[00:02] matta: bad prefix/mask of host? [00:03] hmm, could be an explanation ... [00:04] nothing has changes networking wise.. [00:05] i've checked routing table of the server [00:05] what about the router? is this a cisco? [00:05] motherfuckers [00:05] so i called them up and told them not to reboot the server [00:05] as it doesn't look like a server problem [00:05] i just got a server reboot message... [00:06] lol [00:06] ;) cool stuff ... [00:06] s/stuff/staff/ [00:07] matta (matta@tektonic.net) left irc: Quit: Hey! Where'd my controlling terminal go? [00:09] monako (~monako@194.186.248.64) joined #vserver. [00:10] hi monako! [00:10] hi [00:25] Bertl continue talk ? [00:26] yup, anytime ... [00:28] enrico ? [00:29] shadow: program is nearly finished; display stuff is missing [00:30] perfect ... [00:30] ensc you immediately coding ? [00:31] shadow: what do you expect else? Drawing big UML diagrams? ;) [00:31] who not ? [00:33] is the current list of limits online anywhere? [00:33] 1) cpu [00:33] 1a) cpu time [00:33] 1b) nr processes [00:33] 2) total memory [00:33] 2a) virtual memory [00:33] 2b) locked memory [00:33] 3) resident memory [00:33] 4) disk usage [00:33] 4a) disk space [00:33] 4b) inodes on disk [00:33] 5) network baundwith [00:33] 5a) bandwidth [00:33] 5b) nr connections [00:33] 6) file handles [00:33] 7) terminals (pty/tty) [00:33] I take this as a no ;) [00:34] we can remove points 1,4,5 because it name of resouce group and reorder. [00:36] you forgot the sub-vserver maximum ... (for vserver inside vserver) [00:36] and 2 also.. total memorry == resident + virtual(swap) [00:37] should we reserve alimit for max NR CPUs ? *G* [00:39] bandwidth will be a property of the interface, what do you think alex? [00:39] it realy need ? i don`t confidence.. [00:39] yes. [00:40] same for disk limit , this should be per partition, right? [00:41] hm.. you think vserver have more one partitions inside ? [00:41] could be, the quota hashes are per partition ... [00:44] ahh, we should have a look at the RLIMITS .. some of them might be interesting for per context too ... [00:44] but should be merge it to per vps diskquota ? [00:45] I think a per partition and context quota makes sense .. think about using a separate /var or /tmp ... [00:46] hm.. [00:50] alex, do you need to set maximum hard limits for the RLIMIT things like core size and such? [00:50] i think it realy need only if use unifyted vps and all vps dirs (execpt users home dir and other workplace) be mounted to r/o... [00:51] matta (~matta@69.57.156.208) joined #vserver. [00:52] ok. i open RLIMT.. [00:54] i think from RLIMIT me need only [00:54] so my hosting company hires some poor employees [00:54] they saw Bertl's inode without dqh error on the console [00:54] #define RLIMIT_CPU 0 /* CPU time in ms */ [00:54] rebooted my server, and told me my hard disk is failing. [00:54] hahahaha ... sorry [00:55] then... [00:55] but not in ms - in % [00:55] they tried to tell me when they tried to login the server rebooted itself [00:55] and i'd believe it, except I saw a system shutdown broadcast [00:55] which is only done when reboot or shutdown is ran [00:55] grr [00:58] @alex do you allow to set ulimit inside the vserver? [00:59] yes [00:59] i add it in near future.. [00:59] what about a maximum for those rlimits? [01:00] i think - for first check context limits - after ulimits [01:00] but then, you cannot modify afterwards ;) [01:01] for context limits we can add also [01:01] #define RLIMIT_LOCKS 10 /* maximum file locks held */ [01:01] good point ... [01:01] my current list is now: [01:01] 1 CPU TIME [01:01] 2 NR PROCESSES [01:01] 3 MAX NR CPUs [01:02] 4 MEMORY VM [01:02] 5 MEMORY RSS [01:02] 6 MEMORY LOCKED [01:02] 7 FILE HANDLES [01:02] 8 FILE LOCKS [01:02] 9 TERMINALS [01:02] not time - % of cpu [01:02] yeah, okay, where I would suggest not to use % but something (0-65535) [01:04] ok. not % - but in parts of full cpu time. [01:04] yeah, agreed ... [01:05] it more easy for per context CPU scheduler.. [01:05] Bertl: program is in cvs (SYSCALL_SWITCH branch); src/ctxlimit [01:05] what's wrong