1119484821 M * brc i just thougth there mighbt be any other way so that i could flush those logging rules 1119484822 M * Bertl good 1119484845 M * mugwump fwbuilder has a nice little "accounting" log target for rules 1119484849 M * Bertl as I said, sockets are accounted, but that is not identical to the network info 1119484854 M * mugwump dunno what I does, though 1119484858 M * mugwump s/I/it/ 1119484890 M * brc Bertl: you mean, number of open sockets ? 1119484914 M * brc sorry doesn't know what is socket accounting 1119484976 M * Bertl /proc/virtual//cacct 1119485015 M * Bertl it accounts socket traffic, read/write bytes and packets 1119485057 M * brc isn't that the same as using iptables to log the traafic ? 1119485070 M * Bertl no, for several reasons ... 1119485080 M * Bertl a few simple examples: 1119485105 M * Bertl you send a tcp chunk of 10kb out 1119485119 M * Bertl a) it is broken down according to your mtu 1119485132 M * Bertl b) packets might be retransmitted 1119485144 M * Bertl the socket accounting records 1/10k 1119485192 M * brc ahrm ok 1119485195 M * brc so if the packet is retransmited 1119485200 M * brc socket accounting wont count it 1119485213 M * brc so i guess it would be better for me to use iptables 1119485228 M * Bertl yep, neither will it account for the framing (i.e. ip/tcp, whatever) 1119485245 M * brc ok 1119485251 M * Bertl it's the amount of data the 'user' transmitted, not the system 1119485257 M * brc Another doubt 1119485258 M * brc last one 1119485288 M * brc remember that problem, when one of the vservers was stoped and the ip was taken down, all others were gone to 1119485307 M * brc when we did a vserver down on the one that has a global scope 1119485321 M * brc what was the solution to this prbolem. i forgot 1119485328 M * brc was there a way of making every IP global 1119485330 M * brc ? 1119485336 M * Bertl two solutions: 1119485352 M * Bertl a) reserve a dummy ip for the host, assign and keep it 1119485367 M * Bertl b) use /32 netmasks, they are always separate 1119485375 M * brc good idea 1119485416 M * brc thanks a lot bertl 1119485420 M * brc i think that is it .. hehe 1119485422 M * Bertl yw 1119485507 Q * DigitallyStoned Quit: Off like a prom dress! 1119485546 M * Bertl Aiken: it's a problem with the signal handler (i.e. the unwinding) 1119485621 M * Bertl Aiken: do you have an unpatched (vanilla) kernel to test with? 1119485722 M * Aiken if I dare venture into the cold I can boot that machine with a nomal kernel, back shortly 1119485802 M * Bertl huh, cold? where is it located? 1119485939 M * Aiken east coast australia, we are in winter 1119485960 M * Bertl cool! pun intended :) 1119485971 M * Aiken still segfaults with a std 2.6.11.10 1119485982 M * Bertl okay, that's what I wanted to know ... 1119486002 M * Bertl IIRC, the signal handler was changed slightly in the kernel (some time ago) 1119486018 M * Bertl it was supposed to be just fine for userspace 1119486050 M * Bertl but I'm pretty sure nobody tested it with anything else but glibc on alpha (if they tested at all) 1119486054 M * Aiken how long ago? I also have a 2.6.8.1 kernel for that machine 1119486076 M * Bertl hmm, that could be interesting, IIRC it was past 2.6.8 (sec, checking that) 1119486220 M * Bertl yup, was in 2.6.10 1119486273 Q * Jani Quit: 1119486550 M * Aiken I don't have a 2.6.8.1 after all. I'll compile a kernel of that version and see what happens. I have no other ideas at the moment 1119486569 M * Bertl okay, I'm looking into the code right now ... 1119488253 M * brc bertl, just found out that there is a per vserver loadavg at /proc/virtual/xid/cvirt.. great :) 1119488265 M * Bertl hehe :) 1119488294 M * Bertl brc: if it starts to make you feel bad, you know we have a 'donations' page :) 1119488412 J * Hollow ~Hollow@home.xnull.de 1119488418 M * Hollow heya 1119488419 M * Bertl welcome Hollow! 1119488492 M * Hollow hey Bertl, i have a problem with the syscall wrapper... e.g. if i call ctx_create for a xid which already exists, the syscall should set EEXISTS no? 1119488536 M * Bertl depends, either that or EBUSY ... 1119488537 M * brc Bertl: heheheheh!! as soon as i get money i will :) 1119488559 M * Hollow Bertl: but it always segfaults... 1119488574 M * Bertl then you are doing something wrong, no? 1119488606 M * Hollow the code is: 1119488606 M * Hollow int rc = vserver(VCMD_ctx_create, xid, NULL); 1119488606 M * Hollow if (rc < 0) perror("vc_ctx_create"); 1119488677 Q * shuri Remote host closed the connection 1119489159 M * Bertl http://vserver.13thfloor.at/Stuff/hollow.c 1119489181 M * Bertl in combination with the API-2.0 stuff, and the following compile command: 1119489194 M * Bertl diet gcc -Wall -O2 -Ivserver -I. -o hollow hollow.c 1119489219 M * Bertl results in the 'expected' ./hollow 1119489219 M * Bertl ctx_create: Operation not permitted 1119489232 M * Hollow hm yeah.. just testes sth similar.. semms to fail somewhere else 1119490172 J * rs ~rs@imhotep.rhapsodyk.net 1119490184 M * Bertl evening rs! 1119490198 M * rs evening 1119491657 J * Aiken_ ~james@tooax7-109.dialup.optusnet.com.au 1119491677 M * Bertl wb Aiken_! 1119491759 M * Aiken_ seems the 5 hourly disconnect from the isp just happened 1119491976 Q * Aiken Ping timeout: 480 seconds 1119495557 Q * Doener_zZz Read error: Connection reset by peer 1119495787 J * Doener ~doener@p54876EA9.dip.t-dialin.net 1119495795 M * Doener morning folks 1119495802 M * Bertl morning doener! 1119495862 M * Doener hey Bertl 'sup? 1119495873 M * Bertl another dietlibc issue 1119495885 M * Bertl rt_sigaction() is broken on alpha 1119496251 M * Doener hm, another asm beast? 1119496279 M * Bertl well, it seems that the alpha folks are funny guys ... 1119496292 M * Doener oh wow, it's C code :) 1119496318 M * Bertl they have a 'special' rt_sigaction() syscall, which takes an additional argument ... 1119496325 M * Bertl a so called 'restorer' 1119496492 M * Doener hm, yeah, they moved that from the userspace struct to the kernelspace struct... 1119496805 M * Bertl the funny thing is, somehow strace/gdb shows just wrong information 1119498144 J * Doener` ~doener@p54877999.dip.t-dialin.net 1119498438 M * Doener` Bertl: hm, in dietlibc's struct sigaction for alpha, the order of sa_flags and sa_mask is siwtched 1119498478 M * Bertl yes, I saw that, is it different from the kernel's version? 1119498514 M * Doener` if cscope wouldn't scroll all the way down in the results of my search i could tell you... :/ 1119498543 M * Doener` ok, there we go... 1119498551 M * Doener` yep 1119498582 M * Doener` kernel has: sa_handler, sa_flags, sa_mask. in that order 1119498583 Q * Doener Ping timeout: 480 seconds 1119498589 M * Doener` bye bye me! ;) 1119498684 M * Doener` OTOH, the userspace struct is always order in some other way it seems... 1119498710 M * Doener` but i fail to see how that copy_from_user is supposed to work then... 1119498740 M * Doener` hm, anyways, got to go, have a good one, and good luck! 1119498745 M * Bertl well, no problem, I can try to switch them :) 1119498752 N * Doener` Doener|gone 1119499043 M * Bertl Aiken_: still around? 