1119916807 M * shuri helloo Bertl 1119917753 M * Doener ok, works quite a bit better with select() 1119917964 M * Bertl good ... 1119918008 M * daniel_hozac poll is nicer, IMHO. 1119918060 M * daniel_hozac i can't remember why though... 1119918081 M * Bertl depends on the actual use ... 1119918106 M * Bertl poll is definitely better if you handle files and file I/O 1119918375 J * ehzone ~Michael@CPE000f9f4458f8-CM000a73a0f6ea.cpe.net.cable.rogers.com 1119918512 J * shuri_ shuri@64.235.209.226 1119918512 Q * rs Read error: Connection reset by peer 1119918512 Q * shuri Read error: Connection reset by peer 1119918569 M * Bertl welcome ehzone! 1119918648 J * rs ~rs@imhotep.rhapsodyk.net 1119918733 M * ehzone Bertl, hello, just looking around to see what's available on OFTC. I'll just sit back and listen for a while. 1119918753 M * Bertl np, enjoy your visit ... 1119919786 M * Doener ok, got to get up early tomorrow, so i'm off to bed now... got some basic concept that I'll finish tomorrow, hope to get at least the context setup done, based on the configuration scheme of the alpha tools... 1119919808 M * Bertl sounds good to me, have a good night then ... 1119919816 M * Doener i.e. have the daemon handle "vserver xxx start" except the network setup 1119919837 M * Doener night! 1119923197 M * Bertl guess I'm off to bed now too .. have a good whatever everyone .. cya tomorrow! 1119923238 N * Bertl Bertl_zZ 1119924023 Q * ehzone Remote host closed the connection 1119930039 Q * aba Read error: Connection reset by peer 1119930168 J * Doener` ~doener@p54873E51.dip.t-dialin.net 1119930239 J * aba ~aba@eos.turmzimmer.net 1119930602 Q * Doener Ping timeout: 480 seconds 1119931690 Q * anonymousc Ping timeout: 480 seconds 1119932989 Q * flock Ping timeout: 480 seconds 1119933574 Q * Aiken Ping timeout: 480 seconds 1119935613 Q * are|dinner Quit: Disconnecting 1119937576 J * are|dinner ~are@gateway-dsl.lihas.de 1119937584 N * are|dinner _are_ 1119939115 Q * monrad Quit: Leaving 1119941329 N * BobR_zZ BobR 1119941898 N * BobR BobR_afk 1119942263 J * eXplasm explasm@p549F7C0C.dip.t-dialin.net 1119942343 J * sukria ~sukria@zenon.apartia.fr 1119942742 J * anonymousc ~anonymous@staff.internode.com.au 1119943148 Q * _are_ Ping timeout: 480 seconds 1119943236 M * rt re 1119943665 M * rt anyone runnig djbs daemontools under a vserver with "new config"? 1119943959 Q * eXplasm Remote host closed the connection 1119944127 Q * Psy0rz Ping timeout: 480 seconds 1119944477 J * prae ~prae@ezoffice.mandriva.com 1119947577 J * _are_ ~are@gateway-dsl.lihas.de 1119947584 M * _are_ hi 1119948052 J * marlin ~jimbo@dsl-202-173-187-73.nsw.westnet.com.au 1119948161 Q * marlin Quit: 1119948764 M * Doener` morning! 1119949384 M * Doener` rt: do you experience any problems? 1119949414 M * rt Doener`, yes ... the daemontools won t start in inittab 1119949452 M * Doener` do you use init at all? 1119949472 M * rt i think so 1119949489 M * Doener` the 'default' is the uses sysv initstyle, which starts the vserver in runlevel 3(?) without running an init process in the vserver 1119949523 M * Doener` /etc/vservers//apps/init/style should contain "plain" if you want an init process in the vserver 1119949532 M * Doener` brb 1119949534 M * rt cool, i ll try 1119949538 M * rt thanks a lot 1119950514 M * Doener` back now... 1119951661 Q * sukria Remote host closed the connection 1119952083 M * rt my deamontools are up and runnig ;-) 1119952565 M * Doener` great! 1119952720 M * rt why is there a new config style? now, its harder to clone a vserver ... 1119952987 M * Doener` is it really harder? 1119953079 A * SiD3WiNDR wonders too :) 1119953122 M * sannes heh, have to use cp -a instead of just cp .. hehehe 1119953257 J * erwan_taf ~erwan@AToulouse-105-2-2-69.w217-128.abo.wanadoo.fr 1119953279 M * Doener` cp -a /etc/vservers/v1 /etc/vservers/v2; cd /etc/vservers/v2; sed -i "s/v1/v2/" name uts/nodename; echo `cat context`+1 | bc > nc; mv nc context; 1119953291 M * Doener` and then adjust the symlinks and network config 1119953348 M * Doener` the symlinks are needed AFAICT so there's no way around them, and the rest is just echo'ing into different files 1119954614 M * _are_ is the vserver patch adusted to kernel 2.6.12.1 already? 1119954854 M * rt hmm, ok not realy hard, but it is easier to change one .conf file 1119954909 M * _are_ i prefer the new structure, makes scripting somewhat easier for me 1119955051 Q * anonymousc Remote host closed the connection 1119955101 J * anonymousc ~anonymous@staff.internode.com.au 1119955101 Q * rs Quit: rs 1119955180 Q * erwan_taf Ping timeout: 480 seconds 1119955721 M * gregster hi 1119955754 M * gregster does anybody here know a debootstrap for amd64 ? 1119956343 M * Doener` _are_: doesn't the 2.6.12 patch apply? 