[00:00] and which version of utils I must use? [00:03] hmm, well I adapted it a little, so it should work with ancient tools, but newer tools are suggested ... [00:03] anyone here using gentoo? [00:03] what's the minimum size of a gentoo installation? anyone knows? [00:04] i was using gentoo... [00:04] I was blaming windows :) but Gentoo minimum to me is 1.2GB!!! [00:04] mhepp: was? [00:04] it was about 150MB [00:04] huh? [00:04] how? [00:04] ... [00:04] sigh [00:04] .. [00:05] I just do "emerge system" and voila, 1.2 GB on my HD is occupied :) [00:05] LOL [00:05] just a simple move and my HD is filled [00:05] you don't use Gentoo anymore? [00:06] yes, I am using debian now. [00:06] can you please tell me the basics on how your system was so small? [00:06] Topic changed on #vserver by Bertl!~herbert@MAIL.13thfloor.at: http://linux-vserver.org/ || latest stable 1.24, devel 1.3.6, exp 0.06 [00:06] and why you dropped gentoo for debian? [00:07] it is more simple to install anything on debian. [00:07] ...and faster ;) [00:07] you mean base debian system and then apt-get apps? [00:07] yes [00:08] you can have that with mandrake, RH and suse too, IIRC [00:08] I really like gentoo that has a CD from which you can boot and install things, then take it out and reboot remotely. I love that since my PCs are at a DC [00:08] but RPM-hell never more [00:09] I mean, easily build a system without access to console [00:09] miller7: you can use danix (it is knoppix based distro) [00:09] and you can not install anything [00:10] just move DC to your HDD [00:10] ... CD [00:10] I don't follow you, sorry [00:10] ...ok [00:11] ceskylinux.org is down :( [00:12] my gentoo have 2.2GB it is with KDE and mozilla and OpenOffice.org [00:14] only my /usr/portage dir (without distfiles) is 512MB! [00:14] I go down, I have build new kernel. I go to test it ;) [00:14] miller7: yes, it is so big! [00:14] mhepp (~mhepp@r72s22p13.home.nbox.cz) left irc: Quit: Tak ja padaaaaM [00:14] :((( [00:15] my rescue system (mandrake) is about 300MB ... [00:15] fetch (~fetch@p50851822.dip.t-dialin.net) joined #vserver. [00:16] hi fetch! [00:16] moin all [00:16] hi fetch [00:17] in /usr/src/linux-2.4.24 what should I keep and what should I delete? My kernel runs fine now [00:18] depends ... usually I keep the last kernel, if you add something or change it, you do not need to install a new one (modules for example) [00:19] miller7: do 'make clean' [00:19] that's enought [00:19] well, it's not the kernel I run, it's the stupid gentoo kernel that didn't work!!! [00:19] so I could delete it all? [00:19] yup [00:19] jep [00:20] cool [00:20] rm -r /usr/src/linux-2.4.24 [00:20] hmm, well after all it seems to me that gentoo is an interesting distribution ;) [00:20] should do the job [00:20] Bertl: how? [00:20] Bertl: if u have the time [00:21] miller: first, it builds kernels which don't work, then you have the wrong compiler/headers installed ... [00:21] it also uses the 'broken' 3.2.x compiler ... [00:21] the kernel thing is the stupid automated program :( [00:21] the headers is something bad [00:21] which I might did or not - don't know yet [00:22] the fastes way is to copy a knoppix cd, needs about 10 min for a working system [00:23] wrong, the fastest way is to use the knoppix cd ;) [00:23] Bertl: right ;) [00:23] but I guess, I would not want that installed on a server ... [00:23] me too [00:24] on a server i prefer net-bsd [00:25] Action: miller7 has lowered gentoo installation to 780 MB now [00:26] mhepp (~mhepp@r72s22p13.home.nbox.cz) joined #vserver. [00:26] miller7: most likely there is a command to eliminate the 'build' stuff (sources and unpacked dirs, etc) [00:26] and only keep the built binaries, etc ... [00:26] well, I just deleted some binary things etc [00:26] hehe, like apt-get clean [00:27] packages I mean [00:27] . /usr/lib is 172MB [00:27] perl is 30 MB :-))) [00:27] python 2.2 and 2.3 is another 50MB [00:28] still not usable for vserver :) [00:28] hmm, I don't remember installing python on any of my servers ... [00:28] gentoo needs it for portage [00:29] like RH needs it for installation and update :( [00:29] miller7: in real? rh has pyton for this? [00:29] s/has/use [00:30] yep [00:30] all Redhat tools are in Python [00:30] krass [00:30] Bertl: is it OK? [00:30] #vps [00:30] Debug: sleeping function called from invalid context at mm/slab.c:1856 [00:30] in_atomic():1, irqs_disabled():0 [00:30] Call Trace: [00:30] [] __might_sleep+0xab/0xd0 [00:30] [] kmem_cache_alloc+0x6c/0x70 [00:30] [] alloc_uid+0x94/0x180 [00:30] [] vx_migrate_user+0x2c/0x70 [00:30] [] vx_migrate_task+0x6f/0x180 [00:30] ok, reporting regarding enrico's tools: on a new Gentoo the 0.27.91 compiles perfectly [00:30] [] vc_new_s_context+0x10e/0x240 [00:30] [] sys_vserver+0x112/0x140 [00:30] [] syscall_call+0x7/0xb [00:30] PID CONTEXT TTY TIME CMD [00:30] 757 0 MAIN tty1 00:00:00 bash [00:30] 4494 1 ALL_PROCS tty1 00:00:00 vps [00:31] 4495 1 ALL_PROCS tty1 00:00:00 ps [00:31] mids: http://vserver.13thfloor.at/Experimental/delta-2.4.25-pre7-vs1.24-q0.12-t1.diff [00:31] thanks [00:31] I didn't manage to reproduce it here ... yet [00:31] hopefully this will shed some light on it ... [00:33] mhepp: SMP machine? [00:33] no [00:34] interesting .. brb (investigating) [00:34] but Athlon XP [00:34] (it is supporting SMP) [00:34] hmm i386 or X86_64 ? [00:34] ??? [00:34] i386 [00:34] ah athlon, okay ... [00:35] you see one cpu in /proc/puinfo, right? [00:35] you see one cpu in /proc/cpuinfo, right? [00:35] yes! [00:35] i have only one cpu [00:35] :( [00:36] owners of a xeon board/cpu see two of them (HT) [00:37] well, some alcohol, and you'll see two athlons too ;) [00:37] :) yes, xeon is inside two cpus [00:37] (no, there is VERY big cooler ;) ) [00:39] when did you get that message? [00:39] I mean oops? [00:39] Bertl: when do I have to apply that delta? after the original q0? [00:40] every times, when i write vps command. [00:40] mids: yes, that is ontop of the new q0.12 [00:41] ...and chcontex blabla [00:41] hmm, interesting ... I guess I know why, let me see if I can fix it ... [00:41] but new contex is opened [00:42] you have spinlock debugging enabled I guess? [00:42] (chcontex bash start's new shell...) [00:42] ...maybe... wait... [00:43] yes: sleep-inside-spinlock checking (CONFIG_DEBUG_SPINLOCK_SLEEP) [00:44] good, didn't use that yet, but I guess it's a good idea ;) [00:45] OK ;) [00:47] you know how to edit source code? [00:48] ...? yes? [00:48] okay, in kernel/vserver/legacy.c, line 63 [00:48] Action: _shuri back [00:48] there is a spinlock ... paired with a spinunlock ... [00:49] try to move it into the if() block ... like this: [00:49] if (!