1171929613 J * ensc ~irc-ensc@p54B4FCF9.dip.t-dialin.net 1171929701 Q * dna Quit: Verlassend 1171930081 Q * DreamerC_ Quit: leaving 1171930127 J * DreamerC ~dreamerc@125-225-99-212.dynamic.hinet.net 1171930285 Q * Johnnie Server closed connection 1171930477 Q * olivierk_ Ping timeout: 480 seconds 1171931571 Q * ard Server closed connection 1171931577 J * ard ~ard@82-197-200-127.dsl.cambrium.nl 1171931894 Q * ema Quit: leaving 1171935844 M * CHTEKK lo all 1171935880 M * CHTEKK is it normal that with such sched limits I can make cpuhog in that context use 100% of CPU? http://paste.linux-vserver.org/1182 1171935919 M * CHTEKK if I disable IDLE_TIME stuff, so that instead of the I i get -, it limits the context to 13% like it should 1171935975 M * CHTEKK so is it possible that the idle time, even if set to 1/8, would account for it being able to use all CPU? (there are no other contexts that are particularly active on that system) 1171936293 Q * FireEgl Quit: ... 1171936395 M * CHTEKK the last equations on http://linux-vserver.org/CPU_Scheduler miss a pharentesis 1171936412 M * CHTEKK its (fract1 + fract2) * 100 1171936481 M * CHTEKK but yeah, once that's fixed you actually get a result of 100% for my case 1171936510 M * CHTEKK 1 vserver limited to 1/8, so 7/8's idle, idle settings of 1/8 too, and he basically gets the whole idle time 1171937956 M * CHTEKK hmmm interesting kernel panic... let me see if I can reproduce it 1171938364 M * CHTEKK mhh can't reproduce it, scheudling works well now... was something inside vx_tokens_blabla or somt like that... and doing the same I did before doesn't trigger it again 1171939729 M * CHTEKK grml scheduling doesn't work as it should, and it managed to panic again... I'll have Hollow look at it tomorrow, I don't have the knowledge he has wrt debugging kernel stuff et all 1171939730 M * CHTEKK gn8 1171940644 Q * mountie Server closed connection 1171940650 J * mountie ~mountie@CPE0080c6fe323f-CM000a739acaa4.cpe.net.cable.rogers.com 1171941316 Q * infowolfe_ Read error: Connection reset by peer 1171941367 J * infowolfe ~infowolfe@c-67-164-195-129.hsd1.ut.comcast.net 1171941689 P * loceur 1171941814 Q * hardwire Quit: Coyote finally caught me 1171942315 J * infowolfe_ ~infowolfe@c-67-164-195-129.hsd1.ut.comcast.net 1171942544 Q * infowolfe Ping timeout: 480 seconds 1171942959 N * Bertl_oO Bertl 1171942976 M * Bertl back now ... 1171945724 M * lylix Bertl: the summation of idle-time ratios in the denominator of the fair-scheduling expression only includes those contexts which are being allocated CPU time, yes? 1171945784 M * Bertl sec, have to parse that first :) 1171945845 M * Bertl hmm, no, please try to rephrase ... 1171945876 M * lylix k, let me just ask straighforward 1171945890 M * lylix what values are used in the summation 1171945917 M * lylix http://linux-vserver.org/CPU_Scheduler 1171945991 M * Bertl ah, the mathematics Hollow did ... 1171946031 M * Bertl I'm somewhat certain they are wrong, although they look nice ... but I didn't find the time to think them trough 1171946043 J * FireEgl Atl-NA@68.220.222.136 1171946071 M * Bertl lylix: do you need them for anything or just trying to understand how it works? 1171946085 J * Johnnie ~jdlewis@jdlewis.org 1171946096 M * lylix just trying to understand more thoroughly 1171946127 M * lylix just not sure what ratios get included in the summation 1171946138 M * lylix ALL guest idle settings 1171946153 M * lylix or just the ones grabbing idle time 1171946158 M * lylix but now after looking at it 1171946169 M * lylix the example says 5 contexts 1171946189 M * lylix and there are only 2 idle-time ratios for the two "active" contexts 1171946202 M * lylix so i guess its only those who need idle-time CPU 1171946366 M * Bertl just forget about the summations for a moment 1171946382 M * Bertl the mechanism is quite easy to explain ... i.e. let me try :) 1171946400 M * Bertl you already know how the token buckets work, yes? 1171946467 M * lylix yep 1171946487 M * Bertl okay, so basically the system knows 3 states 1171946503 M * Bertl 1) idle, nothing