * Bertl mess-mate: in the old days, when there was no util-vserver to build the guests out of thin air, we had a kind of well tuned 'guest image' we used for copying everytime we needed a new guest 1124750311 M * Bertl and it is said, that folks still do so nowadays :) 1124750482 M * mess-mate Ahh..so..:) 1124751186 Q * obi Ping timeout: 480 seconds 1124751336 M * mugwump Aiken: that's how I use the unification, I also use a pre-start script to setup the bind mounts 1124751460 J * obi ~obi@asus.saftware.de 1124751472 M * Bertl mugwump: so you are a long time user of the BME stuff? 1124751475 J * renihs ~renihs___@193.170.52.70 1124751479 M * mugwump BME? 1124751490 M * Bertl wb obi! welcome renihs! 1124751518 M * Bertl mugwump: well, or do you use rw --bind mounts with ro mounted filesystems? 1124751536 M * Bertl (or did I get your comment wrong in the first place :) 1124752838 M * mugwump no, I just bind in parts from /vservers/shadow/xxx/{usr,sbin,lib,bin} to /vservers/xxx/{usr,sbin,lib,bin} 1124752852 M * mugwump so that /vservers/xxx is on a different LV to /vservers/shadow 1124752880 M * Bertl yeh, but readonly or read/write or don't care :) 1124753024 P * stefani I'm Parting (the water) 1124753256 J * keyser_soze ~cimarron@host89.201-252-9.telecom.net.ar 1124753319 M * Aiken is there documentation for using the COW or is automatic when using links? 1124753341 M * Bertl it is automatic when using unification 1124753375 M * Bertl so the basic criterion is an unified file (hardlink, immutable flag set, immu unlink set too) 1124753418 M * Bertl (you can get that with setattr --iunlink IIRC) 1124753714 M * Aiken is the setattr on the original file or the hardlink? 1124753752 M * Bertl you are really able to differentiate between them? 1124753799 M * Aiken ok 1124753802 M * Aiken not working 1124753827 M * Aiken I created a file on the host and a hardlink in the guest, used the above setattr 1124753838 M * Aiken does not matter which copy I change they both change 1124753860 M * Bertl really? 1124753930 M * Bertl let me upload a small test script ... 1124754019 M * Bertl http://vserver.13thfloor.at/Experimental/cow_test.sh 1124754039 M * Aiken http://pastebin.com/343492 1124754153 M * Bertl http://pastebin.com/343495 1124754235 M * Aiken I wonder if it is nfs rearing it's ugly head again 1124754243 M * Aiken http://pastebin.com/343497 1124754266 M * Bertl ah, nfs, well, in this case you need a patched server ... 1124754299 M * Bertl looks good to me ... 1124754329 M * Aiken I am confused, /vservers is a nfs mount point http://pastebin.com/343499 1124754333 M * Aiken that seemed to work 1124754356 M * Aiken maybe, don't know, just assumed I am confused 1124754386 M * Bertl with vservers on a remote machine, the remote has to do the COW breaking ... 1124754437 M * Bertl but I have to admit this case wasn't even considered yet :) 1124754446 M * Aiken currently don't have enough storage local which is why I am using nfs 1124754458 M * Aiken I am wondering is iscsi or aoe would be easier 1124754494 M * Bertl iSCSI/nbd would be interesting to test/add .. so if you want to investigate this 1124754526 M * Bertl because iSCSI and nbd will require context awareness 1124754534 M * Aiken already setup with iscsi on several machines 1124754556 M * mugwump why would a block device care about contexts? 1124754561 M * Aiken have played with aoe, marginally faster than iscsi and lower cpu load on the server than iscsi 1124754592 M * Bertl mugwump: because the requests can originate from 'within' a context, where they should not (network context in this case) 1124754672 M * mugwump but, wouldn't you just setup the aoe/iscsi block device in the root context, then mount/partition/etc on top of that? 1124754703 M * Bertl yes, but consider the following: 1124754721 M * Bertl guest opens a file (maybe for direct I/O or whatever) 1124754750 M * Bertl then a guest process writes a block, and the block is transferred to the block device 1124754764 M * Bertl which in turn, sends something over the network ... 1124754779 M * mugwump ok 1124754792 M * Bertl somewhere on the way, the 'context' should have been switched, but wasn't ... 1124754804 M * mugwump why would you want the context to switch for that? 1124754818 M * Bertl that caused a lot of funny issues with NFS for example ... 1124754838 M * Bertl guest IP 192.168.0.2, nfs mounted from 10.0.0.1 (local IP 10.0.0.2) 1124754860 M * Bertl now 'some' nfs requests originate from 192.168.0.2 and, guess what? get denied :) 1124754996 M * mugwump right, I see, you need to UN-switch from a context to do the accessing 1124755023 M * Bertl yes, that's the case for all network based filesystems 1124755032 M * Bertl (unless they are mounted from within a guest) 1124755068 M * Aiken aoe was the quickest to get going, COW with ext3 on top of aoe 1124755068 M * mugwump ok, unless that filesystem/device has a kernel helper process 1124755119 M * Aiken formatting a 1 gig filesystem over 10MBit took awhile 1124755152 Q * daniel_hozac Quit: CDRW insertion 1124755163 Q * Doener Ping timeout: 480 seconds 1124755167 J * Doener ~doener@p5487601E.dip.t-dialin.net 1124755171 Q * mess-mate Quit: bye 1124755180 Q * keyser_soze Quit: Abandonando