[00:02] Hi Jack [00:13] mhepp (~mhepp@r72s22p13.home.nbox.cz) left irc: Remote host closed the connection [00:23] hi [00:24] I have vserver 0.26 with this. I will send it to you (bertl). Now I have a problem with linux-vserver.org. We need a clean download area... [00:25] correct, I would suggest to make a release and experimental dir ... [00:25] or development [00:25] and we should add the releases/devel on the wiki as planned ... [00:29] @jack there is no problem if you release your tools, as I do not maintain them ... [00:29] Bertl - you not have create one dir for stable(release) versions and second for development snapshots.. ? [00:29] @alex yes, this is the idea ... [00:29] we will do that ... [00:44] JonB (~jon@129.142.112.33) left irc: Quit: later [02:03] @jack, did you spend some thoughts on the network virtualization yet? [02:06] No sorry. [02:06] no problem, just wanted to know ;) [02:07] I have to leave. [02:07] okay, when to expect 0.26? [02:07] Just want you to known one thing about vserver and me [02:07] I can send it to you right now [02:08] Anyway about vserver and me [02:08] I was overload with tons of problems, work and what's not and finally start to see the light and this happened exactly the week you suggest to take over the project. [02:09] Since I have managed to stay somewhat on top. This may look like I am reacting to the change in the project and trying to grab it back [02:10] This is not the case. Pure coincidence that I have managed since 2 week to read and answer my email. I hope it will stay like this. [02:10] Just to clarify this. [02:10] hey jack, as I said I'm really glad that you are back, and I hope we will work together on vserver whenever possible .. [02:11] we just have to coordinate ... whatever part you want to take in the development process, it's yours ... [02:11] fine. Now I am leaving (I still have 4 kids, and this is another story) [02:11] have a nice evening ... [02:12] Oh btw, what happend to the mailing list. Last week, I moved all my mailing list to mailman. The noise went down and performance up. [02:12] nice evening, jack.. [02:12] have you seen the new mailing list martin has put up? [02:13] we have threaded archives and he is working on improving them further ... [02:14] Has this been announced ? [02:14] yes ... and we intent to switch at the end of this month ... [02:15] Ok. Just tell me and I will put an alias. [02:15] Go weekend [02:15] jack (~jack@206.162.172.138) left irc: Quit: using sirc version 2.211+KSIRC/1.2.4 [02:16] hi alex .. http://vserver.13thfloor.at/Stuff/syscall2.2.txt, is this okay? [02:16] hi all :) [02:16] I am getting line 19: [: =: unary operator expected [02:16] what programming language? ;) [02:16] when starting services in RH9 [02:17] probably bash [02:17] hmm, sounds like some shell variable is not set ... [02:17] i c [02:19] # Check that networking is up. [02:19] [ ${NETWORKING} = "no" ] && exit 0 [02:21] probably this isn't set in the /etc/sysconfig* [02:21] the variable NETWORKING should be set in /etc/sysconfig/network [02:22] yep. that was it. typo in NETWORKING [02:22] :) thanx [02:23] you're welcome ... [02:27] Bertl> okey. it`s not have new parts - only numbering cels.. [02:27] yes ... [02:32] how i understand - assign commands to cels ? or you want other ? [02:32] no, that's exactly my idea ... [02:33] just assign all present and future commands to those cells/numbers ... [02:35] at col 1 we must place all my "get" commands... it corrent ? [02:36] hmm, could you list them to help me remember? [02:36] correct.. [02:41] http://www.freevps.com/download/command_list [02:43] hmm, you don't mean _all_ your get commands, do you? [02:45] it`s primary commands - other can be added simular. [02:46] it`s primary commands - other can be added similar. [02:47] get max process capablity for example? [02:48] I would put it into the [12] [02:49] get ? [02:50] yes, maybe we should add INFO to that COLUMN? [02:51] for get statistic