[00:00] infowolfe, out o curiosity, what are you getting? [00:06] click (click@gonnamakeyou.com) joined #vserver. [00:07] /usr/lib/util-vserver/vserverkillall: fork: Resource temporarily unavailable [00:07] infowolfe: weeks I know that error! [00:07] eeks I mean [00:07] nathan, could be the nproc thing popping it's ugly again [00:07] infowolfe: I got the same with 1.3.3 [00:07] must be new bug then [00:07] meebey, it was supposed to be fixed in 1.3.4, i guess it wasn't [00:08] nathan, would you let Bertl know that it's happening again [00:08] Herbert thought it was just my normal kernel behavoiur :) [00:08] nathan_, S_FLAGS="lock nproc sched" on that vserver [00:08] infowolfe: lock I used too [00:08] i gotta run, if you're US and need to talk to me, 410-961-8311 [00:08] nproc not, sched not [00:09] infowolfe: I can also tell Herbert [00:09] meebey, i appreciate your help, but nathan knows more about the nuts and bolts of the vserver patches [00:09] :) [00:10] nathan_, even with S_FLAGS unset, it displays same behavior [00:10] and /bin/bash: fork: Resource temporarily unavailable [00:10] so it's a problem with things forking i guess [00:11] i gotta run, nathan, i assume you're us? if you are, call me, let it ring twice and i'll call you back if you need more info on that vserver [00:11] hhmm Herbert let me put some debug printk() into the fork code, but I didnt get any log entries... [00:15] dont under estimate meebay info :) [00:15] sorry i was talking to someone [00:15] ill take a look when i get back i gotta run out [00:18] JonB (~jon@129.142.112.33) joined #vserver. [00:34] tanjix (ViRu_@pD9049C23.dip.t-dialin.net) left irc: [01:14] WSU (~Josh@ny.webpipe.net) joined #vserver. [01:14] Hi [01:15] I am just looking for the VSkel package that Simon Garner released [01:15] anyone have it? [01:19] anyone around? [01:20] i'm here [01:21] You wouldn't happen to have a copy of the vskel script would you? [01:22] nope [01:22] I have a vskel-0.20.tar.gz [01:23] http://swordfish.drown.org/vskel-0.20.tar.gz [01:23] Great! [01:23] thanks [01:23] mind if I add your link to the wiki? [01:23] yeah, make a note that it's a mirror [01:23] or a copy [01:23] -k- [01:32] monako (~monako@ts1-a95.Perm.dial.rol.ru) joined #vserver. [01:33] Hello monako [01:34] hi [02:12] JonB (~jon@129.142.112.33) left irc: Quit: Client exiting [02:12] Nick change: Bertl_oO -> Bertl [02:13] hi WSU! [02:13] hi monako! [02:16] hi Bertl [02:16] Just recompiling kernel to 2.4.24 [02:17] damn security holes [02:17] hey that can be done by recompiling? I always had to patch ;) [02:17] lol [02:17] patch and recompile [02:20] hi Bertl! [02:21] everything fine here? [02:21] no [02:21] everything sucks [02:21] lol [02:22] hmm why do you think so? tell me about your problems! /eliza off [02:31] /usr/lib/util-vserver/vserverkillall: fork: Resource temporarily unavailable [02:31] Action: infowolfe clears throat [02:31] *ahem* [02:32] okay, what versions, what compiler? [02:34] q [02:34] hey bert [02:35] hi nathan_! [02:35] wolfe is here so i guess i dont need to relay his problem [02:39] hmm, not sure he hasn't already shot himself ... [02:41] /usr/lib/util-vserver/vserverkillall: fork: Resource temporarily unavailable [02:41] nathan, could be the nproc thing popping it's ugly again [02:41] nathan, would you let Bertl know that it's happening again [02:41] nathan_, S_FLAGS="lock nproc sched" on that vserver [02:41] nathan_, even with S_FLAGS unset, it displays same behavior [02:41] and /bin/bash: fork: Resource temporarily unavailable [02:41] not sure if you had that in your buffer :) [02:42] hmm, yeah