with Sam's approach ? [01:05] will it compile with 2.95? [01:06] @matt you worked with it, could you explain what you adjusted? [01:07] Bertl: I think... [01:10] thanks ... any syntax/usage example? [01:11] ctxlimit -a -HS --9 30 -M --10 2 --> sets hard and softlimit of 9 to 30, minimum of 10 to 2, and displays settings [01:12] how many syscalls for that? [01:12] the three new ones [01:14] what I meant was, will ctxlimit (hmm maybe we should name it vsrlimit?) -H --9 30 -S --9 20 -M --9 10 do 1 or three calls to set_rlimit? [01:16] set will be called twice [01:16] huh? [01:16] or 1 time in your example [01:16] ahh okay ... [01:17] so you do accumulate the values ... why am I asking ;) [01:18] yes, there is a vc_ctx_limit[32] array which will be set by teh getopt() part and evaluate at the end of the program [01:24] dmistry (~dmistry@su-nat.datapipe.net) left irc: Quit: My damn controlling terminal disappeared! [01:26] @enrico how do I checkout for this branch? [01:27] shadow (~umka@212.86.233.226) (unknown) [01:27] ł ircname : Alex Lyashkov [01:27] | channels : #vserver [01:27] ops. sorry [01:27] cvs -d:pserver:anoncvs@subversions.gnu.org:/cvsroot/util-vserver login [01:28] I've checked out util-vserver already ... how to access this branch? [01:28] cvs -z3 -d:pserver:anoncvs@subversions.gnu.org:/cvsroot/util-vserver co -r SYSCALL_SWITCH util-vserver [01:28] or 'update -r SYSCALL_SWITCH' [01:28] ahh okay .. thanks .. [01:29] how do I compile it? [01:30] autoreconf -i -f [01:30] ./configure [01:30] make [01:30] I do not get a ./configure ... [01:31] which autoconf version? [01:31] or automake? [01:31] 2.13 autoconf [01:31] you will need 2.5 [01:31] hmm ... [01:32] Bertl: do you have an autoreconf2.5 ? [01:33] not yet ... [01:33] you will need automake1.7 also [01:34] hmm .. dont have automake yet ... [01:34] Mandrake seems to be really conservative with development tools... [01:35] nope, I am .. ;) [01:35] well.. i want to sleep.. [01:35] mandrake 9.2 has autoconf-2.13 and automake-1.4 [01:35] but for all. and good evening.. [01:35] Nick change: riel -> unriel [01:35] good night alex! [01:36] Bertl: the newer versions seems to be configured badly [01:36] Thanks Herber and for you too.. [01:36] shadow (~umka@212.86.233.226) left irc: Quit: to bed [01:37] autoconf 2.5 is no problem on mandrake ... [01:38] Bertl: autoreconf-1.5 contains: my $autoconf = $ENV{'AUTOCONF'} || '/usr/bin/autoconf'; [01:38] that means, it chooses the wrong version [01:43] jhh (~heuing@194.245.114.189) joined #vserver. [01:43] hi jhh! [01:47] hi bertl... [01:47] didn't get any further yet, too busy. I thought I just join the channel if anything helpfull passes by :) [01:50] there is one question i still have... it's a fairly small group of people developing vserver, is there any community support in terms of development or money ? [01:50] how do you make sure to keep the project running ? [01:51] hmm, good question(s) ... [01:51] :) [01:51] I thought it is a community project already? [01:52] we are very active in the typo3 sector (webbased cms if you don't know about it...), the community base is very large, but sponsoring the main developer is still not much... [01:52] At least there's more traffic on the ML than on the opter projects I watch, so I don't worry much :) [01:52] if you talk about community, it's just you and a few others, which I would not call community if it's just a hand full of people :) [01:53] communities come in all sizes :) [01:53] in the typo3 project, we didn't get the 10GB listserver and news-traffic, but nearly :) [01:53] Bertl: do you got it compiled? [01:54] hrm ... working on automake/autoconf ... [01:54] vserver is getting *really* useful currently with the per-context quota and all. I think the project will continue to grow. [01:54] am I right with this: you all work on this in your spare time, and there is no professional support like a company sponsoring (you are employed at a company who is paying for you doing so), or companies paying for it in general ? [01:55] jhh: feel free to sponsor Bertl. :) [01:55] yes, *smile* [01:55] difficult right