(ret = vx_migrate_user(p, vxi))) { [00:50] spin_lock(&p->alloc_lock); [00:50] if (old_vxi) { [00:50] old_vxi->virt.nr_threads--; [00:50] if (old_vxi) [00:50] put_vx_info(old_vxi); [00:50] spin_unlock(&p->alloc_lock); [00:50] } [00:50] out: [00:50] put_vx_info(old_vxi); [00:50] return ret; [00:51] wait... [00:52] is there a way I can close my PC but my putty session continue to compile? [00:53] detach it from the ssh? [00:53] if you 'used' screen on the other end, just detach it ... [00:53] no, I'm with SSH [00:53] well, then you ahve a problem ;) [00:53] :) [00:53] Bertl: yes, I found it... [00:53] well, I can always wait 11 more hours to compile glibc [00:54] anyway, I'm off to home guys [00:54] thanks for all your help [00:54] talk to you tomorrow [00:54] cu, have a nice evening ... [00:54] u2 [00:54] Bertl: what is your time? [00:55] miller7 (none@213.239.180.106) left irc: Quit: ring ring it's 7am, move yourself to go again [00:55] current local time is 23:00 CET [00:55] :) it is nice evening ;) [00:56] just move (cut and paste) the locks, nothing else needs to be changed ... [00:56] what? [00:57] spin_lock(&p->alloc_lock); [00:57] if (old_vxi == vxi) [00:57] goto out; [00:57] if (!(ret = vx_migrate_user(p, vxi))) { [00:57] if (old_vxi) { [00:57] ... [00:57] this first spin_lock? [00:57] after out label? [00:58] if (!(ret = vx_migrate_user(p, vxi))) { [00:58] spin_lock(&p->alloc_lock); [00:58] if (old_vxi) { [00:58] old_vxi->virt.nr_threads--; [00:58] atomic_dec(&old_vxi->limit.res[RLIMIT_NPROC]); [00:58] } [00:58] vxi->virt.nr_threads++; [00:58] atomic_inc(&vxi->limit.res[RLIMIT_NPROC]); [00:58] p->vx_info = get_vx_info(vxi); [00:58] p->xid = vxi->vx_id; [00:58] if (old_vxi) [00:58] put_vx_info(old_vxi); [00:58] ok [00:58] spin_unlock(&p->alloc_lock); [00:58] } [00:58] out: [00:59] put_vx_info(old_vxi); [00:59] return ret; [00:59] } [00:59] this spin_...() into the if (!(...) block... [01:01] ok, [01:01] i have it [01:06] this kernel is working well. there is no troubles, except this spinlock! [01:06] Bertl: http://mids.student.utwente.nl/~mids/vserver/quotalog_1.txt [01:08] mhepp: it's a warning only, it means, there might be a deadlock/race in that code if used on SMP or with preemption ... [01:09] I know... but it writes some mess ;) [01:16] ...will be back (rebuilded kernel...) [01:17] mhepp (~mhepp@r72s22p13.home.nbox.cz) left irc: Quit: Tak ja padaaaaM [01:18] mids: great dump, is it complete? I mean a complete run from start to stop? [01:18] yes, quotaoff is now hanging, no more kernel output [01:20] mhepp (~mhepp@r72s22p13.home.nbox.cz) joined #vserver. [01:21] Bertl: no spinlock warning now.. [01:24] there is time to go to bed ;) [01:29] mhepp (~mhepp@r72s22p13.home.nbox.cz) left irc: Remote host closed the connection [01:32] mhepp (~mhepp@r72s22p13.home.nbox.cz) joined #vserver. [01:33] noel- (~noel@pD9FFAAC0.dip.t-dialin.net) joined #vserver. [01:35] mhepp: okay cu tomorrow ... [01:35] thanks for the testing ... [01:36] i will test it tomorrow again, i maybe found a bug. but i am not sure [01:36] mhepp (~mhepp@r72s22p13.home.nbox.cz) left irc: Client Quit [01:40] fetch (~fetch@p50851822.dip.t-dialin.net) left irc: Quit: Client exiting [01:41] noel (~noel@pD9FFA6E4.dip.t-dialin.net) left irc: Ping timeout: 504 seconds [02:01] Bertl: ok, I am going to sleep; let me know if you need more info/tests/etc [02:01] yeah, thank you that report is quite useful ... [02:01] I'm currently analyzing it ... [02:01] have a good sleep ... [02:05] hellekin (~hellekin@menilmontant-2-81-57-191-62.fbx.proxad.net) joined #vserver. [02:05] re [02:06] Bertl : I used a third way to the vserver-on-tty [02:07] on the host, removed (commented) the inittab entry for tty3 [02:07] copied the line in the vserver's inittab [02:09] then cp -ax /dev/tty3 /path/to/vserver/dev/ [02:09] done. [02:10] well, that was what I suggested, only that I use /dev/vc/0 (because I use devfs) [02:10] ah ok, i don't =8) [02:10] and you can name it different inside the vserver, because it only has 'one' console ... [02:11] well, I'm happy with that. Should I post the hint somewhere ? [02:12] sure, maybe someone can use it ... [03:35] okay, enough for me for today, have a good whatever ... [03:36] cu all later ... [03:36] Nick change: Bertl -> Bertl_zZz [04:00] posted in the documentation of the wiki (ttyHowto) [04:49] lilo (levin@lilo.usercloak.oftc.net) left irc: Ping timeout: 492 seconds [05:06] pflanze (~chris@pop-zh-24-1-dialup-202.freesurf.ch) joined #vserver. [05:06] hello [05:15] hi pflanze [05:44] noel- (~noel@pD9FFAAC0.dip.t-dialin.net) left irc: Ping timeout: 504 seconds [06:05] hm. 1.23 and 1.24 are missing md5sums [06:06] Do you hear me, Bertl? (maybe he'll dream about it now) [06:54] grisha (~grisha@ip68-100-225-22.dc.dc.cox.net) joined #vserver. [07:17] noel (~noel@p508599FD.dip.t-dialin.net) joined #vserver. [07:40] grisha (~grisha@ip68-100-225-22.dc.dc.cox.net) left irc: Quit: Leaving [08:25] pflanze (~chris@pop-zh-24-1-dialup-202.freesurf.ch) left irc: Ping timeout: 492 seconds [09:13] serving- (~serving@213.186.190.24) joined #vserver. [09:14] serving (~serving@213.186.190.24) left irc: Read error: Connection reset by peer [09:43] Nick change: Bertl_zZz -> Bertl [09:44] morning everyone! [09:46] AHTOH (~anton@cs3640-m9.ti.ru) joined #vserver. [09:46] hi Anton! [09:47] hi,hey guys, can any one help me with problem: when i start vserver the sshd inside is working but closing connection just after i enetr correct password [09:47] after i restart sshd inside it works fine [09:47] so starting up with the vserver start gives you a 'bad' sshd? [09:48] yep [09:48] but doing vserver enter and /etc/rc.d/initd/sshd restart [09:48] [anton@arbeit anton]$ ssh root@vserv [09:48] root@vserv's password: [09:48] Connection to vserv closed by remote host. [09:48] Connection to vserv closed. [09:48] fixes that? [09:48] yeah, that just what i have [09:48] okay, kernel/patch version? [09:48] 2.4.24 , vserv1.24 [09:49] tools? [09:49] util-vserver 0.27 [09:49] fedora1 outside and inside [09:49] hmm, any messages inside the vserver's /var/log/messages? [09:50] regarding ssh of course [09:50] no [09:51] just some xinetd problems [09:51] also inside i have chkconfiged all services off except xinetd sshd and syslogd [09:52] outside sshd and xinetd replaced with v_sshd and v_xinetd [09:53] sounds a little weird ... I would try to disable the runlevel script inside the vserver and start a sshd -d (debug) from rc.local ... and see if this works/or why it fails ... [09:53] Jan 27 09:54:48 test1 sshd(pam_unix)[1795]: session opened for user root by (uid=0) --- that logs when it closes me [09:54] ok i try now [09:55] lastlog_perform_login: Couldn't stat /var/log/lastlog: No such file or directorylastlog_openseek: /var/log/lastlog is not a file or directory! [09:55] debug1: PAM establishing creds [09:55] debug1: permanently_set_uid: 0/0 [09:56] thats ssh client output [09:58] hmm , i noriced that when i disable service and start sshd manually it works [09:58] okay, try to create the lastlog entry [09:59] (inside the vserver) [09:59] yes, touched it, but sure will not help -- trying [09:59] while can you answer about current restrictions on vserver 1.24 --- what do i have there [10:00] process number, general nice -- i see [10:00] but what about memory and quota? [10:00] depends ... in the stable version, you have no memory limits [10:00] but you can use fair scheduling and a process limit [10:01] (max processes per context) [10:01] quota can be done with 'normal' quota system if you put each vserver on a separate partition (lvm/loop/etc) [10:02] but can i limit each vserver say to 5% of CPU [10:02] there is an addon patch, which allows you to do per context disk limits and quota on a shared partition ... [10:02] is that patch stable? [10:03] the idea of vserver is not to limit resources if they are available, so, you cannot limit the cpu to 5% if nothing else is running, but you can have 20 context, each using the same amount of cpu ... [10:03] patch-2.4.24-vs1.24-q0.12.diff.bz2 this one? [10:03] yep, that one, basically this is about 1 year in production [10:04] but recently some ext3 issues emerged .. probably not patch related ... [10:04] well, I would say, it's at least as stable as the normal quota system ... [10:04] and what about running vserver for 2.6 -- how is it working? [10:05] there are experimental patches, latest is 0.06 .. which do the same on 2.6 [10:05] hmmm sshd still closes my connection if started from rc.sysconfig , but works fine if manually started/restarted [10:06] hmm, the 'debug' version closes the connection on you too? [10:07] without giving any additional explanation, why and what? [10:07] no, if start sshd -d from console manually all works fine [10:08] well, try starting it with -d from the runlevel script ... [10:08] the scripts are handled a little differently, if you do not use fakeinit ... [10:09] if i write sshd -d into rc.local that will do? [10:09] well, it should, just make sure to disable the sshd service ... [10:09] fcourse [10:09] and pipe the output somewhere ... [10:10] yeah i got [10:10] after accepting password i have: [10:10] debug1: monitor_child_preauth: root has been authenticated by privileged processdebug1: Entering interactive session for SSH2. [10:10] debug1: fd 3 setting O_NONBLOCK [10:10] debug1: fd 7 setting O_NONBLOCK [10:10] debug1: server_init_dispatch_20 [10:10] debug1: server_input_channel_open: ctype session rchan 0 win 65536 max 16384 [10:10] debug1: input_session_request [10:10] debug1: channel 0: new [server-session] [10:10] debug1: session_new: init [10:10] debug1: session_new: session 0 [10:10] debug1: session_open: channel 0 [10:10] debug1: session_open: session 0: link with channel 0 [10:10] debug1: server_input_channel_open: confirm session [10:10] debug1: server_input_channel_req: channel 0 request pty-req reply 0 [10:10] debug1: session_by_channel: session 0 channel 0 [10:10] debug1: session_input_channel_req: session 0 req pty-req [10:10] debug1: Allocating pty. [10:11] debug1: session_pty_req: session 0 alloc /dev/pts/1 [10:11] debug1: server_input_channel_req: channel 0 request x11-req reply 0 [10:11] debug1: session_by_channel: session 0 channel 0 [10:11] debug1: session_input_channel_req: session 0 req x11-req [10:11] debug1: server_input_channel_req: channel 0 request shell reply 0 [10:11] debug1: session_by_channel: session 0 channel 0 [10:11] debug1: session_input_channel_req: session 0 req shell [10:11] debug1: PAM setting tty to "/dev/pts/1" [10:11] PAM session setup failed[6]: Permission denied [10:11] debug1: Calling cleanup 0x8059c20(0x8090c20) [10:11] debug1: session_pty_cleanup: session 0 release /dev/pts/1 [10:11] debug1: Calling cleanup 0x8067410(0x0) [10:11] debug1: channel_free: channel 0: server-session, nchannels 1 [10:11] debug1: Calling cleanup 0x806f3a0(0x0) [10:11] debug1: Calling cleanup 0x8063d50(0x0) [10:11] hmm, seems like a pam issue ... [10:12] could you check that /dev/pts/1 is not used in another context ... [10:12] that is only vserver i have [10:12] yeah, but it could be used on the host ;) [10:13] ps ax | grep pts shows only pts/0 [10:13] okay, looks good ... [10:13] and strange -- why after manual restart all is fine [10:13] well, might be a bug in util-vserver ... [10:14] I would suggest to wait a few hours, until Enrico shows up, he might be able to help you debug this issue ... [10:15] (nick: ensc) [10:15] ok thanks for help [10:15] np [10:15] just i ll try to think if i have more questions [10:16] ah, what about memory limit -- i need an additional patch? [10:16] the development release does RSS and VM accounting and allows to limit VM [10:16] that is 1.36 ? [10:17] currently no RSS limit is implemented, because the 2.4 kernel does not allow to limit RSS per se ... [10:17] yes, that is 1.3.6 ... [10:17] there is a