I had ;) [02:43] Action: infowolfe says *BANG* oops, i'm dead [02:43] Bertl, it's 2.4.24-ck1-vs1.3.4... [02:43] gcc 3.2.3 afaik, lemme check [02:44] okay, I assume there is something bad happening to the process accounting ... [02:44] are you interested in locating/fixing that one? [02:44] Action: Bertl hehe good question! [02:48] Bertl, are you talking to me? [02:48] Action: infowolfe isn't a programmer [02:48] Action: infowolfe is merely a guinea pig [02:48] yeah sure I'm talking to you! [02:49] locating/fixing huh? [02:49] i think that kinda thing is outta my league :( [02:49] well you could run some tests for me ... [02:49] i can do that anyway [02:49] and you could describe your setup [02:49] or you could just tell me some jokes, while I'm fixing that issue ;) [02:50] when I know what 'really' happens, then I might be able to reproduce it here, and if this is possible, it's only a small step to fix it ... [02:51] So this pirate walks into a bar... [02:51] :) [02:51] lol [02:51] Action: infowolfe says ARRRRR [02:51] most times fixing something is the easiest part, finding where it is bronken, is hard ... [02:51] Bertl, what would you like to know? [02:51] everything ;) [02:51] gcc (GCC) 3.2.3 20030422 (Gentoo Linux 1.4 3.2.3-r3, propolice) [02:52] you already have the kernel version [02:52] yeah, what did you do ... until the issue popped up out of nowhere? [02:52] http://thor.hardcore-linux.net/weed.conf [02:52] that's the vserver configuration [02:52] um.... [02:52] hrm.. [02:52] i ran vs1.22 and it worked? [02:52] lol [02:53] and when i switched to 2.4.24-ck1-vs1.3.4 it broke when i attempted to do a vserver start [02:53] hmm, okay, so let me note, your fault probably was: running a working 1.22 before ... [02:53] lol [02:54] so I guess I won't get more info out of you, besides 'bertl, it fails!' right? [02:55] i don't know what information you'd like [02:55] here [02:55] hold on [02:55] http://thor.hardcore-linux.net/weed.txt [02:55] enjoy [02:55] that's all i have [02:56] tell me where to look and i might find some more info for you... but i really don't know what you want [02:56] and no... "everything" isn't descriptive enough [02:56] well no problem, I'm good at guessing ... [02:56] lets play some Q&A okay? [02:56] ok [02:57] how many vservers are on that system? [02:57] a few [02:57] how many vserver are started at the same time? [02:57] 5 [02:58] do you enter the vserver with vserver enter? [02:58] Tamama (Pluh@a62-216-20-152.adsl.cistron.nl) joined #vserver. [02:58] oy [02:58] hoi Tamama! [02:58] can i enter it? [02:58] yah [02:58] gutenmorgen :D [02:58] bash: fork: Resource temporarily unavailable [02:58] [root@vserver:weed /] [02:59] do you usually 'enter' them or o you logon via ssh, for example? [02:59] s/o/do/ [02:59] i usually start them via init script and then leave them the hell alone :-p [02:59] lol [02:59] ssh usually [03:00] Action: infowolfe thinks it could be his own ulimit stuff, brb [03:01] heh [03:02] i got my suexec/php to work in a vserver :D [03:02] congrats,! [03:02] by using binfmts in the kernel to a vserver path ;) [03:02] hmm, isn't that some overkill? [03:02] works like a charm [03:03] Bertl, i'm getting the fork: Resource temporarily unavailable all over the place now [03:03] why? i can now start .php files from the command line, like a normal executable.. and it does not need the #!