1119956463 M * _are_ Doener`: seems I am to stupid tofind it, I find a ton of 2.6.11.n patches 1119956468 J * erwan_taf ~erwan@AToulouse-105-2-2-69.w217-128.abo.wanadoo.fr 1119956727 M * _are_ Doener`: I'd expected it to be at http://vserver.13thfloor.at/Experimental or at http://www.13thfloor.at/vserver/d_rel26/ 1119957234 J * sukria ~sukria@213.223.184.202 1119957284 Q * erwan_taf Ping timeout: 480 seconds 1119957456 J * Cain ~wilm.reic@dsl-084-060-037-208.arcor-ip.net 1119957481 M * Cain hi 1119957719 Q * Cain Quit: 1119957745 Q * Hollow Read error: Connection reset by peer 1119957801 M * _are_ Hi Cain 1119957892 J * Hollow ~Hollow@home.xnull.de 1119958443 J * monrad ~monrad@213083190130.sonofon.dk 1119958898 Q * sukria Quit: see you 1119958941 M * Doener` _are_: http://www.13thfloor.at/~doener/vserver/patches/patch-2.6.12-vs2.0-rc4.diff 1119959033 M * rt Doener`, i still have a problem (i know, sounds impossible ;-))): 1119959055 M * rt when i stop the vserver, ip adress still exsists on the server 1119959303 M * Doener` is the vserver really stopped? try: ls /proc/virtual/ and check that there is no directory for its xid 1119959658 J * sukria ~sukria@213.223.184.202 1119959700 J * erwan_taf ~erwan@AToulouse-105-2-2-69.w217-128.abo.wanadoo.fr 1119959888 M * _are_ Doener`: thanks 1119960222 Q * prae Quit: Execute Order 69 ! 1119960511 M * rt Doener`, when i stop the server, ls /proc/virtual/ shows only "info" 1119960520 M * rt but the ip is still assigned 1119960581 M * Doener` ok, what files are in /etc/vservers//interfaces ? (use ls -R, if more than 3-4 lines, user either pastebin or /msg) 1119960587 M * Doener` s/user/use/ 1119960661 M * rt ls -R 1119960661 M * rt .: 1119960661 M * rt apps flags fstab interfaces name run uts vdir 1119960661 M * rt ./apps: 1119960661 M * rt init pkgmgmt 1119960663 M * rt ./apps/init: 1119960663 M * rt style 1119960665 M * rt ./apps/pkgmgmt: 1119960667 M * rt internal 1119960669 M * rt ./interfaces: 1119960673 M * rt 0 dev 1119960675 M * rt ./interfaces/0: 1119960677 M * rt ip name prefix 1119960679 M * rt ./uts: 1119960681 M * rt nodename 1119960732 M * Doener` hmm looks like more than 3-4 lines... 1119960788 M * Doener` anyway, there's no nodev file... hum... 1119960876 M * Doener` try "vserver --debug xxx stop" to see whether there's a call to remove that address 1119961064 Q * erwan_taf Ping timeout: 480 seconds 1119961245 J * erwan_taf ~erwan@AToulouse-105-2-2-69.w217-128.abo.wanadoo.fr 1119961591 M * rt /bin/ip addr del 192.168.240.239/24 broadcast + label eth0:devoutgoing dev eth0 1119961591 M * rt RTNETLINK answers: Cannot assign requested address 1119961944 M * Doener` label is too long 1119961962 M * rt hmm, ok i ll try a short one 1119961970 M * Doener` the whole label may be 16 chars long (incl. the trailing \0 ) 1119961978 M * Doener` and that includes the "eth0:" part 1119962007 M * Doener` so you're one char over the limit ;) 1119962024 M * Doener` /bin/ip addr del 192.168.240.239/24 broadcast + label eth0:devoutgoin dev eth0 1119962028 M * Doener` that should work 1119962043 M * Doener` (label is automagically truncated at the time of creation) 1119962448 M * Hollow hey Doener` 1119962455 M * Doener` hi Hollow 1119962464 M * Doener` what's up? 1119962480 M * Hollow i still can't get rid of these segfaults 1119962488 M * Hollow i've put up another test app 1119962491 M * Doener` hm, please refresh my memory... 1119962496 M * Hollow with libtool and automake 1119962508 M * Hollow http://home.xnull.de/misc/vtest.tar.bz2 1119962520 M * Hollow install and run... it should perror() but it segfaults 1119962555 M * Doener` is there an uninstall make target? 1119962608 M * Hollow iirc yes.. but installs only 2 files 1119962633 M * Doener` but not the executable? 1119962652 M * Hollow it should install libctxcreate.so and vtest 1119962681 M * Doener` ah, in /bla/local/... 1119962688 M * Hollow yup 1119962744 M * Hollow if the app is implemented without the lib (i.e. vc_ctx_create() in vtest.c) the app doesn't segfault 1119962773 M * Doener` doener vtest # /usr/local/bin/vtest 1119962773 M * Doener` doener vtest # 1119962796 M * Hollow hm 1119962820 M * Hollow ah 1119962821 M * Doener` which kernel/arch? 1119962824 M * Hollow sure 1119962840 M * Hollow the VCMD didn't fail 1119962844 M * Hollow try as non-root 1119962900 M * Hollow 2.6.12.1/x86 1119964196 M * rt now i have a second vserver, an get the following error: 1119964196 M * rt set -- 192.168.240.238/24 broadcast + label eth0:rma5 dev eth0 1119964196 M * rt ++ /bin/ip addr add 192.168.240.238/24 broadcast + label eth0:rma5 dev eth0 1119964196 M * rt RTNETLINK answers: File exists 1119964229 M * rt ifconfig don t show me the adderess, so i think it is not assigned 1119964334 M * Doener` try with ip, as i told you, ifconfig can't show addresses without aliases 1119964760 J * ruuth VooDoo@topas.informatik.uni-ulm.de 1119964839 M * maharaja rt: afaik Doener means "ip show addr" 1119964894 M * rt ip addr show 1119964896 M * rt :-) 1119964905 A * Doener` prefers "ip a" 1119964906 M * Doener` ;) 1119964932 J * rs ~rs@Laubervilliers-151-13-4-57.w82-127.abo.wanadoo.fr 1119965241 M * maharaja yeah, sry 1119965243 M * maharaja typo :> 1119965257 M * maharaja "ip a" is a nice shortcut 1119965258 M * maharaja :) 1119965292 A * rt thinks that Doener` is lazy ;-) 1119965386 M * Doener` of course I am! time is money ;) 1119965398 A * Doener` which also means that i always lack both 1119965413 M * Doener` oops... 1119965417 A * Doener` .oO( which also means that i always lack both ) 1119965425 M * Doener` that's what i wanted :) 1119965530 M * maharaja :> 1119966203 J * rt_ ~rt@195.246.161.1 1119966203 Q * rt Read error: Connection reset by peer 1119966326 Q * rt_ Quit: 1119966338 J * rt_ ~rt@195.246.161.1 1119966586 M * Doener` DaPhreak: it seems like the syscall somehow fscks up errno, when used in a lib... 1119966603 M * Doener` eek! I meant Hollow 1119966660 M * Hollow Doener`: any workaround? 