we reserverd 0,8,16,32,... or you have other thoughts about it colum ? [02:51] for get statistic we reserverd 0,8,16,32,... or you have other thoughts about it column ? [02:51] hmm you consider this statistics, okay, then [08] ... [02:52] I thought this was some setting you did with set before ... [02:53] yes. i consider all my "get" commands as statistics command. [02:54] hmm, but it's not statistical if you set it before ... so if this is a readback of a previous set value, we should consider putting it into the control column, and add INFO to make that clear ... [02:55] or if you prefer change the QUERY to INFO (more general) ... [02:56] "set" - is control. but if we get installed values or property - it`s statistic ? or not ? [02:57] okay, let us change the QUERY to INFO and you put it in the first column, yes? [02:58] http://vserver.13thfloor.at/Stuff/syscall2.2.txt [03:01] other way - transfor INFO/QUERY column to 3-dimensional. because many commands have reversed command for get statistic from it. it need 1 bit = one column. [03:03] hmm, well you can always have get/set command in the same cell, that is not a problem ... [03:04] the more I think about it, the better I like that ... [03:04] or do you mean something else? [03:07] but "get" is statitic ? [03:08] and me not change matrix - we rename (transform) one colomn.. [03:08] get is INFO, but it doesn't matter, since we changed the culumn label ... [03:08] which column would you like to rename? [03:10] STATS - but rename can be bad word.. transform to 3D. and 4D is commands. [03:11] hmm, so you want to use one bit from the COMMAND for get/set, right? [03:12] can be command - can be category. but yes. [03:12] we don't have it in category, so it would be command ... if you like that, why not ... 0-127 (set) 128-255 (get) okay? [03:13] okey :) [03:13] but personally I don't see any reason to do it .. because you can also use 17 (set), 18 get for example ... [03:14] for easy identification.. as matrix.. [03:15] as I said, no problem with that .. if you like to extend this to get/set/modify/other (2bits) it's okay for me too ... [03:17] i think more ease do modify in userland - in kernel have more atomic operations... [03:21] okey.. create\destroy context it for cel 01 ? [03:21] yep! [03:23] set schedeling police - it 14 ? [03:24] or [12] but [14] seems better ... [03:24] i agree 14. [03:25] set CPU usage - 12 ? [03:25] agreed! [03:26] first part finisned :) [03:27] sorry i make a tea and return :) [03:27] make it so ... [03:34] i`m back.. [03:34] is/was the tea good? [03:36] yes :) i like ahmand tea with jasmine.. [03:39] i hope you too ? [03:40] I like tee, but I drink it with cinnamon sticks ... [03:41] hm.. not drink with it :) [03:42] well, okay, I don't drink the sticks ;) [03:42] well, returned to fill cells :) [03:43] you have remarks for fill first row ? [03:43] migrate (context change) [03:43] 03 . [03:44] chsaferoot, chnamespace ... [03:44] 02/04 - unused ? [03:44] hm.. namespace is - vfs object... [03:45] okay, right ... [03:45] 04 host/system limits .. hmm ... [03:46] nothing yet ... but we'll see ... [03:46] it can be only extened capablites :-\ if it need. [03:46] ahh, good point, just forgot it ... [03:46] per context capabilities (sub/finer grained) [03:51] okey. [03:52] finish on row ? [03:52] for now, yes ... [03:53] @matt http://vserver.13thfloor.at/Experimental/patch-2.4.23-pre7-O1.2-c17h.diff http://vserver.13thfloor.at/Experimental/patch-2.4.23-pre7-O1.2.diff [03:56] ok.. [03:56] herbert - i want sleep :) [03:56] 3 am :-) [03:56] no problem ... we continue tomorrow? [03:57] of course :) [03:57] have a good night ... [03:58] thanx :) for you too [03:58] @MrBawb could you give the O(1) c17h a try? [03:59] shadow (~umka@212.86.233.226) left irc: Quit: go to bed [04:04] I can when I get home in an hour :) [04:04] perfect ... [04:05] what is the relationship between patch-2.4.23-pre7-O1.2-c17h.diff and patch-2.4.23-pre7-O1.2.diff? [04:05] are