1119499063 M * Aiken_ yes 1119499078 M * Bertl Doener provided an excellent hint it seems ... 1119499107 M * Bertl try the following, get the dietlibc source tree (latest version 0.29) 1119499150 M * Bertl edit the file include/signal.h (line ~446) so that 1119499159 M * Bertl unsigned long sa_flags; 1119499159 M * Bertl sigset_t sa_mask; 1119499168 M * Bertl (not the other way round) 1119499184 M * Bertl it should look like this then: 1119499187 M * Bertl #if defined(__alpha__) 1119499187 M * Bertl union { 1119499187 M * Bertl sighandler_t _sa_handler; 1119499187 M * Bertl void (*_sa_sigaction)(int, siginfo_t*, void*); 1119499187 M * Bertl } _u; 1119499190 M * Bertl unsigned long sa_flags; 1119499192 M * Bertl sigset_t sa_mask; 1119499195 M * Bertl #elif defined(__ia64__) || defined(__hppa__) 1119499215 M * Bertl then compile and isntall the dietlibc (i.e. use it for your test example) 1119499222 M * Aiken_ ok 1119499279 M * Bertl I still don't know _why_ it works without the restorer ... but it seems to do the trick here ... 1119499288 M * Bertl (maybe some kernel 'default' restorer ... 1119499307 M * Aiken_ so it is just swapping sa_flags and sa_mask? 1119499411 M * Bertl yep, well I think so, in my version I modified a bunch of other things too ... 1119499424 M * Bertl will check that now ... 1119499645 M * Aiken_ :) 1119499663 M * Aiken_ it works 1119499768 M * Bertl excellent! 1119499800 M * Bertl well, a lucky chap, we search for a whole night, and he has it at the first glance ... 1119499915 M * Bertl Aiken_: okay, could you please submit the change as diff to the dietlibc folks? 1119499938 M * Aiken_ ok 1119499941 M * Bertl (and send a copy to the linux-vserver ml, so other folks know where to look) 1119499976 M * Aiken_ are there many trying this in alpha? 1119499979 M * Bertl btw, the 20x errors will probably get a workaround soon ... 1119499992 M * Bertl IIRC, martin was trying some time ago ... 1119500063 M * Aiken_ just rebuilding the tools, I want to see a versver start with out needing lockfile compiled against glibc 1119500111 M * Bertl probably ... 1119500148 M * Bertl it's good to see that more and more folks start using linux-vserver on non x86 platforms ... and it pays off that we decided to keep everything as arch independant as possible 1119500185 M * Bertl only the dietlibc is a real issue on most archs ... 1119500206 M * Bertl but I guess we will have fixed that soon :) 1119500287 M * Aiken_ not sure how my poor little sparc IPX would handle this 1119500302 M * Bertl sparc was already tested in several cases 1119500318 M * Bertl it worked in my living room for a few days :) 1119500342 M * Bertl alpha was one of the last archs we had not testing done ... 1119500380 M * Bertl well, arm, m68k, v850 and mips is still missing ... 1119500558 M * Aiken_ apart from rc3.d entries I should remove hoppy starts and stops cleanly 1119500697 M * Bertl well, you might consider switching to sysv init style instead of plain 1119500751 M * Bertl okay, I guess it's working for you so far ... so I'm off to bed now ... will be back this evening ... 1119500767 M * Bertl Doener|gone: great work! thanks a lot! 1119500776 N * Bertl Bertl_zZ 1119500798 M * Aiken_ I started with rh4.1 and have stayed with that style of init system 1119501681 J * xbing ~nb@dsl081-044-121.lax1.dsl.speakeasy.net 1119501706 M * xbing testing 1119508781 J * Aiken ~james@tooax6-142.dialup.optusnet.com.au 1119509048 Q * Aiken_ Ping timeout: 480 seconds 1119509539 J * erwan_ho ~erwan@konilope.dyndns.org 1119509924 Q * erwan_ho Remote host closed the connection 1119510728 J * sukria ~sukria@sargon.lncsa.com 1119511016 M * meebey is the 2.6 branch kinda stable? 1119511024 M * meebey only basic feature set, nothing fancy 1119511093 M * meebey we are upgrading our servers to debian/sarge thus 2.6 kernel is interesting 1119511149 M * id 2.6 rocks - it is stable 1119511156 M * id as far I can tell you 1119511167 M * meebey same goes with vserver for 2.6? 1119511169 M * id with amd64 1119511173 M * id yes 1119511180 M * meebey I know that 2.6 is stable, but I wasn't sure about vserver 1119511195 M * meebey I remember the vserver days where you even couldn't start a vserver on 2.6 ;) 1119511206 M * meebey I tried one of the first patches at that time 1119511225 M * id hehe - grab the latest version 1119511230 M * id should work fine 1119511235 M * meebey ok I will give it a try 1119511241 M * meebey can I use the same tools? util-vserver for it? 1119511248 M * meebey I like then, they are simple and work :) 1119511250 M * id just be shure to compile the util-vserver 1119511250 M * meebey s/then/them 1119511266 M * id newest version 1119511294 M * meebey "[VServer 2.0] coming soon... (rc4 available)" 1119511304 M * id yeah 1119511306 M * meebey so the 2.6 branch is becoming stable (official) that means? 1119511324 M * meebey sounds good :) 1119511330 M * meebey our servers will love it 1119511348 M * id definatly 1119511386 M * meebey hrm question though is if the proc patch went in that 1119511391 M * meebey maybe I have to port it... 1119511408 M * meebey I have applications inside vservers which need write access to proc 1119511431 M * meebey with 2.4 vserver it was not possible, it had a hard check for it 1119511438 M * meebey s/hard/hardcoded/ 1119511452 M * DaPhreak hmm .. no idea if a vserver has write-access to the unhidden entries in /proc 1119511468 M * id vprocunhide ? 1119511486 M * meebey with the proc patch I have, I can allow write access with a CAP 1119511505 M * meebey hm no, it was something different, one sec 1119511634 M * meebey it is a flag 1119511664 M * meebey I will try it 1119511706 M * meebey eeeks again a new kernel released 1119511716 M * meebey 2.6.12.1 1119511745 M * meebey I bet the patch will not fit perfectly :) 1119511788 J * Nik ~Nik@193.193.162.69 1119511792 M * Nik hi all 1119511860 M * id hi 1119512576 M * DaPhreak meebey: naah, only the makefile rejects ;) and thats pretty simple to correct ;P 1119513029 M * meebey I don't like to mess with the vanialla kernel :) 1119513567 M * DaPhreak meebey: then grab this one ;P 1119513568 M * DaPhreak http://phreak.xnull.de/kernel/patches/vserver/patch-2.6.12.1-vs2.0-rc4.diff 1119513620 M * meebey DaPhreak: thanks 1119513654 J * prae ~prae@ezoffice.mandriva.com 1119513897 M * DaPhreak no problem ;) 1119513906 M * DaPhreak morning prae ;p 1119514277 M * prae morning Da :p 1119514310 M * Hollow morning 1119514342 M * DaPhreak morning Hollow ;) 1119514353 M * Hollow moin DaPhreak 1119514593 M * meebey DaPhreak: hm that is your config I think :) 1119514625 M * meebey DaPhreak: it has IBM thinkpad extras acticated etc, I don't think the default confing has them activated 1119514648 M * DaPhreak hmm .. there is no config 1119514653 M * meebey or is that again linus' hardware? 