1119966684 M * Hollow is this related to bertls syscall wrapper or syscall in general? 1119966740 Q * erwan_taf Ping timeout: 480 seconds 1119966842 M * Doener` hm, older ones also don't work... 1119966971 M * Doener` no idea, guess you have to wait for Bertl_zZ 1119967280 J * erwan_taf ~erwan@AToulouse-105-2-2-69.w217-128.abo.wanadoo.fr 1119967341 N * BobR_afk BobR 1119967420 M * daniel_hozac Hollow: what happens if you don't include errno.h in ctxcreate.c? 1119967468 M * Hollow daniel_hozac: bertls syscall wrapper needs errno.h 1119967484 M * Hollow /home/bene/work/vtest/src/vserver.h:47: error: `errno' undeclared (first use in this function) 1119967707 J * stephenM ~stephen@user-2774.l6.c5.dsl.pol.co.uk 1119967727 M * daniel_hozac so what is errno set to after the syscall? 1119968095 Q * erwan_taf Ping timeout: 480 seconds 1119968132 M * jonsmel morning all 1119968368 M * Hollow daniel_hozac: e.g. EPERM 1119968376 M * Hollow if you call the syscall as non-root 1119968389 M * daniel_hozac not -EPERM? 1119968400 M * Hollow yep ;) 1119968469 M * daniel_hozac i think errno is supposed to be positive. 1119968495 M * daniel_hozac but glibc seems to handle the negative case gracefully... 1119968504 M * Hollow vserver(0x9010000, 0x4d2, 0, 0, 0x97fcbca0) = -1 EPERM (Operation not permitted) 1119968512 M * Hollow ah 1119968516 A * Hollow shrugs 1119969069 M * Doener` the syscall return -EPERM, but errno get the absolute value.. 1119969088 M * Doener` ret = syscall(...); if (ret < 0) errno = -ret; 1119969101 M * Doener` (not exactly like that, but you should get the idea) 1119969157 M * Doener` Hollow: you can get the segfault as root as well, just try with xid = -123 ;) 1119969160 Q * rs Quit: rs 1119969193 J * erwan_taf ~erwan@AToulouse-105-2-2-69.w217-128.abo.wanadoo.fr 1119969408 M * Hollow Doener`: or an exisiting one ;) 1119969417 M * Doener` right 1119969587 J * rs ~rs@Laubervilliers-151-13-4-57.w82-127.abo.wanadoo.fr 1119969803 N * BobR BobR_oO 1119970222 M * stephenM Doener`: with your 2.6.12 patch, are you aware of any issues? specifically with segfaulting kernel builds and errors of "insmod: can't read /lib/'? 1119970235 M * stephenM upon bootup* 1119970293 M * Doener` no, it's running on my box since 2.6.12 came out 1119970303 M * daniel_hozac i get that too... i thought it was caused by the Fedora patches. 1119970327 M * daniel_hozac are you using an initrd/initramfs? 1119970384 M * stephenM yeah 1119970472 M * Hollow 2.6.12.1 fine here too 1119970482 M * stephenM I'm using just .12 1119970497 M * stephenM Doener`: could you give me the link to the patch again pls? maybe I picked the wrong one :P 1119970512 M * Hollow hm, i have genpatches applied too 1119970520 M * Doener` http://www.13thfloor.at/~doener/vserver/patches/patch-2.6.12-vs2.0-rc4.diff 1119970530 M * stephenM ah, I didn't use -rc4 1119970531 M * daniel_hozac i'm based on 12.1 as well. 1119970541 M * stephenM do you know where I can get the patch for 12.1? 1119970565 M * daniel_hozac Doener`s patch worked well. 1119970578 M * stephenM for 12.1 as well as .12? 1119970582 M * daniel_hozac s/worked/applied/ 1119970590 M * Doener` stephenM: there's only a -rc4 patch for 2.6.12... 1119970599 M * daniel_hozac Hollow, Doener`: using initramfs? 1119970606 M * daniel_hozac (or initrd) 1119970613 M * Doener` no 1119970705 M * _are_ only reject when using Doener`s 2.6.12 patch against 2.6.12.1 is the extra version in the top level makefile 1119970725 M * _are_ and it compiled and vserver-stat gave reasonable output 1119970743 M * daniel_hozac _are_: using initramfs/initrd? 1119970784 M * _are_ nope, in the guest I have no use for it and in the host I am paranoid ;) 1119970816 M * _are_ fails by far to often for me to trust it and most of my servers are a few 10km away :-> 1119970825 M * Doener` hu? initrd in the guest? how's that supposed to work at all? ;) 1119970827 M * _are_ erm 'few 100km away' 1119970842 M * _are_ Doener`: as said: no use for it :-) 1119971156 M * daniel_hozac Doener`: can you use an initramfs with 2.6.12? 1119971315 J * bbkenshin ~bbkenshin@d213-103-22-109.cust.tele2.fr 1119971320 M * bbkenshin hi every one 1119971358 M * daniel_hozac insmod returned Bad address when it was trying to insert /lib/*.ko 1119971385 M * bbkenshin i'm workingg on Squid , and i've problem with ncsa_auth 1119971404 M * bbkenshin parseConfigFile: line 216 unrecognized: 'authenticate_program /usr/lib/squid/ncsa_auth /etc/squid/users' 1119971527 M * Hollow daniel_hozac: no initramfs here 1119971551 M * daniel_hozac so it's most likely related to that. 1119971603 M * Hollow do you need ramfs? 1119971639 Q * erwan_taf Ping timeout: 480 seconds 1119971684 M * Doener` daniel_hozac: to be honest, i have no idea about initrd/initramfs 1119971725 M * daniel_hozac yeah, i'm rebuilding the Fedora kernels and everything is a module ;) 1119972001 J * Jani ~mail@G956f.g.pppool.de 1119972002 M * Jani *waves 1119972124 M * Doener` welcome Jani 1119972163 N * Bertl_zZ Bertl 1119972173 M * Bertl morning folks! 1119972185 M * Jani Morning Bertl 1119972195 M * Hollow hey Bertl 1119972205 M * Bertl bbkenshin: any indication that it is linux-vserver related? 1119972227 M * jonsmel Morning Bertl 1119972232 M * Bertl hey Doener`, Jani, Hollow, jonsmel! 1119972238 M * bbkenshin mmmm linux mdk ........... 