they both needed? [04:06] yes, as usual (for me) the patch-2.4.23-pre7-O1.2.diff is the scheduler and patch-2.4.23-pre7-O1.2-c17h.diff is the c17h stuff ontop of that ... [04:06] ah ok [04:07] since our schedules seem to be desynced, would you like me to email you the results? [04:08] if you don't find me here .. (Bertl_zZ) then, yes please ... [04:09] it would be great too if you could have it running for some time (lets say a day or so) under some load (e.g. frokbomb in one context) ... [04:10] what's your email? [04:10] herbert@13thfloor.at [04:10] ok [04:21] serving (~serving@213.186.190.62) left irc: Ping timeout: 485 seconds [04:25] serving (~serving@213.186.190.64) joined #vserver. [04:30] morning [04:30] night! [04:31] morning! [04:31] hehe [04:31] as I said, night! [04:31] how are you bertl? [04:31] fine, thanks, how are you? [04:31] good, life is peachy [04:32] how are the vserver plans going? any cool new developments? [04:32] yes, we have a c17h on O(1) scheduler ... wanna test? [04:33] sure, why not...does it have ml, or is that pushing my luck ;) [04:33] not yet, but I guess you can safely add that one ... [04:33] that on your experimental page? [04:33] ml is cool [04:33] http://vserver.13thfloor.at/Experimental/patch-2.4.23-pre7-O1.2.diff [04:33] http://vserver.13thfloor.at/Experimental/patch-2.4.23-pre7-O1.2-c17h.diff [04:33] i use it on my servers and it seems to work quite will [04:34] http://vserver.13thfloor.at/Experimental/patch-2.4.23-pre7-O1-c17g2-ml0.04.diff.bz2 [04:34] that one could work with it ... [04:35] but the O(1) scheduler stuff is very experimental, so be warned ... [04:35] aah [04:35] you might lose your best processes to it ;) [04:35] okay [04:35] hehe [04:35] hmm, what a dilemma [04:35] on the other hand, matt reported that it does miracles on his system ... [04:35] what i really need is a test box [04:36] heh :) that's a big call [04:36] he just had some troubles on his SMP system ... but that was another patch (one he adjusted himself) [04:36] just fyi, i'm running vserver on slackware and the tools all mostly work [04:37] okay [04:37] cool, I thought about putting up some wiki page where everybody can tell how and under what circumstances he/she uses vserver, what do you think about that? [04:38] i think it would be excellent [04:38] wikis are good [04:38] i saw some mention of the utilities being rewritten in C...are those going to replace the existing ones? [04:39] we decided that there will be two branches for userspace tools ... [04:39] ideally, the tools would be distribution aware [04:39] jack agreed to maintain his utilities (and this means adapting them to future stable versions) [04:39] oh [04:40] those are the C ones? [04:40] but the 'mainstream' if we have something like that, are enricos util-vserver ... [04:40] those are 95% C now and might become 100% C in the near future ... [04:41] so this: vserver-0.22.src.tar.gz [04:41] this doesn't include the bash scripts, which are basically the same on both userspace tools ... [04:41] is jacks stuff? [04:41] vserver-0.XX up to 0.24 is jack [04:41] util-vserver is enrico ... [04:41] okay [04:42] hmm, so if i wanted to add slackware support to the toolbase i should do it on util-vserver then... [04:42] I would suggest to do so ... but maybe a mix of both could be advantageous ... [04:43] for example, enrico doesn't include the newvserver stuff, because he isn't able to test it ... [04:43] righto [04:44] why not? [04:44] you have to ask him yourself ... [04:45] are aware of anybody having run vserver on non-x86, eg. sparc64? [04:45] s/are/are you/ [04:46] yes, we have someone on x86_64 and somebody who wanted to test on sparc/ultra sparc ... but got no feedback so far ... [04:46] hmmm [04:46] that is cool [04:46] it should run on sparc/sparc64 as I tried to add those platforms ... [04:46] i might test it at work [04:46] i am running debian on an ultra-10 [04:47] I have no such machines, and no access