is running, nothings wants to run 1171946537 M * Bertl 2) at least one context has tokens left, and is executing 1171946577 M * Bertl 3) all contexts are out of tokens, although processes would execute if the could 1171946584 M * Bertl *they 1171946631 M * Bertl now with the hard scheduler set, the state 3) will result in forced idle, until the first context gets enough tokens (min fill) to be scheduled again 1171946675 M * Bertl the idle time skip (fair scheduling) is what will kick in instead, if compiled in and selected for at least one context 1171946697 M * Bertl and it will simply do the following:# 1171946708 M * lylix which is simply by setting values and no additional flags... k 1171946735 M * lylix * guess it depends on the userlands apps * 1171946737 M * Bertl a) switch all idle time enabled contexts to R2/I2 (instead of R/I) 1171946765 M * Bertl b) artificially advance the time until the first context would be scheduled again (in an instant) 1171946788 M * Bertl c) switch back to R/I and continue scheduling 1171946848 M * Bertl so, let's look at the calculations now, and start with something simple (i.e. a few examples) yes? 1171946896 M * lylix sure (mainly interested in how a single context can get 100%) 1171946920 M * Bertl first, without idle time ... 1171946946 M * Bertl and we also assume min = 0 1171946967 M * Bertl we have now two reulative values, R1 (rate) and I1 (interval) 1171946973 M * Bertl *regulative 1171947001 M * Bertl every I1 msecs/jiffies we will add R1 tokens to the bucket 1171947033 M * Bertl so after T1 msecs/jiffies we will get 1171947058 M * Bertl ceil(T1/I1)*R1 tokens into the bucket 1171947077 M * Bertl hmm, no, actually, it's floor(T1/I1) 1171947108 M * Bertl so, let's now put up an equation, starting with S1 tokens 1171947129 M * Bertl to calculate how long a context running at 100% cpu would be running 1171947204 M * Bertl obviously this time T1 will consist of 1 token per time unit removed from the bucket (S1) plus the amount of tokens added to the bucket within that time 1171947213 M * Bertl so that looks like this then: 1171947241 M * Bertl floor(T1/I1)*R1+S1 = T1 1171947280 M * Bertl if we ignore the rounding down to simplify that equation (and actually solve it) we get to: 1171947347 M * Bertl T1*(R1/I1-1) = -S1 1171947369 M * Bertl and after negating/dividing by the fraction, we have 1171947371 J * muuhDBXi ~foo@a213-22-7-216.cpe.netcabo.pt 1171947395 M * Bertl T1 = S1/(1-R1/I1) 1171947421 M * Bertl lets see if I got that right now with a few intuitive cornercases 1171947451 M * Bertl R1/I1 = 1 means we add one token per unit, which is exactly what is consumed 1171947469 M * Bertl so we will keep the context running for an infinite amount of time 1171947472 M * lylix yep 1171947494 M * Bertl and indeed S1/(1-1) converges towards infinity 1171947517 M * Bertl now the other corner case is R1 = 0 1171947527 M * Bertl i.e. we do not add any tokens 1171947550 M * Bertl and again that give T1 = S1 (the expected time) 1171947637 M * Bertl note: this calculation is only true for the above assumptions, i.e. that min = 0 and 100% cpu is used 1171947670 M * Bertl now let's look at the percentage over a timeframe 1171947715 M * Bertl we can assume that once we reach a 'stable' state, there will be cycles between min fill and zero 1171947753 M * Bertl so introducing the new M1 (minimum fill), we get cycles of length 1171947766 M * Bertl M1/(1-R1/I1) 1171947798 M * lylix which are held processes w/ fair scheduling 1171947802 M * lylix w/o 1171947812 M * Bertl yep, no fair scheduling yet 1171947845 M * Bertl and after each of those periods, we have to wait until M1 is reached again 1171947908 M * Bertl this can be calculated as floor(Z1/I1)*R1 >= M1 1171947921 M * Bertl again, simlifying the equation will give us 1171947952 M * Bertl Z1 = M1*(I1/R1) 1171948000 M * Bertl so we end up of cycles of length T1 + Z1 total, with T1 tokens (or jiffies/msecs for the context) 1171948011 M * Bertl which gives us a percentage of: 1171948025 M * Bertl T1/(T1+Z1) for the cpu usage 1171948040 M * Bertl resubstituing the values now gives: 1171948108 M * Bertl [M1/(1-R1/I1)]/[M1/(1-R1/I1) + M1*(I1/R1)] 1171948118 J * quiksilv Linden@203.176.96.250 1171948372 M * Bertl so after cleaning that up, we can eliminate everything 1171948379 M * Bertl until we end up with 1171948396 M * Bertl I1/R1 1171948412 M * Bertl oops R1/I1 1171948436 M * Bertl welcome quiksilv! 