1119972242 M * Doener` morning Bertl! 1119972252 M * Jani *hmms* Have someone a idea where I can find a gentoo skeleton image? Friend Google haven't found anything. 1119972264 M * Jani Or it really seems that I should create one then ;) 1119972279 M * Hollow http://dev.gentoo.org/~hollow/vserver/gentoo-template.tar.bz2 1119972291 M * Hollow Jani: ^ 1119972303 M * Bertl Jani: "google: gentoo vserver guest image" 1119972307 M * Hollow i686 1119972363 M * Hollow Bertl: i implemented the vtest with makefiles etc, so it installs correctly, but still segfaults, Doener` said the syscall fscks up the errno... 1119972383 M * Doener` (at least it looks like) 1119972391 M * Hollow the new version is at http://home.xnull.de/misc/vtest.tar.bz2 1119972410 M * Jani Oh, Guest Image is the magicaly word :| 1119972419 M * Jani I have tried skeleton... 1119972421 M * Jani Ok, thanks :) 1119972435 M * Bertl Hollow: well, could easily be, you have to 'adjust' the syscalls for the libraries you use 1119972447 M * Hollow Jani: why don't you use a stage3? it's nearly a ready image 1119972449 M * Bertl (it was designed for the non threaded dietlibc use) 1119972459 M * Hollow ah 1119972479 M * Bertl but you can easily adjust that, there are wrappers for setting the errno 1119972479 M * Hollow so with dietlibc it should work? 1119972502 M * id Hi Bertl 1119972538 M * Bertl Hollow: it also seems to work quite fine for non pic glibc 1119972586 M * Jani Hollow: I haven't Gentoo as host system. 1119972599 M * Hollow mhm.. i need to adjust _syscall_error, right? 1119972607 M * Hollow Jani: doesn't matter 1119972631 M * Hollow as long as you have tar and chroot ;) 1119972655 P * bbkenshin Leaving 1119972776 M * jonsmel Bertl: Any luck with lustre, or is it hopeless? 1119972832 M * Bertl no, not hopeless 1119972848 M * Bertl the kernel changes are definitely 'broken' in some aspects 1119972849 M * jonsmel ah cool 1119972874 M * jonsmel what can I do to help 1119972875 M * Bertl but I'm pretty certain the modules need a few changes too 1119972883 M * jonsmel ok 1119972899 M * Bertl but I don't know how we/you want to proceed there ... 1119972932 M * Bertl from my point of view, the lustre is a bad choice, simply because it needs a lot of initial work ... 1119972947 M * Bertl it might be the better solution in the long run ... 1119973025 M * jonsmel ok, hmmm what do you think it will take to get the kernel portion fixed 1119973082 M * jonsmel i will work with Lustre to fix the modules 1119973085 M * Bertl as far as I am concerned, I do not want to spend too much (unpaid) time on it, but I'd suggest to proceed like this: 1119973107 M * jonsmel ok 1119973114 M * Bertl - move the lustre modules into the kernel and make them monolithic 1119973131 M * Bertl - separate everything lustre userspace specific from the kernel part (API) 1119973184 M * Bertl - clean up some changes which are not required in lustre 1119973200 M * Bertl (especially those which clash with vserver changes) 1119973227 M * Bertl - add the non fs specific parts of linux-vserver 1119973237 M * Bertl - add the fs specific changes 1119973260 M * Bertl (of course, between all those steps, some testing/regression checks should be done) 1119973268 M * jonsmel right 1119973302 M * Bertl I would say in total the first 3 steps should take about 1 man-month 1119973317 M * jonsmel gotcha 1119973327 M * Bertl the code doesn't seem especially ugly to me, so I see no big issues there 1119973362 M * Bertl once the modules are integrated and tested, the vserver integration should probably take less than that 1119973382 M * Bertl so including the testing, I'd say 2-3 mm at most ... 1119973396 M * jonsmel ok 1119973402 M * Bertl and of course, a lot of stuff can be parallelized 1119973488 M * Bertl I have no problem with either doing this on paid time or helping/commenting on kernel/vserver related stuff (unpaid) 1119973535 M * jonsmel well, that's good to know, Let me think about what direction I want to go with it. 1119973559 M * Bertl http://www.13thfloor.at/~herbert/delta-2.6.12-vs2.0-lustre2.diff.hl (this is the kernel patch) 1119973612 M * Bertl I identified the following (possible) issues: 1119973648 M * jonsmel now, even if I got all the lustre items built into the kernel, there would still be conflicts with fs items between vserver and lustre correct 1119973659 M * Bertl - it seems that open.flags and it_flags are somewhat synonyms (maybe not precisely) 1119973680 M * Bertl jonsmel: yes, I assume that a lot of issue result from the modules 1119973758 M * Bertl - sys_chmod() the removal of the permission code is wrong 1119973785 M * Bertl - the attributes need handling for both, the lustre and vserver attribs 1119973831 M * Bertl - the thread size cheat seems to me like papering over 4k issues :) 1119973933 M * jonsmel sounds like some of this stuff is going to be tricky 1119973946 M * Bertl ah, I have an easy one too :) 1119973980 M * jonsmel ah, do tell... 1119974040 M * Bertl hmm, can't find it atm ... sec 1119974078 M * jonsmel np, just looking for a upside :) jk 1119974079 M * Bertl ah, yes 1119974086 M * Bertl in fs/ext3/balloc.c 1119974105 M * Bertl the DLIMIT_FREE_BLOCK has to be added back (that's pretty easy) 1119974122 M * jonsmel :) 1119974137 J * erwan_taf ~erwan@AToulouse-105-2-2-69.w217-128.abo.wanadoo.fr 1119974182 M * Bertl evening erwan_taf! 1119974469 M * jonsmel Bertl: let me pose a question, Do you suggest any other options for dfs (distributed file system) besides gfs 1119974670 M * Bertl hmm, I know that drbd as block sync level works (more or less) but that's not a dfs 1119974691 M * Bertl nfs works pretty fine for most cases and is well supporte 1119974693 M * Bertl +d 1119974701 M * Bertl (also no dfs) 1119974739 M * Bertl from all the other fs I know (coda, afs, gfs) the only ever tested is the gfs 1119974777 M * Bertl and IIRC, somebody reported gfs working, might be someone from you folks ... don't remember ... 