1119514926 M * meebey many experimetal things are on too 1119515766 Q * rs Quit: rs 1119515862 J * rs ~rs@imhotep.rhapsodyk.net 1119515865 M * DaPhreak well, its just plain unpacked 12.1 meebey 1119515902 M * meebey k 1119515925 M * DaPhreak try the plain and see if theres all that stuff also 1119515970 M * meebey its not that important 1119515983 M * meebey I guess its the linus' default, was the same with 2.4 :) 1119516085 Q * Aiken Read error: Connection reset by peer 1119516093 J * Aiken ~james@tooax6-142.dialup.optusnet.com.au 1119516440 Q * case Ping timeout: 480 seconds 1119516569 J * tbenita ~tbenita@atreal.net2.nerim.net 1119516575 M * tbenita Hi all ! 1119516620 M * tbenita I have a curious issue with squid : inside a vserver it seems that squid doesn't answer to requests 1119516647 M * tbenita it receives the client request, contacts the server proxied and doesn't answer the client 1119516662 M * tbenita any idea on what's going wrong ? 1119517649 M * TheSeer firewall? 1119517704 M * tbenita there is no firewall rule 1119517752 M * tbenita and I just installed the proxy wwwoffle that just works fine ... So there seems to be something else ? 1119517782 M * tbenita Do somebody use squid inside a vserver ? 1119517952 M * Nik tbenita: try to strace it and look for errors. might be a capabilities-problem like e.g. bind has. 1119517972 M * tbenita ok thx Nik 1119519771 J * Duckx ~duckx@195.75.27.158 1119519789 Q * rs Quit: rs 1119520224 J * hiaslbot ~hiaslbot@83.97.113.18 1119522057 Q * sukria Quit: Leaving 1119522989 J * rs ~rs@ATuileries-153-1-24-237.w82-123.abo.wanadoo.fr 1119523781 Q * Aiken Quit: Leaving 1119524339 J * Jani ~mail@G915e.g.pppool.de 1119524351 M * Jani Hi all 1119524731 Q * Doener|gone Read error: Connection reset by peer 1119524741 J * Doener ~doener@p54877999.dip.t-dialin.net 1119524752 Q * Nik Quit: My damn controlling terminal disappeared! 1119525051 Q * hiaslbot Remote host closed the connection 1119525157 J * hiaslbot ~hiaslbot@83.97.113.18 1119525298 J * rs_ ~rs@ATuileries-153-1-21-183.w82-123.abo.wanadoo.fr 1119525675 Q * rs Ping timeout: 480 seconds 1119525848 Q * hiaslbot Remote host closed the connection 1119525858 J * hiaslbot ~hiaslbot@83.97.113.18 1119526077 J * hiaslboy ~knoppix@pcla-10.mat.ucm.es 1119526227 Q * hiaslbot Remote host closed the connection 1119527108 M * Jani *Yippie* 1119527124 M * id hi 1119527127 M * id Jani, 1119527129 M * Jani vserver runs well on the new SuSe Machine :) 1119527131 M * Jani Heya i 1119527133 M * Jani +id 1119527164 M * id nice - i what is the SuSE slogan : "have fun..." ;) 1119527184 M * id -i 1119527290 M * Jani hehe 1119528244 J * tomi ~tomi@pha-84-242-95-4.nat.karneval.cz 1119528248 M * tomi Hi 1119528399 M * tomi how can I make new vservers 1119528401 M * tomi ?? 1119528419 M * tomi vserver name built ..... and?? 1119528502 J * eXplasm explasm@p549F7978.dip.t-dialin.net 1119528744 M * Jani vserver [name] build -m debootstrap --hostname [hostname] --netdev eth0 --interface [ip] --context 1001 -- -d sarge -m ftp://ftp.debian.org/debian 1119528752 M * Jani For creating a debian based vserver 1119528760 M * tomi ok.. 1119528958 M * Eyck yeah, it's not so good though :( 1119528975 M * Jani *hmms* Have worked for me? 1119529020 M * Eyck yeah, but afterwards there are like 12 different things you need to fix by-hand, that used to be corrected by debian-newsverver script 1119529055 M * Eyck plus, it creates 16-meg /tmp/ partition without asking, 1119529083 M * Eyck which will hunt you down if you don't notice that in time :((( 1119529189 M * Jani *hmms* would limite a vserver to use only 64MB RAM, right? 1119529200 J * hiaslbot ~hiaslbot@83.97.113.18 1119529299 M * tomi vserver quido start 1119529299 M * tomi capset(): Operation not permitted 1119529299 M * tomi capabilities are not enabled in kernel-setup 1119529299 M * tomi Failed to start vserver 'quido' 1119529302 M * tomi ??? 1119529311 M * tomi I have pached kernel 1119529666 M * id but i gues you forget to either load "modprobe capabilities" or you forget to compilie it statically 1119529717 M * tomi I have modular kernel.. 1119529738 J * jsambrook ~jsambrook@host-62-69-64-93.bsve.net 1119529744 P * jsambrook 1119529750 Q * hiaslboy Remote host closed the connection 1119529833 J * case ~case@donpanic.faveve.uni-stuttgart.de 1119530018 Q * Hollow Remote host closed the connection 1119530136 M * meebey has somone a list of required /dev nods for vservers? 1119530191 J * Hollow ~Hollow@home.xnull.de 1119530407 M * meebey hm linux-vserver.org seems to be a victim by kiddies 1119532210 M * daniel_hozac what do you mean? 1119532373 M * daniel_hozac and /usr/lib/util-vserver/distributions/defaults/devs contains a list of device nodes that's created by vserver - build 1119532679 J * sukria ~sukria@213.223.184.193 1119533194 M * _maharaja can you quickly point me to the debug_xxxx flags i have to build the kernel with? 1119533259 M * _maharaja for example MAGIC_SYSRQ 1119533458 Q * tomi Quit: Ukon�uji 1119535156 Q * case Remote host closed the connection 1119535555 Q * Jani Quit: 1119535915 J * jonsmel ~jscottorn@209.33.206.3 1119536158 J * case ~case@donpanic.faveve.uni-stuttgart.de 1119536159 M * case moins 1119536679 Q * Duckx Quit: Leaving 1119536685 Q * Hollow Quit: /dev/zero 1119537268 J * Hollow ~Hollow@home.xnull.de 1119537981 J * hvd ~takeagues@fw-grz.hollomey.com 1119538087 N * Bertl_zZ Bertl 1119538102 M * Bertl morning folks! 1119538154 M * Bertl _maharaja: most .config options related to debugging have DEBUG in their names 1119538168 M * Bertl also with a 2.6 menuconfig you can use the search option ... 1119538646 M * Doener morning Bertl! 1119538655 M * Doener so adjusting the struct worked? 1119538659 M * Bertl hey Doener! congrats for the quick solution! 1119538663 M * Doener thanks :) 1119538687 M * Bertl yeah, it did the trick ... I still don't understand why it works without the restorer, but it seems to do so ... 1119538698 M * Bertl or maybe just the strace info is wrong ... 1119538733 M * Doener hm, dietlibc always passes 0 as restorer, and as far as cscope is concerned that thing isn't used anyways 1119538767 M * Doener The sa_restorer element is obsolete and should not be used. POSIX does 1119538768 M * Doener not specify a sa_restorer element. 1119538778 M * Doener that's what the sigaction manpage has to say... 1119538827 M * Bertl okay, the big big question is, why does the alpha kernel/libc use that thing then? 1119538869 M * hvd Bertl: any timeframe when a diff against 2.6.12 is up ? (tried to merge but failed baldy) 1119538877 M * hvd s/baldy/badly/ 1119538880 M * Doener usually it is in the userspace sigaction struct and i guess for historical reasons, the alpha has it passed in the syscall as an argument instead 1119538883 M * Bertl three days ago! 1119538898 M * Bertl (we'll do some timetravel for 'ya) 1119538930 M * Bertl okay, back, here it is: http://www.13thfloor.at/~doener/vserver/patches/patch-2.6.12-vs2.0-rc4.diff 1119538958 M * hvd uih .. wow . ;) 1119538964 M * Bertl (actually we had to go back 5 days, but that's not a big deal :) 1119539027 M * Bertl but I agree, we are trying to hide them better recently ... 1119539047 M * Bertl (this will change in the near future) 1119539177 M * SiD3WiNDR when you simply don't publish them at all anymore? 