/usr/.... stuff [03:03] with ULIMIT= [03:03] with ULIMIT set from -u 200 to -u 1000 [03:03] okay, infowolfe can we now continue the Q&A? [03:05] sure [03:05] okay, when you start the vserver, how many processes are started (estimation) [03:06] 5-10 daemons [03:06] how long does it take (approximately) until this shows up? [03:06] from what point? [03:07] from the point where a vserver is started [03:07] depends on the vserver, sometimes immediately sometimes not [03:07] but i can't get the one to *STOP* ... [03:07] so i can start it and figure out how long it'll take [03:07] well, you can actually with 1.3.4 [03:08] how? [03:08] there is a new syscall to wipe out an entire context ... [03:08] and IIRC enricos tools support this sycall ... [03:08] so you can send a TERM followed by a KILL to all processes,w hich will stiop the context ... [03:09] using vkill? [03:09] or from inside the verserver [03:09] yes, using vkill [03:09] you can use my tool too ... it does the same ... [03:10] what's your tool? [03:10] the vkill utility ... I thought you where refering to that ... [03:10] http://www.13thfloor.at/vserver/d_release/v1.3.0/vkill-0.01.tar.bz2 [03:10] lol... [03:10] umm, there's another vkill i think [03:11] yeah, probably is .. jacks tools also have one ;) [03:11] but that one requires to 'enter' the context ;) [03:12] ah... [03:12] so, assuming context id is 10... (since i can't comprehend your -h right now) [03:12] what do i type :-p [03:13] hmm could you provide my -h, I can't remember ;) [03:13] vkill: invalid option -- - [03:13] This is vkill V0.01 [03:13] options are: [03:13] -h print this help message [03:13] -k kill (signal num) [03:13] -p target pid (def. 0) [03:13] -x context id [03:13] -V verify interface version [03:13] -- end of options [03:13] okay vkill -x 10 -k 15 [03:13] alright [03:14] should send a term to all processes in ctx 10 [03:14] it says no such process :-p [03:14] ok [03:14] try vkill -x 10 -k 15 -p 0 [03:14] on stop.. [03:15] /etc/rc6.d/S01reboot: fork: Resource temporarily unavailable is after "send all processes TERM signal" [03:15] RoMo (rmoriz@rmoriz.cpan.de) left irc: Read error: Connection reset by peer [03:16] it starts just fine... [03:17] hey, i'm running to the store really quick [03:17] brb in 10 minutes [03:17] hmm, nathan? [03:23] ok [03:23] i'm home [03:24] Tamama (Pluh@a62-216-20-152.adsl.cistron.nl) left irc: Ping timeout: 492 seconds [03:24] hmm, I'd call that quick! [03:24] it's COLD outside [03:24] hmm, how cold? [03:24] umm [03:24] 1 sec [03:25] Action: infowolfe taps out www.weather.com [03:25] wrong unit, it should be degree or such ... [03:25] 30F feels like 18F [03:25] which in centegrade... is umm, i dunno :-p less than 0C :-p [03:25] hmm, cold? it is -10°C here ... [03:26] 30 is -1C [03:26] C = (F - 32) / 1,8 [03:26] but it *feels like* -7C [03:26] F = (C x 1,8) + 32 [03:26] or something like that [03:27] okay, would you be able to run some tests with a patched kernel? [03:29] lol [03:29] sure [03:29] but my kernel is currently patched :-p [03:29] lo [03:29] lol [03:29] Action: infowolfe is being difficult [03:30] yeah, but I'll provide some additional fun patches, okay? [03:31] like? [03:31] like the one, I'm currently hacking together, just for you ;) [03:31] Action: infowolfe doesn't like the sound of that [03:34] http://vserver.13thfloor.at/Experimental/iw01.diff [03:36] what does this one do? :) [03:36] add some proc info ... [03:46] nathan_: I would like to do a backport of the 1.3.3 stuff for 1.23 to remove the locking issues, would you be able and willing to test this? [03:47] Bertl, is that one for me? [03:47] yeah sure iw01 = InfoWolfe 01 =) [03:47] ah [03:47] thanks [03:50] compiling [03:51] Tamama (Pluh@a62-216-20-152.adsl.cistron.nl) joined #vserver. [03:51] weirdness [03:51] yeah, I