memory limits patch for 1.24 somewhere, but I'm not sure it is up to date ... [10:18] is 1.3.6 more unstable than 1.24? [10:18] well, it is development ... so it can break and panic on you, in which case, you should file a bug report ... [10:19] basically development is more feature oriented, stable is more stability oriented ... [10:19] and what about iptables -- only on host it can be run? but can i add there rules for vserver interfaces? [10:20] yes, the current network implementation does not change the way aliases and interfaces work ... [10:20] it only restricts the number of ips you can bind to from within a vserver ... [10:20] in that case an interface as i guess is a eth0:name ? [10:20] with all advantages and disadvantages this approach has. [10:21] and why i say vserver name stop the interface eth0:name is still up? [10:21] yes, correct ... eth0: to be precise ... [10:21] depends how you configure it [10:21] IPROOT="192.168.0.1" for example does not add or remove an alias for you ... [10:22] IPROOT="eth0:192.168.0.1" creates it on start and destroys it on stop ... [10:23] i have IPROOT="eth0:212.x.x.x/255.255.255.192" [10:24] can be the problem my name has a number in the end : test1 [10:24] nope, the number is just added if you have more than one ip, and it has no other meaning than to get a different name ... [10:26] you could do an strace on that sshd, or capture the network traffic, to see if there is any difference between the auto started and the manually started ... [10:27] as an adhoc i can add vserver name service sshd start into vserver script [10:28] you can try, but that should be equivalent to the runlevel script inside ... [10:28] (so basically give the same result/error) [10:29] no that is ok if i disable service sshd, and dor vserver name start and then vserver name sshd start [10:29] hmm, interesting ... [10:30] maybe the sshd depends on something, not present at the startup? [10:30] maybe the sequence of the services is wrong? [10:30] how can i see that sequence? [10:30] /etc/rc4.d/ [10:30] S01usb@ S12syslog@ S56rawdevices@ S80postfix@ S99linuxconf@ [10:30] S03iptables@ S14nfslock@ S56xinetd@ S85gpm@ S99local [10:31] the numbers in front give the sequence ... [10:31] actually those are symlinks to the start/stop scripts ... [10:31] i should look into rc3 ? [10:31] (this is sysv init style ;) [10:31] i thought i have runlevel 3 [10:31] probably, this was just an example ... [10:33] K03messagebus K15gpm K44rawdevices S12syslog S56xinetd [10:33] K05saslauthd K15httpd K92iptables S55sshd S99local [10:33] what is difference between K and S [10:33] means that sshd is started before xinetd [10:33] S = start K= kill [10:34] so maybe changing S55sshd to S57sshd would help ... [10:34] hm is that bad? should i rename to sshd to 57 [10:34] you could try, although I don't know why sshd would depend on xinetd ... [10:34] nothing changed: ( [10:37] try doing an strace -fF -s 10000 of that sshd ... oncein the failing case, another time in the working and put them on the net somewhere, that I can have a look at it ... [10:38] I have to leave now, but I'll be back in about 100 mins ... [10:38] ok, once more great thanks [10:38] np [10:38] cu later [10:39] Nick change: Bertl -> Bertl_oO [10:49] je (~je@hd5e25b7f.gavlegardarna.gavle.to) left irc: Ping timeout: 480 seconds [11:10] serving- (~serving@213.186.190.24) left irc: Read error: Connection reset by peer [11:15] je (~je@hd5e25b7f.gavlegardarna.gavle.to) joined #vserver. [11:27] mhepp (~mhepp@r72s22p13.home.nbox.cz) joined #vserver. [11:32] mhepp (~mhepp@r72s22p13.home.nbox.cz) left irc: Remote host closed the connection [11:33] je (~je@hd5e25b7f.gavlegardarna.gavle.to) left irc: Remote host closed the connection [11:39] je (~je@hd5e25b7f.gavlegardarna.gavle.to) joined #vserver. [11:43] Nick change: Bertl_oO -> Bertl [11:44] hi everyone! [11:44] hi again [11:44] was a little faster than I expected ... [11:44] AHTOH: did you manage to get an strace? [11:51] i ve misunterstood how to do it [11:51] hmm, you know strace? [11:53] negative [11:54] okay, np [11:54] find a rh strace package and install it, if not already done so ... [11:54] i ve read man and i see i have to run it from client ssh? [11:54] strace -fF -s 1000 ssh root@vserv ? [11:54] nope, what you do is: [11:55] strace -fF -s 1000 sshd 2>&1 >/tmp/sshd.strace [11:55] and that inside one of the runlevel scripts ... [11:56] best to add -d to sshd, if this gives produces the failure for you ... [11:56] gives/produces [11:58] that will produce quite some output ... [11:58] so try to do one conenct only, and copy the log somewhere ... [11:59] strace -fF -s 1000 sshd -d 2>&1 >/tmp/sshd.strace ? [11:59] should work ... [12:10] 190K file is a result [12:10] guessed so, can you put it somewhere on the web, maybe compressed? [12:13] http://pueblo.ru/shd.strace.bz2 [12:13] http://pueblo.ru/sshd.strace.bz2 [12:24] hmm, looks okay, until PAM denies the session ... [12:25] setpriority(PRIO_PROCESS, 0, 0) = -1 EACCES (Permission denied) [12:25] that seems to fail ... [12:26] I assume that the setpriority needs more permissions (capabilities) than used for the runlevel scripts (from the vserver tools) [12:30] I checked, it seems CAP_SYS_NICE is required to do that .. are you still there? [12:30] yeah [12:31] but why if i launch that manually all works fine [12:31] okay, then try to add that cap to your server config, jsut for a test ... [12:31] oki [12:31] it might be that the tools use lower caps on the runlevel start ... [12:31] (we have to wait for enrico to know more) [12:32] gut [12:32] it works [12:32] it does? [12:32] yeah [12:32] okay, funny that fedora tries to manipulate the priority settings ... [12:33] but if i dont want to give privileges for nicing [12:33] probably recompiling the pam stuff without that renicing (or sshd, but I suspect pam) will solve that ... [12:34] maybe there is also an option for that sshd/pam not to do/try this ... [12:34] hmmm likely its simplier to have an rh9 inside [12:41] je (~je@hd5e25b7f.gavlegardarna.gavle.to) left irc: Ping timeout: 492 seconds [12:43] great,at least the reason is located [12:44] but explain me what is the difference