1119539179 M * SiD3WiNDR or what ;) 1119539298 Q * hvd Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org 1119539304 M * Bertl SiD3WiNDR: precisely :) 1119539352 M * Bertl we are working at a scheme which will allow to put the releases in arbitrary places ... *G* 1119539477 M * id hi Bertl ;) 1119539477 Q * rs_ Read error: Connection reset by peer 1119539487 M * Bertl hey id! 1119539496 M * id are you going to visit LinuxTag ? 1119539511 M * Bertl nope, will be busy with ICFP 1119539585 M * Bertl meebey: ad 'victim' why do you think so? 1119539605 M * id schade, i hope i can get some time on frieday to get there - it only <50km away 1119539653 M * Bertl yeah, well, I planned to ... 1119539674 M * Bertl (didn't work out, their organization is pretty bad) 1119539726 M * Bertl s/organization/administration/ 1119539744 J * rs ~rs@ATuileries-153-1-21-183.w82-123.abo.wanadoo.fr 1119539751 M * Bertl hmm, wb rs! 1119539844 M * Bertl Doener: can you give me a quick overview of the vserver related changes in 2.6.12 (compared to 2.6.11x)? 1119539864 M * Doener i can try to ;) 1119539878 A * Bertl would appreciate it :) 1119539881 M * Doener most significant are the rss/anon(?) account macros 1119539897 M * Bertl yep, including a bug, which is already fixed, IIRC 1119539914 M * Doener then there's a change in the nfs code (the second pid/gid/xid stuff was removed) 1119539933 M * Bertl in which function/helper? 1119540080 M * Doener the BROKEN_SUID stuff was removed, let me check what the affected struct was 1119540222 M * Doener struct unx_cred, the uc_puid and such were simply removed, so i did the same to uc_pxid 1119540267 M * Bertl hmm ... is there any replacement for that? 1119540317 M * Doener nothing i know of 1119540415 M * Doener hm, forget the BROKEN_SUID thing for now, guess i mixed those two up 1119540424 M * Bertl k 1119540459 M * Doener one of the patches that were already in the vserver patch is now in mainline, so that one is removed 1119540470 M * Bertl excellent, which one? 1119540480 M * Bertl the socket flag? 1119540493 M * Doener PASSCRED or so? 1119540500 M * Doener is there something like that? 1119540505 M * Bertl yep, that's it ... thanks! 1119540506 M * Doener if so, that one is ti 1119540512 M * Doener s/ti/it/ 1119540544 M * Bertl yeah, davem was telling me that it will get in a few weeks ago ... 1119540562 M * Bertl wasn't sure it will be in 2.6.12 though ... 1119540577 Q * jonsmel Read error: Connection reset by peer 1119540609 M * Doener guess that's about it, at least regarding 'not so small' changes 1119540641 M * Bertl okay, let's look a little closer at the aut_cred stuff, I guess you didn't verify the changes there, or? 1119540663 M * Bertl *auth_unix to be precise 1119540702 M * Doener i checked all occurences of uc_puid and friends in 2.6.11.x and compared those function with 2.6.12-rcX 1119540721 M * Doener looked just like a simple complete removal of the code 1119540743 M * Bertl k, I see they 'vanished' do we have a patch (for mainline) which did that? something in the kernel changelog (like an explanation)? 1119540767 M * Bertl I'm pretty sure it's harmless, but this code is not much tested ... 1119540834 M * Bertl http://www.linuxhq.com/kernel/v2.6/11-bk10/net/sunrpc/auth_unix.c 1119540878 M * Doener http://www.frankvm.com/nfs-ngroups/2.6.10/README -- patch 20 1119540891 M * Doener well, and 10 ;) 1119540930 M * Bertl okay, sounds good ... feature removed, so we do not care ... 1119541094 M * Doener hm, timo's back ;) (ml) 1119541099 M * Doener (once again) 1119541099 M * Bertl yes? 1119541116 M * Bertl but I have to say, he improved lately ... 1119541139 M * Doener well, not in spelling your name ;) 1119541145 M * Doener "PLEASE Herbet or somebody else," 1119541152 J * hiaslboy ~knoppix@pcla-10.mat.ucm.es 1119541162 M * hiaslboy Hola Bertl :-) 1119541184 M * Bertl hey hiaslboy! 1119541194 M * DaPhreak lo guys ;P 1119541195 M * Bertl Doener: no, he is not talking to me, is he? 1119541218 M * DaPhreak heh, Bertl I guess he _is_ talking to you ;) 1119541223 M * Doener hm, right, martin is responsible for the list... 1119541233 N * Bertl Bertl_BOFH 1119541235 M * Doener but that's an even worse spelling then *g* 1119541240 M * DaPhreak yeah *g* 1119541249 M * DaPhreak he's going berserk *gg* 1119541258 M * Bertl_BOFH guess we have to ignore it for now *harhar* 1119541268 N * Bertl_BOFH Bertl 1119541272 M * DaPhreak ;) 1119541297 M * DaPhreak irssi++ 1119541307 M * Bertl ok, I'll send an email to martin to remove him from the list ... 1119541415 M * DaPhreak hmm off again :) (rear gymnastics) 1119541423 M * Doener have fun! 1119541429 M * DaPhreak yeah, thanks .. 1119541462 M * Bertl come back in once piece, yeah? 1119541613 A * SiD3WiNDR giggles at timo 1119542032 M * Bertl for german readers (http://vserver.13thfloor.at/Stuff//Timo_14062005.txt) 1119542176 M * id thanx - a good one ;) 1119542179 M * id lol 1119542225 M * Bertl so IMHO his intentions are good, but the methods and ideas ... 1119542307 Q * monrad Quit: Leaving 1119542344 M * Bertl Doener: don't you like to 'port' a simple patch from 2.6.11.x to 2.6.12? 1119542354 M * Doener which one? 1119542378 M * Bertl http://vserver.13thfloor.at/Experimental/FOR-2.0/delta-umarch-fix01.diff 1119542455 J * shuri shuri@64.235.209.226 1119542466 M * Bertl welcome shuri! 1119542482 M * shuri Hello Bertl 1119542682 M * hiaslboy do have anything against a bot you can ask here on the channel? 1119542726 M * Bertl not really, some folks started some bots IIRC ... 1119542749 M * Bertl but our demands to the bots were too high, I guess 1119542790 M * hiaslboy what are our demands :-P ? 1119542828 M * Bertl baiscally we know the lxr bot on kernelnewbies, and the infobot @ #gcc 1119542876 M * Bertl both do relatively sophisticated stuff ... i.e. finding things in the kernel trees, or reporting bug tracker info, etc ... 1119542923 M * Bertl we probably won't use a simple infobot, where you do: 1119542937 M * Bertl hey bot: Doener is smart! 1119542956 M * Bertl and the bot answers: but I heard, doener is something to eat :) 1119542976 M * id ;) 1119542985 M * Doener *lol* 1119542996 M * FaUl hrhr 1119543009 M * hiaslboy for me it records the talks and makes it searchable. 1119543010 M * FaUl hmm, doener :-) 1119543029 M * FaUl .oO( maybe i should go to the next doenersmith ) 1119543034 M * Bertl hiaslboy: http://irc.13thfloor.at/LOG 1119543054 M * Bertl and google is pretty good at searching, btw ;) 1119543099 M * hiaslboy I know, but I am trying to write a little more sophisticated tool to analyse a conversation. for that a talk in the chatroom is the ideal input :-) 1119543120 M * Doener like when we talk about a problem, it gives us the solution? 1119543125 M * Bertl hiaslboy: well, then take the logs, no? 1119543127 M * Doener that would be great :) 1119543162 M * Doener OTOH espdiff already does that 1119543241 M * Bertl btw, it seems the googlebot comes here every day ... 1119543532 J * erwan_taf ~erwan@81.80.43.67 1119544433 Q * eXplasm Remote host closed the connection 1119544491 J * eXplasm explasm@p549F7978.dip.t-dialin.net 1119544548 M * Bertl welcome erwan_taf! wb eXplasm! 1119544606 M * erwan_taf hey Bertl 1119544613 M * erwan_taf Bertl: how are you ? 