1119974847 M * sannes how can I mount something into an already running vserver without editing fstab and restarting? 1119974867 M * _are_ Bertl: nfs has issues when you use locking, have not managed to get it to run within a vserver 1119974868 M * sannes and why isn't this in the FAQ ? 1119974878 M * sannes :P 1119974882 M * daniel_hozac sannes: vnamespace -e mount... 1119974892 M * Bertl _are_: hmm, details 1119974908 M * sannes daniel_hozac : thanks 1119974911 M * Bertl sannes: just 'enter' the namespace and mount it 1119974937 M * _are_ Bertl: sorry, no access to the system atm, it had to do with File locking of an IRIX Client running Catia, it claimed the file was not avaliable all the time 1119974987 M * Bertl and that worked with a compareable non-linux-vserver kernel? 1119975024 M * Bertl (ah, and another important thing, did you use tagged nfs or not?) 1119975024 M * _are_ yes, ofc, it works even with the vserver patch if you ran the nfs server on the host instead of inside a vserver 1119975048 M * _are_ no tagged nfs, help file says I have to patch the client then and I doubt I get access to IRIX sources 1119975050 M * Bertl ah, you are running the nfs server inside a guest 1119975069 M * Bertl okay, that's probably not what jonsmel will do :) 1119975079 M * jonsmel correct 1119975102 M * _are_ yes, inside the guest NFS works fine as long as you don't try and lock anything, outside/on the host it just works, regardless if with or without vserver 1119975104 M * jonsmel hmm, well, I guess my search will continue 1119975126 M * jonsmel and yes, gfs was noted by us that it was working 1119975133 M * Bertl jonsmel: IMHO a good idea would be to look at coda 1119975137 M * jonsmel which we did have it working but then developed other issues 1119975141 M * jonsmel internal to gfs 1119975149 M * jonsmel ok 1119975153 M * jonsmel I'll look into it 1119975173 M * Bertl basically you will have to face the following issues with _all_ dfs 1119975205 M * Bertl - handling the ip restrictions (and permitting dfs network traffic) 1119975205 M * jonsmel btw, if you want to email me about what you think it would cost for us to have you fix lustre and vservers, that might be what still choose to do 1119975268 M * Bertl jonsmel: no problem, I can send an estimation of costs to you ... 1119975281 M * sannes Bertl: mount: permission denied, if I do it with enter .. and with vnamespace I can't get at the thing I want to bind mount 1119975292 M * jonsmel thanks 1119975299 M * Bertl - handling the additional filesystem attributes 1119975316 M * Bertl - handling the xid tagging of files and the permissions 1119975338 M * Bertl sannes: how so? (with namespaces) 1119975342 M * jonsmel Now if we do something like coda fs, have the vserver patches taken in account for that 1119975357 M * jonsmel since coda is builtin 1119975358 M * Bertl coda is in the kernel already 1119975406 M * Bertl so I can easily modify/test that, and if somebody is actively testing/using that, it is of interest to me 1119975421 M * jonsmel ok, that might be a possibility 1119975429 M * jonsmel let me read up on coda for a bit 1119975438 M * jonsmel I'll get back with you on it. 1119975442 M * _are_ I think I have intrest in coda, too, wanted to check it out for years anyway ;) 1119975475 M * Bertl don't get me wrong, I'm not against lustre, maybe you can convince the lustre folks (especially P Braams) to make an all-in-one kernel patch for 2.6.x ... 1119975488 M * jonsmel right 1119975510 M * Bertl with that present, it is equiv to grsec for me ... 1119975519 M * sannes Bertl : vnamespace -e vs1 mount -o bind,ro /root/testdir /root/fromhost 1119975520 M * jonsmel I have been talking with a few of them to try to get them to do it "correctly" :) 1119975549 M * Bertl sannes: try with vnamespace -e vs1 bash 1119975560 M * Bertl then look around and see what is where :) 1119975568 M * sannes Bertl : it obviously won't work, but that is what I want .. ok 1119975598 M * Bertl jonsmel: for me the following guidelines apply (atm): 1119975599 Q * erwan_taf Ping timeout: 480 seconds 1119975626 M * Bertl - is it in the kernel right now and somebody wants to use it? 1119975647 M * Bertl - can it be easily patched into the kernel and/or does it require tools? 1119975667 A * sannes looks around .. DOH! and thanks 1119975678 M * Bertl - is a test setup simple, and do a lot of folks want to use it/need it 1119975711 M * Bertl jonsmel: basically nfs was the first category, so it was done 1119975773 M * Bertl the lustre currently doesn't fulfill any of them ... 