now, but thought about it [01:56] one thing I could easily do: we are hosting the mailinglists and newsserver for typo3, we could do that for typo3 as well. as you just go the new mailinglists, I could still add a news gateway for those of you interested in using the thing via newsgroup... [01:56] Action: Hurga is right now moving a server to vserver. So I guess I'll continue to contribute one or other way. [01:56] sorry, exchange the last typo3 with vserver ;) [01:57] news://news.netfielders.de [01:58] hmm, do you see any value in a news gateway, (don't get me wrong, have nothing agains it) [01:59] @enrico hey autoconf2.5/automake1.7 does the trick ... [01:59] well, yes. I am annyed of all these mails via mailinglist. and you can easily follow the threads on the news server. I always said things like you just did, adn I agreed on setting it up. after I gott used to it I realy like it more than the mailinglist way :) [01:59] both lists will be in sync anyway [01:59] jhh: Get a threading newsreader [01:59] hurga ? [01:59] like, mutt. [01:59] hmm, well mutt/procmail does the threading/sorting for me ... [02:00] I don't think the vserver crowd is big enough for netnews... [02:00] Besides, the gateway might gate spam back to the list. [02:00] ah, you meant he archive ? the archive for typo3 is a very powerful thing included on typo3.org. it builds it's own archive. you can setup faq items through mailinglist mails, the list-managers can do that [02:01] @jhh what about spam protection? [02:01] matta (~matta@69.57.156.208) left irc: Ping timeout: 492 seconds [02:01] hm, never heard that spam is sent via newsgroup... but anyway, there is one small problem: if you have to authenticate for the mailinglist, you could just write to the news server, but the mail will not be sent to the listserver. we do no authentication, but will add spam filter [02:02] spam: spamassin will be enabled shortly [02:02] just did the update to the new mailman version first... just finished [02:03] automatic archive: http://typo3.org/1422.0.html [02:03] okay, I would say, make a suggestion on the mailing list, contact Martin (this is where the mailing list is and will stay for now), and we add a news.linux-vserver.org ... what do you think? [02:04] @enrico /usr/src/ENRICO/util-vserver/src/ctxlimit.c:127: undefined reference to `vc_get_rlimit_mask' [02:04] good idea. I'll write a mail to the mailing list some time soon. I'll discuss the authentication thing with martin, as this might be a little problem [02:06] Bertl: uups, forgot to checkin the lib directory; update cvs, call automake && config.status --recheck again... [02:09] okay, I have a binary ;) [02:10] @enrico could we change the name to vsrlimit or vslimit later? [02:11] no problem, but what is wrong iwth ctxlimit? [02:12] I'm currently moving away from ctx, and try to use vs where possible ... like in the patches ... [02:14] what is with 'vlimit'? [02:14] if you like .. (as a matter of fact, I don't care too much ;) [02:48] monako (~monako@194.186.248.64) left irc: Remote host closed the connection [02:50] monako (~monako@ts1-a64.Perm.dial.rol.ru) joined #vserver. [02:50] monako (~monako@ts1-a64.Perm.dial.rol.ru) left irc: Client Quit [02:50] hmmm. [02:50] hydrogen[root]:~> quotacheck -maug [02:50] quotacheck: Cannot get quotafile name for /dev/hdv1 [02:51] what tools, is this inside the server? [02:51] Yes. Old distro, I recompiled quota though... [02:52] [V] cat /etc/mtab [02:52] hydrogen[root]:~> cat /etc/mtab [02:52] erf [02:52] /dev/hdv1 / ext3 rw,usrquota 0 0 [02:53] change ext3 to ufs ... [02:53] [V] ls -la /dev/hdv1 [02:53] no change [02:53] brw-r--r-- 1 root root 4, 0 Nov 7 04:46 /dev/hdv1 [02:54] okay this vroot device is set up, I assume ... [02:54] it should. lsmod says module vroot is busy, too. [02:55] [V] quotacheck -Fvfsv0 -maug [02:55] the same [02:56] do you ahve strace? [02:56] yes. [02:56] which version? [02:56] version 4.2 [02:57] hmm, is it possible to upgrade to 4.5? [02:57] I can try :) [02:57] (this is inside vserver) [02:57] *nod* [02:58] before I do.. this is what I did. [02:58] It's a RH 