1119544629 M * Bertl fine thanks! and you? 1119544654 M * erwan_taf fine too, not to much time for vserver but still listening about ;) 1119545002 M * Doener Bertl: hm, seems uml now includes the syscall tables from the 'real' archs... didn't come much further with my research 1119545013 M * Doener got to go now, back later (probably) 1119545018 M * Doener cya folks! 1119545020 M * Bertl okay, thanks, and cya! 1119545022 N * Doener Doener|gone 1119545815 Q * prae Quit: Execute Order 69 ! 1119545936 J * jonsmel ~jscottorn@209.33.206.3 1119545975 M * Bertl welcome jonsmel! 1119546114 M * jonsmel hi 1119546143 M * jonsmel I've been the one trying to get vserver to start on a lustre file system 1119546153 M * Bertl yep, I know ... 1119546162 M * jonsmel I'm working on getting upgraded to 2.6.12 right now 1119546214 M * jonsmel does it matter if i am running on 2.6.11.11 or 2.6.12 much 1119546231 M * jonsmel i have everything installed and ready to test on 2.6.11.11 1119546231 M * Bertl it might, a lot of things changed in 2.6.12 ... 1119546241 M * jonsmel but 2.6.12 is giving me a little grief 1119546254 M * jonsmel ok, i will push forward and get it all installed on 2.6.12 1119546259 M * Bertl the thing is, I need to get a 'working' setup too .. otherwise I can not check it ... 1119546266 M * jonsmel ok 1119546278 M * jonsmel what do you need 1119546278 M * Bertl so it would be helpful to know what I need to do to get lustre working ... 1119546294 M * jonsmel ok, not too dificult for the most part 1119546298 M * Bertl i.e. does it require patches? (probably) 1119546304 M * jonsmel yes 1119546317 M * Bertl what do they change? 1119546322 M * jonsmel basically what I have done thus far is patch in vserver to 2.6.12 1119546334 M * jonsmel mainly fs itmes 1119546351 M * jonsmel ext3 to be more exact 1119546391 M * Bertl why isn't lustre part of mainline? 1119546398 M * Bertl (just asking :) 1119546437 M * jonsmel supposedly the people developing the kernel did not want that much of an alteration to the ext3 fs or something like that 1119546467 M * Bertl so lustre is an extension of ext3? 1119546486 M * jonsmel basically, it is kinda like nfs on steroids 1119546488 M * Bertl where can I get the patch(es)? 1119546504 M * jonsmel i can ftp them to you 1119546515 M * jonsmel it is basically the patches and the utils all in one 1119546521 M * Bertl what about putting them somewhere where I can download them? 1119546535 M * jonsmel sure, just a sec 1119546568 N * BobR_oO BobR 1119546752 M * Bertl jonsmel: a) what timeframe do you see for the lustre stuff, and b) when are you usually awake/interested in working on that? 1119546772 M * jonsmel I usually have all day 1119546778 M * jonsmel during the day time that is 1119546800 M * Bertl and currently it's XX:XX for you? 1119546850 M * jonsmel 11:15am MST 1119546953 Q * tbenita Quit: Leaving 1119547030 M * Bertl k, because it's 19:17 here :) 1119547051 M * Bertl your work-day ends at 18:00 or later? 1119547079 M * jonsmel usually between 17 and 18:00 yes 1119547094 M * jonsmel I am usually here at 6 or 7 as well 1119547188 M * Bertl okay, I'll leave pretty soon, but I'll be back in 4 hours or so ... would be great if you could get the patches up somewhere and your test scenario working (with 2.6.12-vs2.0-rc4) 1119547208 M * jonsmel will do 1119547222 M * jonsmel i'm just getting them tar'd up right now 1119547229 M * jonsmel and put to a public place 1119547281 M * Bertl okay, perfect ... so if you find some time before you leave, we'll take a look at a test setup, good? 1119547363 M * jonsmel good np 1119547373 M * jonsmel They are almost ready to go 1119547394 M * Bertl I'm too ... but don't worry, I'll read the backlog ... 1119547412 N * Bertl Bertl_oO 1119547421 M * jonsmel cool 1119547423 M * jonsmel thnx 1119547440 N * BobR BobR_oO 1119547575 Q * erwan_taf Ping timeout: 480 seconds 1119548210 Q * shuri Remote host closed the connection 1119548406 Q * hiaslboy Ping timeout: 480 seconds 1119548712 M * jonsmel you still there bert? 1119548746 M * jonsmel you will have the link in your email 1119549717 J * erwan_taf ~erwan@81.80.43.67 1119550611 Q * sukria Quit: see you 1119552333 J * Psy0rz ~psy0rz@195.169.61.235 1119552394 M * Psy0rz what style to use with vserver 2.0 to boot a slackware vserver? 1119552398 M * Psy0rz initstyle 1119552747 Q * eXplasm Remote host closed the connection 1119552906 Q * rs Quit: rs 1119553073 M * pusling I have made som devicenodes in a vserver (/dev/hdd) but root in the vserver does not have permissions to do hdparm and stuff on hdd. is there anything else I need to do ?` 1119553161 M * Psy0rz pusling why would you do that in the vserver? 1119553175 M * Psy0rz the whole idea is that you can't use those tools inside the server 1119553246 M * pusling it is a test server for me and some friends... one of my friends wants a extra disk, so I plugged it in.... and he should have control of the disk. (less work for me ;) 1119553255 M * Psy0rz oooh 1119553266 M * Psy0rz i wonder if thats possible at all 1119553635 M * Psy0rz how to create a lo with 127.0.0.1 in each seperate vserver 1119553636 M * Psy0rz ? 1119553695 M * pusling i do not think that is possible 1119554022 M * daniel_hozac it is, with ngnet ;) 1119554168 J * hiaslboy ~matthias@23.Red-83-52-229.pooles.rima-tde.net 1119554464 J * tomi ~tomi@pha-84-242-95-4.nat.karneval.cz 1119554477 M * tomi Hi 1119554491 M * tomi I have trobles with ip.. in virtual server.. 1119555188 M * Psy0rz hmm doesnt seem to be a problem at all with 2.0 1119555211 M * Psy0rz just add a interface device in /etc/vserver/bla/interfaces/0/dev 1119555215 M * Psy0rz uh just add a interface device in /etc/vserver/bla/interfaces/1/dev 1119555221 M * Psy0rz soomthing like it 1119555760 J * Aiken ~james@tooax7-182.dialup.optusnet.com.au 1119556701 M * tomi now is ok.. 1119556714 M * tomi how can i managed new-vserver ip? 1119556724 M * tomi ifconfig show nothing 1119556735 J * are|linuxtag ~are@dsl-084-056-148-219.arcor-ip.net 1119556751 M * are|linuxtag Hi 1119556756 M * tomi hi 1119557171 J * yarihm ~yarihm@217-162-112-102.dclient.hispeed.ch 1119557954 J * newz2000 ~newz2000@12-226-91-204.client.mchsi.com 1119558166 M * jonsmel you set the ip within the /etc/vsererver/bla/interfaces/0/ip 1119558205 M * jonsmel same with netmask, brdcast, etc.. 1119560953 Q * newz2000 Ping timeout: 480 seconds 1119561209 N * Bertl_oO Bertl 1119561224 M * Bertl evening folks! 1119561318 M * jonsmel hey bert 1119561488 M * Bertl hey, I'm ready in a few minutes, just answering email 1119561505 J * rs ~rs@imhotep.rhapsodyk.net 1119561514 M * jonsmel ah, np still trying to get lustre running on 2.6.12 1119561869 M * Bertl under which license is the lustre stuff? 1119561994 M * jonsmel GPL 1119562024 M * Bertl do you know their homepage? 1119562035 M * jonsmel www.lustre.org 1119562056 M * Bertl tx 1119562065 M * jonsmel also with clusterfs.com 1119562154 M * Bertl okay, so they apply their patches with quilt .. and do they apply fine against 2.6.12 (without vserver)? 1119562216 M * Bertl l-patches-2.6.12.tgz <- are those the 2.6.12 patches? 