1119975817 M * Bertl it might also be interesting to fix up gfs, which would at least qualify for the second one, no? 1119975818 M * jonsmel ok, I will see what I can do, I think that this will be a very useful way of combining vservers with dfs to provide basically no down time with these vservers 1119975830 M * jonsmel right 1119975857 M * jonsmel or like you said to look at coda or afs 1119975864 M * jonsmel which I currently am 1119975879 M * Bertl okay, good, keep me informed about that stuff ... 1119975901 M * _are_ jonsmel: I do it with heartbeat, drbd, vserver 1119975917 M * _are_ you can make about everything redundant thatstores data on disks 1119975918 M * jonsmel cool, thanks for all the info and help bert, ps.. please email me what you think it would cost as well for lustre 1119975956 M * jonsmel heartbeat, drbd ?? haven't heard of them 1119975963 M * _are_ for a samba server the failover time is <2m, nfs might be a bit tricky, depending on how the clients mount the export 1119975976 M * Bertl jonsmel: yes, will do you, expect an email in a few hours ... 1119975982 M * Bertl s/you*so/ 1119975995 M * jonsmel thanks bert 1119975998 M * _are_ jonsmel: www.linux-ha.org should be the domain to get info, it is tools to do failover between machines 1119976011 Q * Jani Ping timeout: 480 seconds 1119976013 M * Bertl yep, drbd works on the block level 1119976038 M * Bertl current implementation does support only 2 nodes and the fs has to handle the failover 1119976047 M * _are_ and vserver just makes the live a few doozen times easier as you can just run a 'normal' server for enything and it magically is redundant 1119976087 M * _are_ well, atm I have a box that runs 2*UML, the UML runs vserver, drbd, heartbeat to demonstrate this, works really nice 1119976141 J * erwan_taf ~erwan@AToulouse-105-2-2-69.w217-128.abo.wanadoo.fr 1119976165 M * Bertl okay, have to leave now .. will be back later ... 1119976173 N * Bertl Bertl_oO 1119976217 M * sannes Bertl : does that mean that vnamespace has both namespaces available, and doesn't change what is root? 1119976262 Q * BWare Quit: using sirc version 2.211+KSIRC/1.3.12 1119976328 M * sannes Bertl : or that a namespace of a vserver includes the host namespace? 1119976330 M * _are_ ot doesn't change into the chroot, if I remember correctly 1119976422 M * sannes pivot root with 2.6? 1119976506 M * _are_ pardon? 1119976546 M * sannes well, if it uses chroot, the barrier still has to be set, right? 1119976615 M * sannes but I thought we didn't need that anymore with 2.6 vserver? 1119976720 M * _are_ eh? I think this is different things 1119976747 M * _are_ the namespace is a partition in the kernel, the chroot is a partition in the filesystem, the vserver-tools bring this together 1119976764 M * _are_ you can use one without the other, just in most cases you don't. 1119977351 M * Doener` sannes: alpha tools create a namespace, setup the mounts at /vservers/foo/... and then rbind mount the whole vserver stuff to /, followed by a chroot 1119977416 M * Doener` sannes: and there you go for docs: http://linux-vserver.org/Namespaces ;) 1119977482 M * Doener` when you enter the namespace, you get the pre-chroot state, i.e. you're still on the old / (or somewhere below it) not the 'new' / 1119977729 M * sannes so if I mount something in the host after I have started the vserver, would it still be visible? 1119977771 M * sannes in the vnamespace -e 1234 case? 1119977885 M * _are_ to my understanding: yes, and if I remember right, I already did it that way 1119978671 M * sannes how come? when they are separate namespaces? then it is "seeing" two namespaces? 1119978753 A * sannes is trying to understand this .. heh, hope I am not bothering anyone .. 1119979174 M * _are_ no, the namespace is the same, but it is not in chroot 1119979193 M * _are_ as said: kernel namespace and filesystem chroot are independent in the first instance 1119979237 M * _are_ gtg shopping and home, cu later 1119979240 Q * _are_ Quit: bbl 1119979265 Q * erwan_taf Ping timeout: 480 seconds 1119979406 Q * sukria Ping timeout: 480 seconds 1119979746 Q * sannes Read error: Connection reset by peer 1119980880 M * daniel_hozac hmm. wasn't the named issue supposed to be gone with 2.0-rc? 1119980953 M * stephenM in order to switch to 2.6.12.1 do I need to change anything in /lib? I keep getting errors of "insmod: can't read '/lib/': Bad address" 1119980962 M * stephenM and then the kernel panics 1119981108 M * stephenM it's fine with 2.6.11.5-vs1.9.5 1119981465 M * Doener` that's the lib in the initramfs 1119983785 J * _are_ foobar@dsl-084-056-137-250.arcor-ip.net 1119984112 M * stephenM Doener`: any idea how to fix? 1119984144 M * _are_ Hi 1119984179 M * Doener` no 1119984182 M * Doener` wb _are_ 1119984200 M * Doener` i never used initrd/initramfs 1119984253 M * Doener` stephenM: you should probably just ask on the ml, maybe someone knows 1119985047 Q * ruuth Quit: Nettalk6 der Freeware IRC-Client 1119985198 Q * Loki|muh Quit: Lost terminal 1119985264 J * Loki|muh loki@satanix.de 1119986734 Q * rs Quit: rs 1119987058 J * sannes ~ace@cm-84.118.217.070.chello.no 1119987303 Q * shuri_ Remote host closed the connection 1119987382 M * sannes is vps supposed to work in 2.6 vserver? 1119987437 M * _are_ yes, an dit works 1119987448 M * Doener` yes, needs legacy support 1119987478 M * _are_ erm, i though just about everything needs legacy support and it is a real bad idea not to compile in legacy support? 1119987501 M * sannes I have legacy support, I'm quite sure, but I got realloc(): Out of memory when doing so 1119987531 M * _are_ with alpha-utils? 