agree, I thought you where already here?! [03:51] somehow the network interface stopped responding on this computer [03:52] had to reset my router as well [03:52] a vserver issue? [03:52] no this is a w2000 machine [03:52] ah a Micro$oft issue ... [03:54] something like that [03:55] well seems that i got most of this to work so far... [03:55] does anyone currently have redhat kernel + vserver packages? [03:55] you mean, precompiled or patches? [03:56] well, either patches against recent redhat kernels. or packages [03:56] precompiled [03:56] hmm, what is a recent RH kernel? [03:57] and what redhat distribution? [03:57] redhat 9 [03:58] afk, brb [03:58] hmm, no kernel version? [03:59] hm so far i got most of the stuff i need to work in vservers [03:59] sounds good, what doesn't work yet? [03:59] no idea, im still installing :) [04:00] ahh, godd keep installing ... [04:00] but mysql/apache/bind/php/etc all work fine [04:01] i dont even want nasty stuff like hardlinked socket inodes [04:01] heh [04:02] hm [04:02] Bertl, rebooting with the new patch [04:02] 246 MB used memory [04:03] but 155MB is in buffers [04:03] whatever that means heh [04:05] that means, that linux was smart enough, not to throw away, in memory copies of files you accessed some time ago ... [04:06] so it started using my swap instead :D [04:06] Bertl, didn't work [04:06] infowolfe: how could you tell? [04:06] /etc/rc6.d/S01reboot: fork: Resource temporarily unavailable\ [04:06] bash doesn't give the fork error [04:06] and things *look* a bit cleaner [04:07] I was hoping that you get that one, this patch didn't fix anything, it just added some instrumentation to get a closer look at the issue ;) [04:07] lol [04:07] lol [04:07] what a butthead :-p [04:07] but a small placebo effect can be observed, nice ;) [04:08] okay, vserver is running? [04:08] xid = ? [04:10] gah.. im getting those stupid errors again ... heh [04:11] 10 again [04:11] okay, what does 'cat /proc/virtual/10/limit' output? [04:12] VM: 0/-1 [04:12] VML: 0/-1 [04:12] RSS: 0/-1 [04:13] hmm, you might have forgotten to apply the patch at all, or compiled the wrong kernel, or booted the old one ... [04:14] i forgot to apply the patch :( [04:15] *oops* [04:16] rebuilding now [04:16] how do you 'rebuild' [04:16] make dep clean modules modules_install install [04:16] hmm, you could save some time by just doing [04:17] make dep bzImage modules install modules_install [04:20] save time? [04:20] oh [04:20] yah... [04:20] but i don't like bzImage... :-p [04:20] lol [04:21] i am sure bzimage doesnt like you either :D [04:24] nathan_: still around? [04:25] hm so far vservers have been a great help to me... every time i really screw up with an install, i just delete the vserver and remake it :D [04:26] I'm glad you like it ;) [04:27] PROC: 1/500 [04:27] VM: 0/-1 [04:27] VML: 0/-1 [04:27] RSS: 0/-1 [04:27] hmm, we are getting somewhere ... [04:28] and cat /proc/virtual/10/status ? [04:30] ensc: enrico, are you there? [04:31] yep [04:31] regarding the initpid issue on 1.22 [04:32] RefC: 3 [04:32] Flags: 00000007 [04:32] Ticks: 0 [04:32] nr_threads: 3 [04:32] nr_running: 0 [04:32] I'm looking at the code, and it seems like it requires a second syscall with xid=-2 to set this .... [04:32] max_threads: 0 [04:32] total_forks: 49\ [04:32] infowolfe: okay, and this context gives you the for() message, when you do what exactly? [04:33] s/for/fork/ [04:33] attempt to stop it [04:34] that is the only way to cause this? [04:34] afais, chcontext is doing this second syscall... [04:34] could you verify? [04:34] but without the flags... [04:35] hmm, wrong, the