when i say service sshd restart from rc scripts or when i do it manually [12:46] hmm, maybe you should do some more tests ... [12:47] remove that CAP_NICE again and output /proc/self/status from the runlevel script and from vserver enter, and upload the output ... [12:47] ok [12:55] [root@vserver:test1 /]cat /proc/self/status [12:55] Name: cat [12:55] State: R (running) [12:55] Tgid: 1128 [12:55] Pid: 1128 [12:55] PPid: 1061 [12:55] TracerPid: 0 [12:55] Uid: 0 0 0 0 [12:55] Gid: 0 0 0 0 [12:55] FDSize: 256 [12:55] Groups: 0 1 2 3 4 6 10 [12:55] VmSize: 3556 kB [12:55] VmLck: 0 kB [12:55] VmRSS: 416 kB [12:55] VmData: 148 kB [12:55] VmStk: 4 kB [12:55] VmExe: 16 kB [12:55] VmLib: 1320 kB [12:55] SigPnd: 0000000000000000 [12:56] SigBlk: 0000000000000000 [12:56] SigIgn: 8000000000000000 [12:56] SigCgt: 0000000000000000 [12:56] CapInh: 0000000000000000 [12:56] CapPrm: 00000000d40c04ff [12:56] CapEff: 00000000d40c04ff [12:56] CapBset: 00000000d40c04ff [12:56] s_context: 49153 [ -16383] [12:56] ctxticks: 25 5 5 [12:56] ctxflags: 7 [12:56] initpid: 0 [12:56] ipv4root: 05e001d4/c0ffffff [12:56] ipv4root_bcast: 3fe001d4 [12:56] ipv4root_refcnt: 2 [12:56] that was from console [12:56] --------------------- [12:56] okay we only need the CapPrm,CapEff,CapBset [12:56] [root@vserver:test1 /]cat /tmp/stat [12:56] Name: cat [12:56] State: R (running) [12:56] Tgid: 1565 [12:56] Pid: 1565 [12:56] PPid: 1563 [12:57] TracerPid: 0 [12:57] Uid: 0 0 0 0 [12:57] Gid: 0 0 0 0 [12:57] FDSize: 256 [12:57] Groups: 0 1 2 3 4 6 10 [12:57] VmSize: 3556 kB [12:57] VmLck: 0 kB [12:57] VmRSS: 416 kB [12:57] VmData: 148 kB [12:57] VmStk: 4 kB [12:57] VmExe: 16 kB [12:57] VmLib: 1320 kB [12:57] SigPnd: 0000000000000000 [12:57] SigBlk: 0000000000000000 [12:57] SigIgn: 8000000000000000 [12:57] SigCgt: 0000000000000000 [12:57] CapInh: 0000000000000000 [12:57] CapPrm: 00000000d40c04ff [12:57] CapEff: 00000000d40c04ff [12:57] CapBset: 00000000d40c04ff [12:57] s_context: 49154 [ -16382] [12:57] ctxticks: 6 1 6 [12:57] ctxflags: 7 [12:57] initpid: 0 [12:57] ipv4root: 05e001d4/c0ffffff [12:57] ipv4root_bcast: 3fe001d4 [12:57] ipv4root_refcnt: 7 [12:58] ooops to late i read you :( [12:58] that was from rc.local [12:58] np [12:58] hmm, it is interesting anyway, this is the same server, but different startups? [12:59] yeag [13:00] first i cat it [13:00] okay, try to use a static context id, just for a test, add S_CONTEXT=100 to your vserver config file [13:00] then stopped modified rc.local and started [13:00] ok [13:01] 100 doesnt improves the situation :( [13:02] I was hoping that it doesn't change anything ... ;) [13:02] now you need self-status? [13:02] no, not required ... [13:02] the funny thing is, the permissions are the same (capabilites) [13:03] so either you forgot to remove that CAP_NICE, or the pam/sshd stuff is complaining in one case, and ignoring it in the other ... [13:05] no i havent forgot removing CAP NICE [13:06] serving (~serving@213.186.190.24) joined #vserver. [13:06] well, this means, that both invocations are done with the _same_ capabilites, and in one case, the pam complains, in the other not ... funny stuff ... [13:07] yeah real fun :) but where should i search for it [13:08] you could do the strace again for the working case, and compare it to the non working ... [13:08] ok i ll try [13:17] AHTOH (~anton@cs3640-m9.ti.ru) left irc: Remote host closed the connection [13:18] je (~je@hd5e25b7f.gavlegardarna.gavle.to) joined #vserver. [13:18] hi je?! [13:22] hi Bertl. my isp is doing some network maintenance so my connection is a bit unstable. [13:22] np, we are used to unstable ;) [13:23] ;D [13:34] je (~je@hd5e25b7f.gavlegardarna.gavle.to) left irc: Ping timeout: 480 seconds [13:43] je (~je@hd5e25b7f.gavlegardarna.gavle.to) joined #vserver. [14:30] miller7 (none@213.239.180.106) joined #vserver. [14:30] hello people [15:04] hi miller7! [15:04] hey bert [15:15] how is your gentoo system? [15:16] the one you have access to? I have to re-build it [15:16] some other ones I'm building, they go great [15:16] no problem there, I'm not going to access it anymore ... [15:16] ok, I will rebuilt it then :) [15:16] and get myself a normal /vservers partition [15:16] and a normal gcc [15:17] good idea ... [15:17] and probably use ext2 for /vservers, right? [15:17] for the quota to work inside vs [15:17] well, ext2 seems to be the best solution atm ... [15:17] cool [15:18] did you follow the ext3/quota issue 'mids' reported? [15:18] I read part of it but didn't see the last parts [15:18] what was the result? [15:18] still freezing? [15:18] well, funny thing is, I can't reproduce it here ... [15:18] :-) [15:19] I think it was easily reproducable 4-5 months ago I had tried the old patches before the project takeover [15:20] well, I mkfs a new partition for ext3, touch some files, add the quota hash, turn quota on, let it run a little ... turn quota off, remove the quota hash, no problem there ... [15:20] not even a single message that was reported :( [15:21] haha [15:21] If I have time one of these days I'll try to put vs quota [15:21] and test [15:21] or else after I return to homeland I will [15:22] homeland means greece? [15:22] yep [15:23] how is the ip infrastructure there? [15:23] normal, why? [15:23] probably better than the telephone network, right? [15:23] The telephone network is fine I think :) [15:23] did you have any problems? [15:24] well, I heard that telephone is very expensive and unreliable ... but might be a misinformation ... [15:24] well, not really [15:24] when did you hear that? [15:24] okay, so just another urban legend ;) [15:25] a friend of mine, who was some time (4 month IIRC) there told me ... [15:25] well, I don't know [15:25] what is he doing here? [15:25] is he into telco businesS? [15:25] he was looking after his subsidiary business location ... [15:26] ic [15:26] well, the telco quality I've experienced from companies in UK, Netherlands and Germany do not even match to ours [15:27] I mean, ours is better :) [15:27] well good to hear ... [15:27] but the telco has high prices to call abroad, that's true [15:27] maybe he was only speaking about the price/service ratio ... [15:28] calls to