1119987542 M * sannes 0.30.207 1119987544 M * _are_ I had that once when I used the 'stable' ones 1119987587 M * _are_ same here 1119987611 M * sannes ok, thing is last time I got it and did it twice and the machine hung, so I'm not so keen on doing it again .. hehe 1119987624 M * _are_ :-> 1119987681 M * sannes I am going to try when I know that I can have physical access to it 1119988724 J * eyck eyck@81.219.64.71 1119988870 J * erwan_taf ~erwan@AToulouse-105-2-2-69.w217-128.abo.wanadoo.fr 1119989090 J * eXplasm explasm@p549F7C0C.dip.t-dialin.net 1119989420 J * rs ~rs@imhotep.rhapsodyk.net 1119989428 J * hiaslboy ~matthias@249.Red-83-52-226.pooles.rima-tde.net 1119989445 Q * erwan_taf Ping timeout: 480 seconds 1119989447 M * hiaslboy Good evening everybody :-) 1119989529 M * id good evening hiaslboy 1119989738 M * hiaslboy don't we have a topic in this channel ? 1119989761 M * Doener` hm, we used to have one... 1119989774 M * id gone since the last server split 1119989816 T * id : http://linux-vserver.org/ | latest stable 1.2.10, devel 1.9.5, 2.0-rc4, ng9.5 -- He who asks a question is a fool for a minute; he who doesn't ask is a fool for a lifetime -- share the gained knowledge on the wiki, and we'll forget about the minute ;) 1119989821 M * id ;) 1119989889 M * Doener` thx 1119989920 M * Doener` ha! finally i got my simple pty something running :) 1119990210 J * flock ~restless@l192-117-111-12.broadband.actcom.net.il 1119990546 Q * alexx Remote host closed the connection 1119992289 Q * Doener` Ping timeout: 480 seconds 1119992530 M * FaUl bist du sicher das das ne mpq war? 1119992532 M * FaUl oh 1119992536 M * FaUl sry 1119992600 Q * flock Ping timeout: 480 seconds 1119993407 J * matti matti@linux.gentoo.pl 1119993831 Q * hiaslboy Ping timeout: 480 seconds 1119994645 Q * case Ping timeout: 480 seconds 1119994725 M * kevinp anyone run a chkrootkit inside of a vserver? Does it report back false problems because of the different contexts? 1119994777 M * _are_ shall i check? I have a freshly installed server 1119994786 M * kevinp that would be great 1119994805 M * kevinp I just have one problem but on an older server 1119994830 M * _are_ chkrootkit 0.45 ok? 1119994870 M * kevinp yeah, I ran .44 1119995186 M * _are_ Checking `sniffer'... eth0: PF_PACKET((null)) 1119995200 M * _are_ Checking `chkutmp'... => possibly 1 deletion(s) detected in /var/run/utmp ! 1119995200 M * _are_ The tty of the following user process(es) were not found 1119995200 M * _are_ in /var/run/utmp ! 1119995206 M * _are_ (then a few lines) 1119995238 M * _are_ I am pretty sure the system is clean ;) 1119995261 M * kevinp okay, thanks 1119995619 M * kevinp _are_ what version are you running? 1119995626 M * kevinp kernel, tools... 1119995735 M * _are_ 2.6.12.1 with Doeers` 12.1 patch, 2.0 rc something, tools: 0.30.207-8 (debian package) 1119995752 M * kevinp *very* new 1119995782 M * _are_ as said, freshly installed, aka today, 30m ago, kernel compiled 5h ago 1119995801 M * kevinp you don't have and old 2.4 box around that you could check do you? 1119995807 M * _are_ erm, no 1119995842 M * kevinp ok, thanks 1119995843 M * _are_ only 2.6 in variuous hardware flavours from single P4 - 2*Xeon to dual opteron 1119996442 J * Aiken ~james@tooax6-041.dialup.optusnet.com.au 1119996856 M * jonsmel Ok, does anyone know if vservers workd with enbd? 1119996863 M * jonsmel workd = works 1119996901 Q * eXplasm Ping timeout: 480 seconds 1119998320 M * id _are_ still alive ? 1119999095 N * Bertl_oO Bertl 1119999106 M * Bertl evening folks! 1119999122 M * jonsmel hey Bert 1119999141 M * Bertl hey jonsmel, preparing your email now ... 1119999160 M * jonsmel cool, np, do you know if vservers works with enbd 1119999205 M * Bertl what's that? 1119999230 M * jonsmel another network file system 1119999234 M * jonsmel similar to lustre 1119999248 M * Bertl live and learn ... any details? 1119999257 M * jonsmel it's the only other one that is remotely close to what we want to use 1119999337 M * jonsmel you can check out www.it.uc3m.ex/~ptb/nbd 1119999345 M * jonsmel enhanced network block device 1119999352 M * jonsmel is what it stands for 1119999357 M * Bertl ah, yeah, heard of that ... 1119999374 M * jonsmel you know if vservers will work with that 1119999381 M * Bertl but doesn't that need a filesystem ontop? similar to iSCSI? 1119999393 M * jonsmel you can use loopback files 1119999410 M * jonsmel other options besides iscsi 1119999431 M * Bertl hmm ... my nameserver cannot resolve that url :/ 1119999460 M * Bertl but I have to admit, that I never heard of .ex ... 1119999494 M * jonsmel just do google: enbd 1119999502 M * jonsmel that will bring up the site 1119999509 M * Bertl yess, sir! :) 1119999518 M * jonsmel ;) 1119999545 M * Bertl I see, 2.4, 2.4, 2.4 ... 1119999582 M * Bertl but it suuports 2.4.31 :) 1119999586 M * jonsmel that is the version of enbd, i got a little confused on that as well. They say that they have it working with 2.6.10 or 2.6.11 1119999608 M * Bertl http://freshmeat.net/projects/enbd/ 1119999635 M * jonsmel if you google: enbd 2.6 1119999707 M * Bertl http://lists.community.tummy.com/pipermail/enbd/2005/004018.html 1119999772 M * jonsmel right 1119999778 M * Bertl yes, but as I expected, it does only bring a remote disk to you (local) 1119999794 M * Bertl so how would that help you with the dfs issues? 