6.2 running on a RH 9 host. [02:59] I copied the host's (RH9) kernel stuff over, took the RH9 source RPM and recompiled on the RH 6.2 [03:00] anything badly wrong with it so far? [03:00] should be okay ... [03:03] uhh, do I need to give CAP_QUOTACTL? [03:03] yup! [03:03] gah [03:04] just a moment :) [03:08] didn't help. [03:08] what version is this quotactl? [03:09] quotacheck I mean ... [03:09] Quota utilities version 3.06. [03:09] okay add CAP_SYS_ADMIN too ... [03:09] ook... [03:10] although I suspect your problem lies somewhere else ... [03:10] I used patch-2.4.22-vs1.00.diff and patch-2.4.22-ctx17a-vr0.13.diff on plain 2.4.22 [03:10] And I run vserver on LVM [03:11] okay, if CAP_SYS_ADMIN doesn't work, strace it is ... [03:13] so it is strace. [03:14] compilation error, gah. [03:15] hmm, you got the sources? tar.gz? [03:15] strace-4.5.tar.bz2 [03:15] configure/make? [03:15] sure. [03:15] time.c:363: `TIMER_ABSTIME' undeclared here (not in a function) [03:15] time.c:363: initializer element for `clockflags[0].val' is not constant [03:15] make[1]: *** [time.o] Error 1 [03:16] okay, sounds strange ... [03:16] do you try inside the vsever, and what kernel headers are there? [03:16] inside the vserver, and the kernel headers of the host [03:17] hmm not good ... [03:17] kernel headers correspond to the glibc ... [03:18] I can use the old ones, which were native to the vserver [03:18] ok? [03:18] that would be better ... [03:18] ok. [03:19] same goes for compiling the quota tools by the way ... [03:19] same error... [03:19] strange ... and if you compile strace on the host? [03:21] compile strace on the host works [03:21] okay, copy the strace into the vserver ... [03:22] but you have to check why this fails ... it seems that the vserver is somewhat fishy ... [03:23] copy the strace into the vserver: well, glibc is too old [03:23] well, well, well ... you could compile it static ... [03:23] is 4.2 the latest RH package for that? [03:24] I think so. [03:24] let me look for it ... [03:24] And I could need a hint on how to compile staic :) [03:24] static, even [03:25] ftp://ftp.nluug.nl/pub/os/Linux/distr/RedHat/ftp/redhat/linux/beta/severn/en/os/i386/RedHat/RPMS/strace-4.5-1.i386.rpm [03:25] wow. [03:26] uh. [03:26] error: failed dependencies: [03:26] libc.so.6(GLIBC_2.3) is needed by strace-4.5-1 [03:26] this is RH 6.2 [03:26] hum, in the vserver? [03:27] The vserver. Not really recent... [03:27] *nod* [03:27] ahh okay, I thought your server was 9.0 and the host 6.2 ... [03:27] no! Did I say that? sorry... [03:28] well, do you have the rpm build system on that vserver? [03:28] Yes. [03:28] strace SRPM... [03:28] ftp://ftp.nluug.nl/pub/os/Linux/distr/RedHat/ftp/redhat/linux/rawhide/SRPMS/strace-4.5-1.src.rpm [03:30] same error as when I tried to compile from source. [03:30] (well, that somehow was to be expected) [03:30] what gcc is this? [03:30] egcs-2.91.66 [03:31] just tell me how to do a static build of strace... [03:31] well, if you do the dynamic build, then there should be a line almost at the end .. which does a gcc -o strace ... [03:32] this might go over several lines actually ... [03:32] do you see it? [03:32] yes. [03:33] okay copy that line and add a -static after gcc ... [03:34] ok, let's rock... err, strace. [03:35] fine ... [V] strace quotacheck -Fvfsv0 -maug >/tmp/qc.log 2>&1 [03:37] dcc? [03:37] http? [03:37] well, my web server is the vserver, you see... :) [03:38] okay dcc [03:39] hrm, hrm what about the 'ext3'->'ufs' change I advised? [03:39] didn't work. ok, again [03:41] no significant cange. Do you want that strace, too? [03:41] nope ... [03:44] okay I'll try to reproduce ... [03:44] could you recompile the kernel (in the meantime) with the vr0.14 released for vs1.00 ? [03:45] just to make sure we have the same codebase ... [03:45] sure. - Just a thought. Should I try with insecure quota? [03:46] well you can do that, but I doubt that it changes anything ... [03:46] ahh, wait, just a moment [03:46] where is the lvm volume mounted? [03:47] /dev/volg1/lvol2 10079084 8158152 1408932 86% /vservers/tig [03:47] and tig is the