to them, but I'm willing to help if you encounter any issues with vservers on that kind of hardware ... [04:48] okay, i'll give it a go on monday and let you know how it turns out :) [04:48] perfect ... [04:48] I would suggest testing c17h, c17f in this order ... [04:49] okay [04:49] you will require latest tools (util-vserver) for c17h .. [04:49] but they will work for older kernel patches too ... [04:49] latest, as in cvs? or latest release? [04:50] he has his cvs on savannah, so pay a visit there, link is on linux-vserver.org ... [04:51] yep [04:52] so why has jack stepped back from maintaining vserver? real-life? [04:53] I guess it just wasn't fun anymore ... [04:53] :( [04:54] probably some (read a lot) troubles in real-life ... [04:54] but I'm glad, he still is working on this project ... [04:55] yeah [04:55] i'm glad the project is still going...it opens up a lot of possibilities [04:56] yeah?, tell me about them ... [04:56] i particularly like the idea of high-availability vservers [04:56] floating from host to host...very cool [04:56] i read in the archives that some guy had accomplished this on top of drbd and heartbeat [04:56] yes, I actually did some tests of this scenario ... with drbd ... [04:57] was it effective? [04:57] well, it worked reasonably good ... [04:57] blueshoe (~cocoons@phylogenomics.Berkeley.EDU) joined #vserver. [04:57] probably you can do without drbd, if you sync via rsync or something similar ... [04:58] hi blueshoe? [04:58] yeah, that was another possibility [04:58] gotta go, relatives have arrived [04:58] cu! [04:59] i just used debain's make-kpkg to compile a new kernel with the ctx patch, and it compiles fine, but when i boot and try chcontext 1 it complains [04:59] hmm, what exactly is in this kernel? [04:59] "can't set the new security context" ": Function not implemented" [04:59] 2.4.21 [05:00] vanilla, with what patches? [05:00] linux, debian, i386 [05:00] just the ctx patch (the "kernel-patch-ctx" deb) [05:00] I did a debian patch some time ago, maybe you should try that one? [05:00] yeah, i was going to try manually patching the kernel next [05:01] http://vserver.13thfloor.at/Experimental/patch-deb0.2-2.4.22-c17f.diff.bz2 [05:01] that is for 2.4.22-3 IIRC ... [05:01] i was just hoping maybe there was something i was missing that i could easily fix :) [05:01] we can have a look at your kernel if you want ... [05:02] well, i don't want you to go to that much trouble [05:02] what does 'grep __NR /proc/self/status' return? [05:03] nothing [05:03] hmm, maybe you didn't install this kernel then? [05:03] maybe you didn't patch it? [05:03] yeah, i'm running it now [05:03] what ctx patch did you use? [05:03] in the compile output i see "Patch /usr/src/kernel-patches/i386/apply/ctx processed fine" [05:04] i used the "kernel-patch-ctx" deb from the unstable tree [05:04] 17-1 [05:04] hmm, can you give me an url? [05:05] i just did an "apt-get update" and selected it through aptitude... but i'll see if i can find it on the web somewhere [05:05] maybe you can put it somewhere on the web then? [05:06] http://packages.debian.org/unstable/devel/kernel-patch-ctx.html [05:07] by the way, vserver is a really cool tool [05:07] thanks for making it happen :) [05:07] thanks for using it ... [05:08] my pleasure [05:09] + buffer += sprintf (buffer,"__NR_new_s_context: %d\n",__NR_new_s_context); [05:09] + buffer += sprintf (buffer,"__NR_set_ipv4root: %d rev2\n",__NR_set_ipv4root); [05:09] those two lines are added to your kernel .. if the patches are applied correctly .. [05:10] so you should see those lines in /proc/self/status ... [05:10] if not, they wheren't applied or the kernel wasn't booted ... [05:11] what does 'cat /proc/version' return, by the way? [05:12] Linux version 2.4.21.031024a (root@arcturus) (gcc version 3.3.2 20030908 (Debian prerelease)) #1 Fri Oct 24 17:21:49 PDT 2003 [05:12] that's the kernel i just compiled [05:13] hmm, well it should have a ctx-17 in the version ... [05:13] really? [05:14] I don't know the debian kernel building process, but the patches suggest so ... [05:14] what i did was: "fakeroot make-kpkg --append-to-version=.031024a --added-patches=ctx kernel_image modules_image" [05:16] well, I assume that is the right thing to do (I really don't know) but in all patches on the url you gave me the kernel version is changed to blabla-ctx-XX ... [05:16] but I don't see a 2.4.21 patch in there, only 2.4.18/19 and 20 .. so maybe this is your problem ... [05:17] IIRC debian stable (woody) is 2.4.20 kernel, right? [05:17] don't know [05:17] i only run unstable [05:17] okay, in this case the 2.4.22-3 should be yours ... [05:18] i do remember that there used to be ctx patches just up to the 2.4.20 kernel [05:18] so maybe that is the problem [05:18] well there are patches up to 2.4.23-pre7 ... [05:18] there are just no debian packages for that ... [05:18] i meant debs [05:18] right [05:19] well, it looks like i'll be compiling a kernel by hand [05:19] Action: blueshoe sighs [05:20] make-kpkg has spoiled me [05:20] http://packages.debian.org/unstable/base/vserver.html [05:21] that's 0.23-3 [05:21] i have 0.23-2 [05:21] that's only the tools, though [05:21] not the kernel patch [05:21] ah okay ... sorry ... [05:22] yeah, the package descriptions are a little confusing sometimes [05:22] http://www.linux-vserver.org/index.php?page=DebianVserver [05:22] nice :) [05:23] how is the unstable release called? [05:23] i think it's called "SID" [05:24] but don't quote me on that [05:24] there's also "testing", so maybe that's "SID" [05:24] okay .. you don't know what you use then? ;) [05:24] i know i use unstable... i just don't know the proper name for it :P [05:25] "the branch that breaks all the time" [05:25] well, as shakespeare said: "what's in a name ..." [05:31] ensc (~ensc@134.109.116.202) left irc: Ping timeout: 513 seconds [05:57] RH (~john877@CC3-24.171.21.47.charter-stl.com) joined #vserver. [05:57] Nick change: RH -> netrose [05:58] hi bobi! [07:00] okay .. I call it a day ... [07:00] Nick change: Bertl -> Bertl_zZ [07:02] goodnight Bertl :) [07:54] blueshoe (~cocoons@phylogenomics.Berkeley.EDU) left irc: Ping timeout: 485 seconds [09:25] shadow (~umka@212.86.233.226) joined #vserver. [09:25] hi all [09:26] hi [11:42] serving (~serving@213.186.190.64) left irc: Ping timeout: 480 seconds [12:21] JonB (~jon@129.142.112.33) joined #vserver. [13:20] mhepp (~mhepp@r72s22p13.home.nbox.cz) joined #vserver. [13:45] serving (~serving@213.186.189.91) joined #vserver. [13:51] mhepp (~mhepp@r72s22p13.home.nbox.cz) left irc: Ping timeout: 480 seconds [14:53] alekibango (~john@b59.brno.mistral.cz) left irc: Ping timeout: 485 seconds [15:00] alekibango (~john@b59.brno.mistral.cz) joined #vserver. [15:03] JonB (~jon@129.142.112.33) left irc: Quit: Client exiting [15:24] mhepp (~mhepp@r72s22p13.home.nbox.cz) joined #vserver. [17:45] Nick change: Bertl_zZ -> Bertl [17:46] hi all! [17:46] Bertl: Hi [17:47] Hi Herbert :) [17:49] hi alex! ready to continue? [17:53] after half hour.. it`s ok ? or i finish test and be ready.. [17:53] no problem .. let me know when you are ready ... [18:06] hmm, any interest in binding vps to CPUs? [18:07] mhepp (~mhepp@r72s22p13.home.nbox.cz) left irc: Remote host closed the connection [18:13] for run programs from vps to selected CPU`s ? [18:13] serving (~serving@213.186.189.91) left irc: Ping timeout: 485 seconds [18:13] yes .. like cpu_affinity only stronger ... [18:14] we can write it to scheduler command group - and if any peoples sending request - add it. [18:15] well, the thing is somebody wrote something called CPUSETS for 2.6 and probably 2.4 .. which does exaxctly this via a syscall ;) [18:33] herbert - what you opinion about open files in one context but file inode be owned in other context.. we can disable it or not ? [18:34] you mean with tagctx files? [18:35] (persistant context tagged files, like UID24/GID24) [18:36] for me - all