mobiles cost 30-40% less than the calls to mobiles in .de/.nl/.uk :) [15:28] well, yeah perhaps [15:28] for whatever reason, I jsut remembered that part ... ;) so no offense meant [15:28] no, it's ok - [15:44] hello [15:44] hi! [15:44] hi hellekin [15:44] x-term (~x-term@cs3640-m9.ti.ru) joined #vserver. [15:45] Some crazy idea hit my brain... [15:45] 2 Bertl, hi its AHTOH here [15:45] Imagine a network of machines running a similar virtual network of vservers. [15:45] Nick change: x-term -> AHTOH [15:46] is Enrico here? [15:46] There must be a simple way to allow a vserver to be dynamiccaly transfered from one machine to another, in a quite transparent way, with dynamic dns updates and proxying... What do you think? [15:46] I love the idea of "nomad servers" [15:47] hellekin: you just need a NAS and vservers that reside there [15:47] you can do load balancing easily by just stopping the vserver and starting it from another machine [15:47] without data transfer [15:48] but if the NAS goes down... [15:48] assuming that all contain static information ,) [15:48] using different machines we allow multiple points of failure [15:48] is this stupid or worth invetigating ? [15:49] it's not stupid but moving a vserver from one machine to another is trivial [15:49] just stop it, copy data and start it :) [15:49] (unless you have compiled apps for a specific CPU) [15:50] miller7 : yes, vserver-copy -s old new_host:new is ok [15:50] but if you do that, there will be a gap that has to be filled using proxying eventually to maintain the link to service between DNS updates... [15:51] are you talking about moving it to new IP? [15:51] yes [15:51] ah [15:51] ok [15:51] well you just have to install some port forwarding thingy I guess [15:51] but changing IP is never nice [15:53] you can't do much about it, no matter what way you go [15:53] that's my personal opinion [16:25] pflanze (~chris@195.176.0.209) joined #vserver. [16:25] hi pflanze! [16:39] hi bertl [16:39] any advances [16:40] is this a question? [16:40] i have another problem: httpd not loading in vserv :mmap2 /dev/zero fails [16:40] yeah that was question, enrico still offline? [16:40] you mean /dev/zero does not exist? [16:41] no it exists [16:41] so what fails? [16:41] open("/dev/zero", O_RDWR) = 8 [16:41] mmap2(NULL, 24, PROT_READ|PROT_WRITE, MAP_SHARED, 8, 0) = 0x4002a000 [16:41] close(8) = 0 [16:41] gettimeofday({1075210111, 449275}, NULL) = 0 [16:41] write(7, "[Tue Jan 27 16:28:31 2004] [emer"..., 93) = 93 [16:41] close(7) = 0 [16:41] close(4) = 0 [16:41] close(6) = 0 [16:41] close(5) = 0 [16:41] close(3) = 0 [16:41] exit_group(0) = ? [16:41] Process 3297 detached [16:42] mmap2() didn't fail..? [16:42] but why it terminates [16:42] doesn't seem so ... [16:42] what is the complete error message? [16:42] at least 0x4002a000 looks like a good mapping ... [16:44] [Tue Jan 27 16:28:31 2004] [emerg] (38)Function not implemented: Couldn't create accept lock [16:45] hmm, search for something containing lock in the strace, you used -fF didn't you? [16:45] -f only [16:47] -fF -- nothing about lock [16:47] grep -I lock ?! [16:47] might be something different ... is the strace output available somewhere? [16:48] yeah grep empty [17:15] rehi [17:15] hi mids ... [17:15] we have to find out what's the difference between your system and my test setup ... [17:16] I'll boot it up [17:16] pflanze (~chris@195.176.0.209) left irc: Ping timeout: 501 seconds [17:20] Pentium III 550Mhz [17:20] 512MH Ram [17:20] /dev/hdf1 on /disk2 type ext3 (rw,tagctx,usrquota,grpquota) [17:20] euh, what else matters? [17:20] what context tagging setup did you select? [17:21] S_NICE="" [17:21] S_FLAGS="lock nproc" [17:21] ULIMIT="-HS -u 256 -n 1024" [17:21] S_CAPS="CAP_QUOTACTL" [17:21] S_CONTEXT="101" [17:21] rm -f /vservers/test1/dev/hdv1 [17:21] vrsetup /dev/vroot/0 /dev/hdf1 [17:21] cp -fa /dev/vroot/0 /vservers/test1/dev/hdv1 [17:21] cqhadd -v -x 101 /dev/vroot/0 [17:21] chctx -R 101 /vservers/test1 [17:21] BLOCKS=`du -skx /vservers/test1|awk '{print $1}'` [17:21] INODES=`find /vservers/test1|wc -l` [17:21] cqdlim -x 101 -S $INODES,10000,$BLOCKS,524288,10 -v /dev/hdf1 [17:22] hmm, maybe you could put that all somewhere on the web? [17:22] ok [17:22] especially the .config file of the kernel ... [17:22] dd if=/dev/hdf1 of=dump.txt :) [17:22] äh ... [17:23] please use xxd /dev/hdf1 >/tmp/dump.txt 8-) [17:24] http://mids.student.utwente.nl/~mids/vserver/ [17:31] not sure if it matters, but I mount disk2 under /disk2 [17:32] and /vservers is a symlink to /disk2/vservers [17:34] hellekin (~hellekin@menilmontant-2-81-57-191-62.fbx.proxad.net) left irc: Ping timeout: 501 seconds [17:37] mhepp (~mhepp@r72s22p13.home.nbox.cz) joined #vserver. [17:38] mids: okay, I'll adapt my setup a little ... [17:38] Bertl: I found bug! When I open a new contex, its hostname is set in whole system [17:39] hmm, this is vs0.06, right? [17:39] yes [17:39] hmm, yeah, the hostname virtualization isn't done yet ... [17:40] but it's good to know that somebody actually tested it ;) [17:41] :) [17:43] Bertl: ok, I tried this: [17:43] yup? [17:44] mounting disk2 as /vservers: no lockup, but in_use / dirty flag output upon quotaoff [17:44] Bertl: just wondering... why are you spending some much tmie on this? [17:44] tmie == time [17:44] mounting disk2 as /disk2, symlinking to /vservers: no lockup, but in_use / dirty flag output upon quotaoff [17:44] Zoiah: because I'm nuts! and it's fun! [17:44] Bertl: no commercial interests? [17:45] mounting disk2 as /disk2, symlinking /disk2/vservers to /vservers: lockup~, in_use / dirty flag output upon quotaoff and quota in use/dquout line [17:45] Zoiah: nope, I just want to live and of course ... world domination, fast! [17:46] so the lockup upon quotaoff has something to do with the vservers directory beeing in the root of the disk or not... or so it appears [17:46] okay just to make sure I understood that ... [17:47] fun, the BAD_INODE error also only appears in case #3 (symlink to subdir) [17:47] if you reference a subdirectory with the link, you get htat, right? [17:48] yes: /vservers -> /disk2/vservers/ [17:48] hmm ... [17:48] okay, I can try that ... sec [17:49] vat (vat@pD9E374CC.dip0.t-ipconnect.de) joined #vserver. [17:49] hi. [17:49] hi vat! [17:49] hi vat [17:49] ah. hi mids, hi Bertl. [17:49] Bertl, query. [17:50] vlinux08:/var/log# uname -a [17:50] Linux vlinux08 2.4.24ctx-17c #1 SMP Thu Jan 8 20:34:04 CET 2004 i686 unknown [17:50] okay. [17:50] problem: eth0-aliases are often "away" ;). [17:50] any known bugs about that? [17:50] hmm, spontaneously disappearing? [17:50] yep. [17:50] after that, you do a vserver id enter, and they are back. [17:51] hmm, sounds strange ... [17:51] by disappear, you mean the host isn't responding or the interface is really not there? [17:52] I had this a few times when I ran ctx-17.. I've recently upgraded to vs1.24 and haven't noticed it anymore. [17:52] hmm, this is 2.4.24ctx-17c ? [17:52] yep. [17:52] what is that? [17:52] the interface is not there. [17:52] well, normalle we got 2.4.22ctx-17c on the old systems.. [17:52] .oO(its vs1.2 iirc) [17:52] well, I didn't release a 17c for 2.4.24, right? [17:53] yep ;) [17:53] sooo? [17:53] but it's compatible ;P [17:53] it is? [17:53] well, it works... [17:53] just every two weeks a silly bug *g* [17:53] well, seems, except for some disappearing interfaces, right? [17:53] maybe will downgrade to 2.4.22 again. [17:54] what is the problem with vs1.24 ? [17:54] tried it, but did not boot. [17:54] AHTOH (~x-term@cs3640-m9.ti.ru) left irc: Quit: Client exiting [17:54] hrm, you are going to tell me, that vs1.24 doesn't boot, but 2.4.24-ctx17c does? [17:55] I'm almost sure that is an issue between keyboard and chair ;) [17:55] vat: what error you get when booting? [17:55] miller7, no error. just uncompressing kernel and halting. [17:55] Bertl, ehm. [17:56] sounds to me like a CPU error [17:56] vlinux08:/var/log# uname -a [17:56] Linux vlinux08 2.4.24ctx-17c #1 SMP Thu Jan 8 20:34:04 CET 2004 i686 unknown [17:56] vlinux08:/var/log# uptime [17:56] 15:57:19 up 17 days, 20:29, 2 users, load average: 0.29, 0.61, 0.67 [17:56] kernel is compiled for wrong CPU [17:56] ;-) [17:56] miller7, nak. 2.4.24ctx17-c got the same .config. [17:56] I would suggest you look at the .config carefully [17:57] well.. it works with the normal 2.4.24 and with ctx-17c. [17:57] vat: after copying the .config what where your steps to build the kernel? [17:57] and which compiler did you use? [17:57] gcc version 2.95.4 20011002 (Debian prerelease) [17:57] vat: I have installed the 1.24 patched kernel 3-4 times the last days [17:57] that does not matter, since 2.4.24 works with that. [17:58] 2.95.4 is okay ... [17:58] about the steps? [17:58] Bertl, patching, make blah? [17:58] what about oldconfig or menuconfig? [17:58] same patching like any of the other 20 kernels. [17:58] cp .config, make menuconfig, save, blah [17:58] ;) [17:58] okay ... [17:59] what CPU you have in that box? [17:59] sounds strange, vs1.24 doesn't even come near to the boot stage code ... [17:59] model name : AMD Athlon(tm) XP 2400+ [17:59] Bertl, i know. got no idea why it was so, but it was so. [17:59] so with the same config the 2.4.24 vanilla boots but with the 1.24 patch it does not? (same .config both times?) [18:03] yep. [18:03] does not sound good :) [18:03] will try it tomorrow again, will be in the noc tomorrow. [18:04] please do [18:04] it is very strange [18:04] what FS you use? [18:04] ext3. [18:04] .oO(and yes, i did not forget ext3 support in the kernel ;)). [18:05] well, since it works on vanilla and the config is same... :) [18:05] if you get uncompressing kernel, but nothing else this means that either console is not enabled or early cpu detection/apic code is busted ... [18:06] or you compile for wrong CPU as I did earlier this morning by accident :) [18:06] nah. [18:06] as i told, it is 100% the same .config ;). [18:06] vat we believe you [18:06] just think it's strange [18:06] .oO(okay, Makefile diff's with EXTRAVERSION ;-)). [18:12] vat: problem is, if I fix that for you, you end up with a patch like vs1.24, which, in turn, won't work either ;) [18:13] hehe. [18:13] will do it tomorrow again. [18:13] let's see. [18:16] root tty1 Mon00 12:19m 0.07s 0.07s -sh [18:16] hrhr. 2.4.23-xfs rocks also ;) [18:18] is xfs cool? [18:19] xfs is ... interesting ;) [18:20] Action: mids looks for a reiser vs ext3 vs xfs vs jfs comparison sheet [18:21] ..nah.. xfs rocks ;).. no lame filesystem errors like on ext3 ;) [18:21] but only working with that at home on my LFS. [18:25] but recovery on a bad disk is 'hard' [18:25] :) [18:25] hmm, well, at SGI they throw bad disks away 8-) [18:26] hehe [18:37] hm. [18:37] which capatibility is a must, if you want to bind own eth-aliases in the vserver? [18:38] you mean define your own network interfaces ... [18:38] fuck, now quotaoff locks up even with vserver mounted as /vserver [18:38] eyp. [18:38] AAARGH [18:38] eh, yep. [18:38] guess NET_ADMIN ... [18:38] ah, yes. [18:45] click (click@gonnamakeyou.com) left irc: Quit: server maintainance in progress - back in 10 minutes. [18:52] hellekin (~hellekin@menilmontant-2-81-57-191-62.fbx.proxad.net) joined #vserver. [19:19] mhepp (~mhepp@r72s22p13.home.nbox.cz) left irc: Remote host closed the connection [19:23] exit [19:23] hellekin (~hellekin@menilmontant-2-81-57-191-62.fbx.proxad.net) left irc: Quit: BitchX: coming soon to a theatre near you! [19:23] shutdown -h now [19:24] system is going down now ... [19:24] ????? [19:24] who's rebooting my PC? [19:24] :) [19:46] mids: I have an idea, but it doesn't work here ... [19:46] well, appears that the behaviour is more random than I thought [19:46] well, as a matter of fact, it does work, because the problem doesn't exist here [19:47] the list of BAD INODE ... [19:47] it seemed to me, that the files are all generated/opened by syslog ... [19:48] could you, just for a test, disable the syslog inside the vserver ... and see if the BAD INODE go away? [19:48] yes [19:48] remaking a fresh vserver was we speak [19:51] better now [19:51] only: [19:51] Jan 27 17:51:56 cia kernel: ··· BAD INODE: add_dquot_ref(de65d000[101], 1) d503b5e0[101](00000000) »null« [19:51] Jan 27 17:51:56 cia kernel: ··· BAD INODE: add_dquot_ref(de65d000[101], 0) d503b5e0[101](00000000) »null« [19:51] cron 3927 root 0r