1119562233 M * jonsmel yes 1119562249 M * jonsmel their patches are through bugzilla 1119562383 M * Bertl okay, so what's the point in using quilt? 1119562408 M * Hollow quilt is great 1119562426 M * Hollow i love doing the kernel patches with quilt 1119562431 M * Hollow it's easy and fast 1119562477 M * jonsmel they just suggest to use quilt to patch the kernel with some changes they need in the running kernel before you install 1119562495 M * jonsmel but I think it will do it by itself if you run make with Lustre 1119562514 M * Hollow hi Bertl btw 1119562518 M * Bertl hey Hollow! 1119562530 M * jonsmel I had problems trying to do it my own way so I followed their instructions on getting it installed 1119562536 M * Bertl jonsmel: okay, I don't plan to do so, I want to _know_ what needs to be applied 1119562582 M * jonsmel there is a series file within lustre/kernel_patches/series for 2.6.12 1119562591 M * jonsmel those are the patches that need to be applied 1119562599 M * Bertl yeah, I know quilt so that should be fine ... 1119562625 M * jonsmel once those are applied and you go to run the make 1119562626 M * Bertl is the 2.6.12 'lustre' version making problems, or the mixture of vserver and lustre` 1119562644 M * Bertl s/`/?/ 1119562645 M * jonsmel I think a little of both 1119562649 M * Hollow Bertl: I'm still fighting with these segfault issues.. you have a sec to look at a sample code? 1119562658 M * Bertl Hollow: sure ... 1119562676 M * Bertl jonsmel: okay, so you did not try a 2.4.12+lustre (no vserver) yet? 1119562677 M * Hollow Bertl: http://home.xnull.de/misc/vserver-apitest.tar.bz2 1119562679 M * jonsmel I think I almost have it running on 2.6.12 1119562681 M * Hollow just run make.sh 1119562747 M * jonsmel I have not tried just lustre and 2.6.12 yet 1119562752 M * jonsmel reason being.... 1119562787 M * jonsmel as soon as I patch the vserver in I have to reinstall Lustre which then leads me down the same path I am on 1119562823 M * Bertl aha, why is that so? 1119562847 M * Bertl Hollow: ever heard of gnu make? 1119562859 M * jonsmel haven't 1119562860 M * Hollow Bertl: yep and i use it.. 1119562866 M * Hollow just not in that example ;) 1119562902 M * Bertl jonsmel: what is the issue, why do you need to 'reinstall'? 1119562906 M * jonsmel whenever I change anything with the kernel and redo the modules, lustre gets wiped out and causes me to resetup 1119563044 M * jonsmel now if we want to try on 2.6.11.11 there are no issues there except vserver won't start 1119563083 M * jonsmel They just barely developed patched for 2.6.12 a couple of days ago, that is why there are so many bugs with them 1119563232 M * Bertl well, you do not know if 2.6.12/lustre works or not, right? but you state that it has bugs? 1119563255 M * Bertl somehow I can't really follow your argumentation ... 1119563296 M * jonsmel Let me back up then, I will quickly redo a fresh 2.6.12 kernel and lustre source for 2.6.12 and see if that goes 1119563303 M * jonsmel take me about 15 mins 1119563314 M * jonsmel without vserver patches 1119563340 M * Bertl Hollow: ./vtest 1119563340 M * Bertl ./vtest: error while loading shared libraries: libcreate.so: cannot open shared object file: No such file or directory 1119563358 M * Hollow has to be in in ld path 1119563361 M * Bertl maybe you are using an older version? 1119563372 M * Hollow cp libcreate.so /usr/local/lib/ 1119564645 M * Bertl Hollow: works fine once you enable optimization 1119564691 M * Bertl otherwise the perror stuff causes the segfault ... 1119564701 Q * rs Quit: rs 1119564765 M * Hollow Bertl: uh... and how could i know? :) 1119564783 M * tomi Bertl, how could I set up using ram for vservers.. 1119564796 M * tomi for expample 64MB 1119564814 M * Bertl you can 'limit' RSS, VM and VML ... 1119564824 M * tomi Yes 1119564830 M * tomi where? 1119564865 M * Bertl either with the vlimit tool, or with the config 1119564881 M * Bertl see http://www.nongnu.org/util-vserver/doc/conf/configuration.html for details 1119564903 M * Hollow hm.. which CFLAGS did you use? 1119564914 M * Bertl -O2 -g 1119564932 M * Hollow still segfaults here 1119564957 M * Bertl well, I didn't use your 'cryptic' make.sh 1119564968 M * Hollow instead? 1119564989 M * Bertl the cc lines ... 1119565023 M * tomi etc/vservers/vserver-name/rlimits 1119565023 M * tomi 1119565023 M * tomi A directory with resource limits. Possible resources are cpu, fsize, data, stack, core, rss, nproc, nofile, memlock, as and locks. This configuration will be honored for kernel 2.6 only. 1119565030 M * tomi this? 1119565040 M * Bertl yep 1119565052 M * Bertl rss, as and memlock 1119565066 M * tomi what are setting words? 1119565071 M * tomi inside file? 1119565082 M * Bertl you specify the pages IIRC 1119565105 M * Bertl have a look at the /proc/virtual//limits, it will show you the current values 1119565115 M * Bertl (and the observed maximum in the second column) 1119565164 M * tomi tihis table I must write to file? 1119565171 M * Bertl Hollow: gcc -Wall -g -o vtest vtest.c create.c (works too) 1119565187 M * Hollow yep, that's the point... 1119565202 M * Bertl so something with your shared lib is broken, no? 1119565207 M * Hollow yep 1119565208 M * tomi what is first nubmer and secound? 1119565290 M * Hollow Bertl: can you give me your makefile? 1119565305 M * Hollow or the command which built the lib 1119565556 M * Bertl no, because I did not use the library it seems 1119565568 M * Bertl well, you should first make a 'sane' makefile 1119565602 M * Hollow ok.. i'm no makefile expert.. ust learned some basic automake rules 1119565676 M * Hollow but i used a automake based library, where the segfault existed as well 1119565742 M * Hollow anyway... gtg to bed.. cu tomorrow.. 1119565808 M * Bertl k, cya 1119565820 M * tomi Bertl, can you send me some file with limits set up?? 1119565908 M * Bertl just do 'echo "100000" >/etc/vservers//rlimits/rss 1119565923 M * Bertl for example ... this will add an rss limit 1119565939 M * Bertl (on the next restart, it will be shown in proc too) 1119565956 M * tomi aha.. cool 1119566182 M * tomi files is quota for files? 1119566307 M * Bertl well, the nofile is the number of open file handles 1119566348 M * Bertl disk limits are not covered by the tool config yet, but they are supported by the tools (see vdlimit) 1119566453 M * jonsmel ok forgive my jumping the gun, patched with a fresh 2.6.12 kernel everything works fine 1119566471 M * Bertl okay, sounds good to me ... 1119566472 M * jonsmel I have it up and running right now without vservers 1119566488 M * Bertl good, give me a minute to check the patches 1119566494 M * jonsmel so I guess not the question is how to get the vservers patched in 1119566499 M * jonsmel ok, np 1119566556 M * jonsmel One though might be is that when I patch the vservers in it changes the version to whatever-vs2.0-rc4 1119566573 M * jonsmel and then in turn changes where the modules go for the running kernel 1119566585 M * jonsmel which is what causes me to resetup lustre 1119566606 M * jonsmel would that possibly be on track at all 1119566640 M * Bertl what modules? 