files taged.. but ones in context_0, other in context with id 3 ... etc.. [18:37] okay, I had a poll, on the mailing list, how to handle inter context file access .. (means file is tagged ctx 7, ctx 3 wants to access) ... [18:38] and against my suggestion to handle this gracefully (silently migrate the context) the community decided that it would be far better to handle this as access violation ... [18:38] primary context 0 can be opened this files.. or not.. [18:38] context 0 can open all files, and context 0 files can be openend by all contexts ... [18:39] same goes for the other inodes ... [18:39] hm.. for open files in all context we can add special flag SHARED.. it do more security... [18:40] hmm, where would you like to put that, and how to handle? [18:40] and what exactly should it do? allow inter context access? [18:40] #define S_IMMUTABLE_LINK 128 /* Immutable links */ [18:40] #define S_SHARED 256 /* it's shared between contexts */ [18:41] hrm, I'm sure I don't want to do that, especially because the S_IMMUTABLE_LINK is not necessary with context tagging anyway ... [18:41] but in current i use it only for control diskusage in context.. [18:41] i not count shared files.. [18:42] but me can`t use same value for it. [18:43] no, what I mean is, I don't see a reason to have any one of those flags ... [18:44] look there is only one case, which actually benefits from S_IMMUTABLE_LINK, where one inode is linked into two contexts ... [18:44] mhepp (~mhepp@r72s22p13.home.nbox.cz) joined #vserver. [18:45] this case can be described as (file in ctx0) and (n_links > 1) [18:45] assuming that you are in a context != ctx0, this file should behave like S_IMMUTABLE_LINK set on it ... [18:46] when the S_IMMUTABLE_FILE is set ... (maybe also without it) [18:47] clearly the without S_IMMUTABLE_FILE case would fit your S_SHARED or if you prefer, both would ... [18:50] or do you see any 'other' use for S_SHARED / S_IMMUTABLE_LINK? [18:50] shared files not count in per context diskquotas [18:51] well ctx0 files (those would be shared) don't count either, right? [18:51] and can be in other types quots.. [18:51] but you can do ctx5 files shared for all ? [18:52] what would be the use of that? beside that it would screw up the UID/GID? [18:56] not :) but can. and check one bit more fast. [18:56] well, I would prefer _not_ to modify 6 filesystems with a flag which isn't used 8-) [18:58] also please apply for you small fix for disquota statitics.. [18:59] mdaur (mdaur@p50917C44.dip.t-dialin.net) joined #vserver. [18:59] what fix is this? [18:59] hi [18:59] in sync_diskquota [18:59] dqput(dquot); [18:59] put_dquot_ref(dquot); [18:59] remove dqput(dquot); [18:59] and add put_dquot_ref(dquot); [18:59] it fix statitics after sync [19:00] do 'sync' - and you seen "Number of still active inodes with quota " be lower [19:01] hihmm, we are talking about mq0.11? [19:01] than before 'sync' [19:01] or about vanilla kernel code? [19:01] i talk about _all_ diskquota [19:01] vanila.. rh.. other patchs.. [19:01] okay, give me a second to verify ... [19:01] okey. [19:02] for veryfy do [19:02] 1) mount -o remount,usrquota,grpquota .... [19:02] 2) quotaon .... [19:02] 3) start some program [19:02] 4) do sync [19:03] 5) quotaoff [19:04] my test [19:04] [root@uml root]# mount -o remount,usrquota / [19:04] [root@uml root]# quotaon -augv / [19:04] start mc... \ finish mc [19:04] [root@uml root]# sync [19:04] [root@uml root]# quotaoff -aug / [19:04] [root@uml root]# quotastats [19:04] Number of dquot lookups: 113 [19:04] Number of dquot drops: 114 [19:04] Number of still active inodes with quota : -1 [19:05] also do mount -o remount / - do not stop using diskquota but quotatools say - quota not used on fs. [19:07] hmm, where do you see this code? (the sync_disquota labeled one?) [19:07] fs/dquot.c [19:08] you mean vfs_quota_sync? [19:08] for new - can be it. [19:08] yes. [19:08] get_dquot_ref(dquot); [19:08] if (dquot->dq_flags & DQ_LOCKED) [19:08]