CHR 1,3 4358148 /dev/null [19:51] cron 3927 root 1w CHR 1,3 4358148 /dev/null [19:51] cron 3927 root 2w CHR 1,3 4358148 /dev/null [19:51] inetd 3924 root 0u CHR 1,3 4358148 /dev/null [19:51] inetd 3924 root 1u CHR 1,3 4358148 /dev/null [19:51] inetd 3924 root 2u CHR 1,3 4358148 /dev/null [19:51] disable cron and inetd too? [19:51] hmm, no [19:51] maybe the syslog isn't stopped propperly ... [19:52] before the quota is turned off .. [19:52] syslog is stopped [19:52] oh [19:52] well [19:52] ah [19:52] I see [19:53] I did type: quotacheck -maug in the console [19:53] that is when the initial aquota files were created [19:53] did you start more than one vserver, or is it necessary to have more than one to make the setup crash? [19:53] I mean hang ;) [19:53] I am only running one at the same time [19:54] okay, so there is no chance that something outside that vserver/context has some files open, right? [19:54] well, the host system has all kind of files open ofcourse [19:55] yeah, but probably not on that partition ... [19:55] Action: mids lsofs [19:55] hm [19:55] quotaoff appears in the host lsof output [19:55] is that correct? [19:56] hmm, not really ... [19:56] well, it works in and outside, but it should be called inside only ... [19:56] so the quotaoff is done outside, hmm ... [19:57] oh wait [19:57] I was inside the vserver :/ [19:57] Action: mids goes to the host and retries [19:57] okay ;) [19:58] only the vserver process uses /vservers/test1 [19:58] good, that was what I expected ... [19:58] lsof|grep vservers shows nobody else using that partition [19:59] well, quota _is_ allowed to hang around, until the last user of quota stuff/files releases them ... [19:59] well, if I run vserver stop it keeps hanging too [20:00] when does the quotaoff happen? [20:00] when I type it [20:00] or when I the vserver is stopped [20:00] Action: mids goes to the rc0.d [20:01] okay, so could you (just for a test) make sure that nothing like syslog is having some files open, before you do the quotaoff ? [20:01] rc0.d is halt isnt it? or what rcS.d that? [20:01] S is start/single [20:02] test1:/etc/rc0.d# ls [20:02] K11cron K20exim4 K20inetd K82quotarpc K85quota K90sysklogd S20sendsigs S31umountnfs.sh [20:02] ok [20:02] let me restart the box and try again [20:02] 0 is halt/reboot [20:02] can I kill a vserver the hard way? [20:03] hmm not really ... [20:03] so I dont have to reboot? [20:05] how can I specify the vserver with vps/vtop? [20:05] you can use chcontext --ctx top/ps [20:05] root 4500 0.0 0.2 2284 1064 pts/1 S 18:05 0:00 /bin/sh /etc/rc6.d/K85quota stop [20:05] root 4501 0.0 0.1 1780 548 pts/1 D 18:05 0:00 /sbin/quotaoff -aug [20:06] ok, it definitly hangs at quotaoff [20:06] well, yes, but did you stop syslog and co before? [20:06] no, it was already borken [20:06] so I have to reboot [20:09] ok, cron/syslog/inetd stopped... [20:10] quotaoff doesnt lock [20:10] still log of debug output [20:11] hmm, it seems like (only on ext3?) quotaoff waits for all files to close, which leads to some kind of deadlock ... [20:11] I could move K85quota to K91 quota [20:11] so it is after syslogd [20:12] try that one ... [20:13] hm [20:13] ah [20:13] rc6.d [20:13] dinner time [20:13] I'll try later [20:13] ok, cu [20:14] bert, I don't remember if I asked you before: the disk quota (host) and the vserver quota patch are in one patch? [20:15] yes, they depend on each other, at least somewhat ... [20:15] ok [20:15] and ext2 fs for /vservers does not hang? [20:15] with quotaon/off I mean [20:16] well, I can't reproduce the hang on ext3 here (yet), but it looks like it only happens on ext3 ... [20:16] ok [20:17] btw, the bot that existed on freenode.net irc that used to connect both networks... what happened to it? [20:17] I don't know ... [20:17] too bad, it was helpful [20:18] hmm, well you could use a client which is able to connect to more than one server (like irssi for example) [20:20] needs cygwin to run under win :) [20:20] well I'm sure there is a win* client which can do this too ;) [20:21] probably [20:21] i'll look into it [20:21] thanks [20:22] ur welcome ... [21:01] re [21:02] xchat can [21:02] and mirc too [21:03] mirc can? [21:03] how/ [21:03] on mirc 6+ there is a checkbox on the server dialog [21:03] just found it [21:03] /connect -m server [21:04] talk about ignorance... it works great [21:04] :) [21:06] Bertl: first time on start/stop it works with quotaoff after shutting down, but second time it still hangs [21:07] I guess I know what happens .. [21:08] you do? [21:08] did some source code reading ... but I have to verify it ... [21:09] I asume that something is acessing the files (log files/syslog) before the quota is enabled ... [21:09] this later, when discovered leads to the BAD INODE messages, where I supect the quota code to do the wrong thing ... [21:10] so how comes it doesnt happen for you? [21:10] again later, this leads to blocked/locked inodes, which can't be released and to the quota lockup ... [21:10] well, in all my testscenarios, I have nothing which opens any file before the quota is turned on ... [21:11] and more important, keeps it open during the quota on ... [21:11] ah [21:11] can you try quotaon on the commandline? [21:13] not at the moment, I have to get ready, leaving in a few minutes ... but I will try and analyze in the evening ... [21:14] Tue Jan 27 18:15:45 UTC 2004 [21:14] ok :) [21:15] well evening is 23:00 CET in my vocabulary ... [21:15] okay, cu later ... [21:15] bye bert [21:15] Nick change: Bertl -> Bertl_oO [21:23] click (~click@virtualdesignz.net) joined #vserver. [21:33] bertl_oo: there? [21:33] hm, out [21:38] well evening is 23:00 CET in my vocabulary ... [21:38] okay, cu later ... [21:38] :p [21:39] 30 minutes ago ;) [21:39] darned [22:41] AHTOH (~Anton@212.1.230.115) joined #vserver. [22:41] hi ppl [22:51] *mumble* [23:01] ploing [23:27] Doener (~doener@p5082D9D4.dip.t-dialin.net) left irc: Ping timeout: 492 seconds [23:36] Doener (~doener@p5082D3CF.dip.t-dialin.net) joined #vserver. [23:57] Topic changed on #vserver by ChanServ!services@services.oftc.net: http://linux-vserver.org/ || latest stable 1.24, devel 1.3.6, exp 0.06 [00:00] --- Wed Jan 28 2004