1119566657 M * Bertl does lustre have out of kernel modules? 1119566693 M * jonsmel well, lustre when you do the make, goes and does a install from whatever you passed to it in the ./configure 1119566716 M * Bertl okay, let's for a moment forget about the 'lustre make' 1119566717 M * jonsmel ie... I did ./configure --with-linux=/usr/src/linux-2.6.12 1119566723 M * jonsmel ok 1119566730 M * Bertl we have a kernel, vanilla 2.6.12 ... 1119566736 M * jonsmel yep 1119566737 M * Bertl now we apply some patches ... 1119566741 M * jonsmel correct 1119566753 M * Bertl I saw that those patches seem to aptch stuff in luste too 1119566763 M * Bertl no idea what the intention behind that is/was 1119566763 M * jonsmel yes 1119566777 M * Bertl but I just applied the parts to the kernel for now 1119566787 M * jonsmel ok 1119566813 M * Bertl now I'd expect to find a 'lustre' option in the kernel menu, no? 1119566822 M * jonsmel lustre-1.4.2.1 was for rh and suse with 2.6.5 kernel 1119566826 M * jonsmel no there is not 1119566834 M * jonsmel they don't do that, don't know why 1119566844 M * Bertl so where does the lustre part come from? 1119566856 M * Bertl I just mount my ext3 as lustre or what? 1119566864 M * jonsmel basically 1119566871 M * jonsmel that is where the modules come into play 1119566880 M * Bertl which modules? 1119566902 M * jonsmel after you patch and reboot the kernel 1119566973 M * Bertl yes? 1119566975 M * jonsmel you then install lustre which during that it installs modules in the /lib/modules/what_ever_you_passed from ./configure 1119566997 M * Bertl okay, where do they come from, where is the module source? 1119567001 M * jonsmel so if the modules dir changes it wipes out lustre stuff 1119567014 M * jonsmel within the lustre source tree 1119567029 M * jonsmel ie.. lustre-1.4.2.1/lustre/ 1119567047 M * jonsmel there are a bunch of dirs where the modules come from 1119567082 M * jonsmel is that helping at all or am i just making a mess of this 1119567095 M * Bertl somehow I'm pretty confused by your tar actually, do they distribute the lustre in that format? 1119567108 M * Bertl I mean it contains various patches and tars and other stuff 1119567114 M * Bertl in all kind of places ... 1119567147 M * jonsmel sorry, i gave you a source that had all the patches and such that I was working on 1119567184 M * jonsmel I can put up the original sources if you want those 1119567202 M * Bertl yeah, please 1119567208 M * jonsmel k half a sec 1119567212 M * Bertl and what's the 1.2 vs 1.4 issue? 1119567222 M * jonsmel what do you mean 1119567244 M * Bertl on their site they have 1.2 and 1.4 where 1.2 is called public, and 1.4 production? 1119567300 M * Bertl also I do not like the: "This release of Lustre is available to customers, partners, and evaluators." part of the 1.4 1119567463 M * jonsmel Not sure, I have been working with Johann and Evan getting this working and the software is GPL, if you get 1.4 they are begining to charge for consulting 1119567468 M * jonsmel that's as much as I know 1119567515 M * jonsmel Those are ready to go 1119567521 M * jonsmel same site 1119567529 M * jonsmel lustre-1.4.2.1.tar.gz 1119567541 M * jonsmel l-patches-2.6.12.tgz 1119567551 M * jonsmel lustre-2.6.12.patch 1119567579 J * rs ~rs@imhotep.rhapsodyk.net 1119567597 M * jonsmel those three files are the sources for everything 1119567646 M * Bertl evening rs! 1119567668 M * Bertl jonsmel: okay, and the license of all those files is GPL or similar? 1119567693 M * jonsmel yes 1119567764 M * jonsmel It covers the license in the COPYING txt file within the lustre source if you want to look at it 1119567767 M * Bertl okay, but the lustre-1.4.2.1.tar.gz also contains a bunch of (different) 'kernel patches' 1119567788 M * rs hi there 1119567793 M * jonsmel yeah, they are for different kernels 1119567822 M * jonsmel to get it to work, you have to patch in the lustre-2.6.12.patch with in the lustre source 1119567833 M * jonsmel that updates lustre to work with 2.6.12 1119567855 M * jonsmel and untar the l-2.6.12 patches within the lustre source as well 1119567878 M * jonsmel it will put the correct kernel patches under lustre/kernel_patches 1119567907 M * Bertl what does your 'working 2.6.12 luste' kernel output when you do: 1119567909 M * Bertl cat /proc/sys/kernel/tainted 1119567935 M * jonsmel 0 1119567938 Q * yarihm Quit: Leaving 1119567946 M * Bertl and what modules are loaded atm? 1119567968 M * jonsmel nfsd, exportfs, lockd, sunrpc, r8169 1119567983 M * jonsmel when I start up lustre though it loads a tone more modules 1119568012 M * Bertl okay, could you do that, and upload output of lsmod somewhere (e.g. pastebin)? 1119568034 M * tomi vdlimit is in MB 1119568077 M * Bertl jonsmel: let me explain why am I so picky about this, and why I do not really like what I see (yet) 1119568107 M * jonsmel it is on that site as lsmod.log 1119568152 M * Bertl the vserver kernel patches do change a lot of low level things (i.e. pid mapping and such) and we have almost 'zero' chance to get that right if I do not have the complete code (i.e. all pieces belonging to the kernel which are required for lustre to work) 1119568176 M * Bertl I do not care about the userspace part, on the contrary I'd like to avoid it where possible ... 1119568232 M * jonsmel ok 1119568257 M * Bertl the 'best' starting point IMHO would be a single 'kernel' patch, which I can apply to vanilla-2.6.12 and compile the kernel (including all lustre kernel modules) and then, put that probably monolithic kernel somewhere, you can download it, and your lustre setup will work as expected ... 1119568280 N * _maharaja maharaja 1119568318 M * Bertl so if the lustre folks think that building their kernel modules outside the kernel tree, that's fine for me, but we have to get it inside, to test and change stuff ... 1119568333 M * jonsmel ok 1119568362 M * jonsmel i am glad to help in any way, I am, however, not that well versed in code and such 1119568421 M * Bertl quite a bunch of modules ... 1119568432 M * jonsmel yes it is 1119568438 M * Bertl are they loaded with some 'magic' options? 1119568458 M * jonsmel nope, they get loaded when I do anything with lustre 1119568473 M * jonsmel right now I am just running their test commands 1119568687 M * tomi what is flgs in vdlimit ?? 1119568691 M * tomi flags 1119568705 M * Bertl it's unused for now ... 1119568728 M * tomi there is number ctx? 1119568736 M * tomi or what is on flags? 1119568755 M * tomi foe examps.. 1119568765 M * tomi I want limit 2G 1119568808 M * Bertl you have to specify the limit for a vserver (xid) _and_ a filesystem (as there can be more than one) 1119568824 M * Bertl you also ahve to mount the filesystem with tagxid, there is a howto about this 1119568848 M * Bertl http://vserver.13thfloor.at/Linux2.6/index.php?page=Per+Context+Disk+Limits 1119568862 M * Bertl it's a little outdated, but it should give you an idea 1119568914 M * Bertl jonsmel: ./configure --with-linux=../linux-2.6.12-l1.4.2.1-P1 1119568932 M * Bertl just stops with: Run make config in ../linux-2.6.12-l1.4.2.1-P1. 1119568941 M * Bertl which was already done ... 