vserver, right? [03:47] Yes. [03:48] hmm, okay, that should be fine ... [03:50] hmm. I have a nearly identical machine where quotacheck -maug [03:50] works. [03:50] with the new kernel? [03:50] the vroot quota device is not set up there, thiugh. [03:51] hmm, and if you set it up? [03:51] no LVM there... [03:51] ahh okay ... [03:52] it's just on a partition... [03:52] shared partition you mean ... [03:53] Yes. [03:53] Nick change: surriel -> riel [03:55] By the way, RH9 on the RH9 host works fine. [03:55] including the quotacheck? [03:55] Yes. [03:55] well then it's probably the tools, that won't work with this kernel ... [03:56] I used the same quota, recompiled on the vhost... [03:56] on both servers? [03:56] ahh, damn, I forgot, check if /proc is mounted in the failing server ... [03:57] Yes. RH9 quota (v. 3.06) everywhere. [03:57] /proc is mounted and works [03:58] hmm, hard case ... try to strace both, and we look for the difference ... please make sure that mtab is the same, etc, etc ... [03:58] ok. [04:04] grr! [04:04] quotacheck -cmaug [04:04] -c Don't read existing quota files. [04:04] hmm a feature? [04:05] I did not have any quota file there (quota was on another partition before). [04:05] well good to know ... we've learned something ;) [04:06] I think I had that before even :( Just didn't remember... [04:09] quotaon: Quota format not supported in kernel. [04:09] *siigh* [04:10] you can specify the format with the -Fvfsv0 or -Fvfsold ... [04:11] ! [04:15] excellent. [04:19] Thanks a lot for your time! [04:19] well, you will do some testing for the project, right? [04:19] I think I am already. I'll whine a lot if this server crashes or something ;) [04:20] so you'll give something back if you report your findings/results ... [04:20] I will. [04:22] Action: Hurga needs to port quota info from one host to another. Well, this has nothing to do with vserver at least. [04:22] hmm, could be useful for others too .. [04:37] uh, simple. copy the quota file and run quotacheck again. without the c this time... [04:38] hmm, will give you troubles if you want to change to quota format ... [04:38] and I would advise to go for vfsv0 in the long run ... [04:39] I got errors when I tried the default, which is the newer version. But I just want to run this vserver until I manage to port everything over to one with a recent distro... [04:40] yeah, but maybe a quotarep | script stuff would be useful to transfer the quota ;) [04:40] definetely. [05:31] jhh (~heuing@194.245.114.189) left irc: Ping timeout: 493 seconds [05:33] ah, per vserver uptime patch announcement :) [05:33] yup! [05:34] cool :) [05:42] i like the sys_reboot one too, that sounds like a good solution [05:43] hmm... are there any plans to get the vserver stuff into the main kernel, BTW? [05:44] hmm, well yes, there where, but I doubt it will be before 3.0 ;) [05:44] aww. [05:45] I don't think that the whole vserver stuff belongs into the mainstream ... and the kernel maintainers are _very_ creative in what goes in and what not ... [05:45] heh [05:46] creative is a euphemism for random eh? [05:46] Is BSD jail an addon? [05:47] not yet ... why? [05:48] I mean, how do the BSD folks handle this? [05:48] If it doesn't hurt, I don't see a reason why this shouldn't be in the kernel... [05:49] well, okay, let me give an example ... [05:49] you know mount --bind ? [05:49] yup [05:49] ever tried to mount something with mount -o ro --bind ? [05:50] no, I read it doesn't work. [05:50] would be quite useful, though. [05:51] well there is a patch since 2.4.22pre4 or so, which I try to get into the kernel 2.4 and 2.6 since more than 3 month now ... without any succes ... not even a response from the maintainer ... [05:52] not good... [05:52] Action: Hurga wonders if Rik might be able to help :) [05:53] the kernel maintainers said, check with Al Viro ... but he doesn't respond ... [05:53] Keep trying, sometimes folks are busy or somesuch... [05:54] This is the drawback to linx as a community project :( [05:54] well, you can be damn sure that I will keep trying ... ;) [05:54] Sometimes it's just chaos. [06:21]