1119999818 M * Bertl (it looks similar to drbd to me) 1119999850 M * jonsmel well, we have a backend data server that pushes the device to the front ends where the vservers will be run 1119999875 M * jonsmel so the frontends will be getting the remote device 1119999898 M * Bertl yes, but you cannot start/mount a non dfs more than once (securely)? 1119999920 M * Bertl let's reduce this setup to something simpler .-.. 1119999936 M * Bertl you have two nodes, both connect via fiber to some SCSI storage 1119999954 M * jonsmel yep 1119999960 M * Bertl you can get access to the raid in the storage from both 'frontends' 1119999973 M * Bertl now how do you proceed with your setup 1120000082 M * jonsmel basically what happens is that the /vservers dir on the backend is pushed to all the frontends so that at any time I can start any particular vserver up on any of the fronends 1120000152 M * Bertl with what fs? 1120000161 M * jonsmel say for example, if frontend 1 has some issue and brings a vserver down, I can, in a matter of seconds bring up those vservers on frontend 2 and cause little down time 1120000165 M * jonsmel doesn't matter 1120000176 M * jonsmel what ever will acomplish the job 1120000183 M * jonsmel that is what we are trying to find 1120000192 M * Bertl well, it does matter, because it has to be aware that there are other mounts ... 1120000219 M * Bertl so you are back to square one with (e)nbd ... in the search for a gfs 1120000223 M * jonsmel we wanted to try and stick with ext3 that is why we were looking into lustre 1120000239 M * Bertl yes, but ext3 will not work here, for example 1120000314 J * rt ~rt@p54B23D15.dip.t-dialin.net 1120000331 M * Bertl jonsmel: OTOH, nfs will work (as it does for lycos) in your scenario 1120000336 M * Bertl evening rt! 1120000350 Q * rt_ Read error: No route to host 1120000482 M * jonsmel the issue we had with nfs is: it is slower, it has database writing issues 1120000547 M * jonsmel the best situation that would fit our scenario was lustre so I am trying to find something similar that does not have all the integration issues 1120000570 M * Bertl did you have a look at coda? 1120000584 M * jonsmel yes, our concern there was the local disk cacheing 1120000622 M * jonsmel that say if a system went down, potentially we could lose a lot of data depending on when the system was last sync'd 1120000640 M * Bertl that's right ... 1120000654 M * jonsmel that turned us off from coda 1120000660 M * Bertl but the same for sure happens with drbd and enbd too, no? 1120000672 M * jonsmel then there is intermezzo which come to find out was the predecessor to lustre 1120000682 M * Bertl if you use any of them in sync mode, they will be dog-slow 1120000702 M * jonsmel enbd is not supposed to do that, it is similar to a nfs mount 1120000726 M * jonsmel right with the slowness, that again is why we liked lustre 1120000732 M * jonsmel unfortunately 1120000748 M * Bertl yeah, well, back to my email :) 1120000792 M * jonsmel anyway, that is basically our only other option is enbd, so I will test to see if that will work with vservers and let you know how that goes 1120000809 M * jonsmel I don't think anyone has done it yet. 1120000820 M * jonsmel or at least no one has responded. 1120000823 M * jonsmel :) 1120000843 M * Bertl I'm pretty sure it will not work out of the box, but it still leaves you with the choice of the fs, which again will have dependancies 1120000866 M * jonsmel right 1120000881 M * jonsmel but maybe it might be easier to fix that lustre 1120000887 M * jonsmel that = than 1120000905 M * Bertl what fs options do you see? the only I see is gfs ... 1120000983 M * jonsmel well, why won't ext3 work, maybe I'm missing something.. 1120001000 M * Bertl simple, try this on one of your nodes: 1120001025 M * Bertl mount -t ext3 /dev/sda1 /mnt/disk1 1120001028 M * Bertl mount -t ext3 /dev/sda1 /mnt/disk2 1120001039 M * Bertl now access the _same_ fs via both pathes and watch 1120001044 J * are|lunch foobar@dsl-084-056-128-130.arcor-ip.net 1120001063 M * jonsmel well, we aren't going to be mounting like that though 1120001091 M * Bertl you will be, because you mount _one_ network block device on more than one nodes, no? 1120001095 M * jonsmel it's more of mounting a directory that is the entire block device 1120001122 M * jonsmel right but it is only mounted once on any node 1120001134 M * Bertl ah, then why not iSCSI? 1120001136 M * jonsmel ie.. frontend 1 has it mounted once 1120001164 M * Bertl would IMHO be the natural choice right after NFS ... 1120001175 M * jonsmel well, we were going to go that route, I will have to check with the other person working on this project, he did not what to go that way 1120001190 M * jonsmel don't know why for sure but I will find out when he is in tomorrow 1120001446 Q * _are_ Ping timeout: 480 seconds 1120001497 J * rt_ ~rt@195.246.161.1 1120001907 Q * rt Ping timeout: 480 seconds 1120002972 M * Bertl jonsmel: well, JFYI, iSCSI would be something which could be arranged ... (at slow pace) because IMHO it might be something of interest to other folks too ... (NAS and SANs are the future :) 1120003158 M * jonsmel ok 1120003191 M * jonsmel I'll see what the details are with that tomorrow. I am outta here for tonight, thanks again bert, laterzzz