1119568980 M * jonsmel hmm 1119568988 M * Bertl nevermind, they check for the version file 1119568989 M * jonsmel does .config exist there 1119568995 M * jonsmel ah ok 1119569033 M * jonsmel what is ATTR_XID used for 1119569057 M * jonsmel within include/linux/fs.h 1119569059 M * Bertl propagating the xid information for files 1119569065 M * jonsmel hmm 1119569078 M * jonsmel I am just trying to patch in the vservers patch again 1119569103 M * jonsmel there are two rejects which ultimately be why vservers don't start on lustre 1119569124 M * Bertl well, I doubt that this is the only issue ... 1119569210 M * jonsmel vserver patch wants to add in #define ATTR_XID 8192 1119569218 M * jonsmel lustre has already patched in: 1119569228 M * jonsmel ATTR_RAW 8192 1119569239 M * jonsmel ATTR_FROM_OPEN 16384 1119569263 M * jonsmel what would happen in this case 1119569283 M * Bertl strange things :) 1119569294 M * jonsmel well that is one reject 1119569328 M * hiaslboy Good night - se�ores y se�oritas :-) 1119569333 M * jonsmel could XID value be changed 1119569335 M * Bertl night hiaslboy! 1119569344 M * Bertl jonsmel: yes, move it up to 32k 1119569356 M * jonsmel so 32768 1119569361 M * Bertl yep 1119569434 M * jonsmel next is in mount.h 1119569442 M * Bertl configure: error: Lustre does not support kernels with preempt enabled. 1119569447 M * Bertl is that a joke? 1119569457 M * jonsmel that is what we thought as well 1119569516 M * Bertl configure: error: Lustre requires that CONFIG_KALLSYMS is enabled in your kernel. 1119569517 M * jonsmel The bad thing is after testing the speed and reliabity out we want to stick with it over other fs 1119569520 M * Bertl another one ... 1119569535 M * Bertl did you test with GFS? (just asking) 1119569540 M * jonsmel yes 1119569553 M * Bertl and? performance is so much worse? 1119569553 M * jonsmel that is what we were using 1119569579 M * jonsmel performance was fine, we just kept dealing with cluster issues 1119569601 M * jonsmel couldn't reliably keep it running 1119569602 M * Bertl of what kind, if I may ask? 1119569645 M * jonsmel if one of the nodes went down it would cause the main server to lose the gfs mount 1119569661 M * jonsmel which in turn would cause all servers to go down 1119569703 M * jonsmel lustre I can shut off any system and no issues 1119569727 M * jonsmel that was what the biggest issue was for us 1119569767 M * Bertl llite,mds,lov,osc,mdc,obdfilter,fsfilt_ldiskfs,ost,ptlrpc,obdclass,lvfs,ksocknal,portals,libcfs (are those all lustre modules?) 1119569801 M * jonsmel yes 1119569962 M * jonsmel is all vserver stuff built into the kernel or are there some modules as well 1119569990 M * Bertl there are modules too, but you can decide if you want them as module or not 1119570012 M * jonsmel It is all under linux vserver under make menuconfig correct 1119570031 M * Bertl most of it .. the vroot is not (for example) 1119570053 M * jonsmel oh, ok 1119570063 M * hiaslboy say is the 9pfs like these filesystems or totally different? 1119570078 M * jonsmel I was going to see if I can just build it all in so that I don't have to mess with the modules 1119570090 M * Bertl you can do that 1119570097 M * jonsmel ok 1119570113 M * jonsmel one error i get when building the kernel is 1119570129 M * jonsmel __kcrctab___iget previously defined 1119570137 M * jonsmel within fs/inode.c 1119570149 M * jonsmel looking at that file 1119570211 M * jonsmel there is a line that says EXPORT_SYMBOL_GPL(___iget) 1119570215 M * jonsmel and 1119570225 M * jonsmel EXPORT_SYMBOL(___iget) 1119570236 M * Bertl well, remove the second one 1119570241 M * jonsmel ok 1119570285 M * Bertl do you need xid tagging in your setup? 1119570298 M * jonsmel what would it be for 1119570312 M * Bertl disk limits on shared partitions 1119570340 M * Bertl if not, I'd suggest to disable it .. 1119570354 M * Bertl (also for the nfs support) 1119570384 M * jonsmel well, basically what we use this for is mainly of course vservres that can have seperate databases, jboss servers, etc.. 1119570414 M * jonsmel users usually will not access the vservers 1119570422 M * jonsmel other than through the ewb 1119570425 M * jonsmel web 1119570479 M * jonsmel where would modules be set off fresh vserver patch 1119570482 M * tomi vdlimit --xid 1101 --set inodes_used=300000 /var/lib/vservers/qqq 1119570490 M * tomi does not work.. 1119570490 M * jonsmel that way I can go have them built in 1119570502 M * jonsmel I've got all under linux vserver 1119570577 M * Bertl tomi: what does it tell you? 1119570594 M * tomi Im make vdlimit 1119570601 M * tomi on 3g 1119570634 M * tomi /dev/hdv1 3100000 0 0 - / 1119570634 M * tomi none 16384 0 16384 0% /tmp 1119570645 M * tomi without 1119570703 M * tomi dev/hdv1 12491804 2609720 9247520 23% / 1119570703 M * tomi none 16384 0 16384 0% /tmp 1119570720 M * tomi where is mistake? 1119570726 M * Bertl check the output of 1119570732 M * Bertl vdlimit --xid 1101 /var/lib/vservers/qqq 1119570747 M * Bertl you will see that various settings are 'wrong'/missing 1119570775 M * tomi vdlimit --xid 1003 /var/lib/vservers/quido 1119570775 M * tomi vdlimit: vc_get_dlimit(): No such process 1119570802 M * Bertl that's because you are now using a different xid, which you did not configure at all 1119570830 M * tomi 1003 is ok.. 1119570839 M * Bertl yeah, it has _no_ limits set ... 1119570875 M * tomi vdlimit --xid 1003 --set space_total=3000000 /var/lib/vservers/quido 1119570875 M * tomi borg1:~/vdlimit-0.02# vdlimit --xid 1003 /var/lib/vservers/quido 1119570875 M * tomi 1003 /var/lib/vservers/quido 1119570875 M * tomi space_used=0 1119570876 M * tomi space_total=3000000 1119570877 M * tomi inodes_used=0 1119570879 M * tomi inodes_total=0 1119570881 M * tomi reserved=100 1119570883 M * tomi borg1:~/vdlimit-0.02# 1119570885 Q * hiaslboy Ping timeout: 480 seconds 1119570904 M * tomi what is inodes? 1119570925 M * Bertl the number of inodes, check 'df' vs 'df -i' 1119570969 M * tomi uido:/# df -i 1119570969 M * tomi Filesystem Inodes IUsed IFree IUse% Mounted on 1119570969 M * tomi /dev/hdv1 0 0 0 - / 1119570969 M * tomi none 32073 5 32068 1% /tmp 1119570969 M * tomi quido:/# df 1119570970 M * tomi Filesystem 1K-blocks Used Available Use% Mounted on 1119570972 M * tomi /dev/hdv1 3000000 0 0 - / 1119570974 M * tomi none 16384 0 16384 0% /tmp 1119570976 M * tomi quido:/# df 1119571002 M * tomi but on disk is 360MB full 1119571006 M * tomi i think.. 1119571017 M * Bertl if so, then you have to tell the kernel that 1119571030 M * Bertl by specifying the space_used value 1119571031 M * tomi 2.6.11.6 1119571061 M * Bertl --set space_total=3000000 space_used=360000 1119571132 M * tomi vdlimit --xid 1003 --set --set --set space_used=360000 /var/lib/vservers/quido 1119571132 M * tomi vdlimit: unknown limit in '--set' 1119571178 M * tomi quido:/# df 1119571178 M * tomi Filesystem 1K-blocks Used Available Use% Mounted on 1119571178 M * tomi none 16384 0 16384 0% /tmp 1119571178 M * tomi quido:/# df -i 1119571178 M * tomi Filesystem Inodes IUsed IFree IUse% Mounted on 1119571180 M * tomi none 32073 5 32068 1% /tmp 1119571182 M * tomi now is ok?