VX_INFO_INIT is required ... [04:36] well, I don't think that the way it works in 1.22 is smart anyway, and therefor I removed this specialities in 1.3.x [04:37] Jacques used this second call to set the hostname (first call with full capabilities, second one reomves them) [04:37] those quotas ar weird [04:37] ensc: yeah, actually the -2 call does the following: [04:37] - vx_set_initpid(vc_data.flags); [04:37] if ret=0 [04:38] current->cap_bset &= (~vc_data.remove_cap); [04:38] if (vc_data.flags & VX_INFO_INIT) [04:38] current->s_info->initpid = current->tgid; [04:38] current->s_info->flags |= vc_data.flags; [04:38] so it is quite weird ... [04:39] would it hurt when second call has all flags, or should I set INIT only? [04:40] doesn't hurt, if the flags are appropriate ... [04:41] hm [04:41] ensc: I'm not even sure that this is required at all, let me dig a little deeper ... [04:45] nathan_, hey im here [04:45] im heading over to my girlfriend sbut ill be back on when i get there [04:46] Bertl, yea i can test the 1.23 [04:46] okay .. cu l8er [04:46] im heading over there now, ill be on in like 20m. [04:46] see you soon [04:46] nathan_ (~nathan@209-6-130-26.c3-0.sbo-ubr1.sbo-ubr.ma.cable.rcn.com) left irc: Quit: [BX] Reserve your copy of BitchX-1.0c20cvs for Windows CE today! [04:48] ensc: I guess jack never tried (or even considered) this code path, or the possibility of having a static context and a fakeinit ... [04:48] Bertl, let me know what to try next [04:49] infowolfe: it would be interesting to do the shutdown process line by line if that is possible ... [04:49] Tamama (Pluh@a62-216-20-152.adsl.cistron.nl) left irc: Ping timeout: 499 seconds [04:49] or do you know any other way to 'get' those fork() messages? [04:50] Bertl, i'll investigate the script that seems to be causing the problems [04:51] try to narrow it down with some printf (yeah bash can do that) or sleeps ... [04:51] i know it can [04:53] yummy [04:54] infowolfe: add 'set -x' at the top of the script [04:54] Bertl, if i vserver weed stop and it fails out due to the fork() problems... and then i attempt to enter and edit /etc/rc.d/rc it freaks out [04:55] okay, please explain 'freaks out'? [04:55] and *everything* gives a bash fork error [04:55] perfect ... [04:55] show me the proc values now [04:55] Tamama (Pluh@a62-216-20-152.adsl.cistron.nl) joined #vserver. [04:55] hm.. [04:56] definately a router prob [04:56] + egrep -q '(daemon |action |success |failure )' /etc/rc6.d/S00killall [04:56] Bertl, with it alive or dead [04:56] /etc/rc.d/rc: fork: Resource temporarily unavailable [04:56] what does chcontext --ctx 10 true [04:56] return? [04:57] new security context is 10 [04:57] from inside or outside the vserver [04:57] outside ... inside would probably be very complicated to do ... [04:57] lol [04:57] [root$thor]::root> chcontext --ctx 10 true [04:57] New security context is 10 [04:57] okay, chcontext --ctx 10 bash [04:57] then? [04:58] ooh, thor :) [04:58] that's it [04:58] RefC: 2 [04:58] Flags: 00000007 [04:58] Ticks: 0 [04:58] nr_threads: 2 [04:58] nr_running: 0 [04:58] max_threads: 0 [04:58] total_forks: 143 [04:58] that's the status [04:58] okay, and the /proc/virtual/10/limit ? [04:59] PROC: -4/500 [04:59] VM: 0/-1 [04:59] VML: 0/-1 [04:59] RSS: 0/-1 [04:59] aha, there we go ... [04:59] accounting issue on the NPROCs ... [04:59] now we have to find the cause for PROC being below 0 [05:00] uninitialised variable? *whistles* [05:00] I'll give you some commands to execute, and after each one, you do the following and report the result: [05:00] ok [05:00]