1171948447 M * Bertl lylix: could you follow so far? 1171948461 M * lylix yep, n/p 1171948468 M * quiksilv hi there 1171948501 M * Bertl lylix: btw, I did the elimination right now, so it actually gives the desired result 1171948515 M * Bertl (just to verify that I'm not talking rubbish here :) 1171948522 M * lylix :) 1171948539 M * Bertl okay, now for the fair scheduling part 1171948560 M * Bertl instead of doing the 'refill' cycle (Z1) 1171948579 M * Bertl we simply skip some idle time 1171948597 M * Bertl Z2 = M1*(I2/R2) 1171948617 M * Bertl within a single jiffie 1171948634 M * Bertl and continue from there with the T1 cycle 1171948659 M * Bertl what does that mean now for a 100% cpu hog? 1171948669 Q * infowolfe_ Quit: Leaving 1171948675 M * Bertl it will get 100% cpu and run all the time 1171948690 M * quiksilv can i ask a vserver support question ? 1171948708 M * Bertl quiksilv: sure, that's what this channel is for (among other things) 1171948788 M * quiksilv ok thanks, ive got a vserver enabled kernel up and running on Redhat EL4 but now I am trying to build a guest... 1171948809 M * Bertl what tools did you install? 1171948818 M * quiksilv util-vserver from source 1171948832 M * lylix k, makes sense... so essentially when we hit 0, the next jiffie puts us up to M1, which then satisfies the T1 cycle, reducing tokens to 0 again, etc, etc.... got it 1171948832 M * Bertl 0.30.212 or later? 1171948875 M * quiksilv yea 0.30.212 1171948892 M * quiksilv do i also need apt-get on my redhat distro or any other tools ? 1171948904 M * quiksilv i noticed i needed to create /vservers and /vservers/.pkg dir 1171948912 Q * Greek0 Server closed connection 1171948925 M * Bertl no, those dirs will be created on the make distribution-install 1171948927 J * Greek0 ~greek0@85.255.145.201 1171948948 M * quiksilv hmm i only did a 'make install' 1171948973 M * Bertl so you want to do that too then, as sugested at the end of the install process :) 1171949003 M * Bertl also make sure that you do use dietlibc (you have been warned) and you have only one copy of the tools installed :) 1171949043 M * Bertl http://linux-vserver.org/Building_Guest_Systems 1171949070 M * Bertl this is only one example, there are many others, sec 1171949080 M * quiksilv excellent, i did miss make install_distribution 1171949102 M * Bertl http://oldwiki.linux-vserver.org/VServer+installation+Fedora+Core+5#g4 1171949129 M * Bertl basically you can use one of the build methods 1171949147 M * Bertl rpm, yum, apt-get, debootstrap, template, archive ... 1171949150 M * quiksilv been reading that but was stuck on the earlier process i think :) 1171949169 M * Bertl the easiest test is a debian guest with debootstrap 1171949177 M * quiksilv leave it with me for a few mins ill let you know how i go.... my fault for not reading the rest of 'make install' 1171949182 M * Bertl as it doesn't require any installed support tools 1171949191 M * Bertl okay, np 1171949229 M * Bertl lylix: okay, so the fair scheduling does the expected in this special case (i.e. nothing else wants to run) 1171949282 M * Bertl now let's looks at a rather complex case, and simplify away the details by referring to existing formula 1171949288 J * DoberMann_ ~james@AToulouse-156-1-119-97.w90-30.abo.wanadoo.fr 1171949346 M * Bertl let's assume we have two guests with the followin ratios in a stable state, with two hogs competing for a single cpu 1171949381 M * Bertl R1a/I1a = 0.2 R1b/I1b = 0.5 1171949394 Q * DoberMann[ZZZzzz] Ping timeout: 480 seconds 1171949400 M * Bertl R2a/I2a = 0.1 R2b/I2b = 0.5 1171949402 J * Johnsie ~jdlewis@h130.158.88.75.ip.alltel.net 1171949447 M * Bertl this means we start with M1a and M1b for both guest 1171949497 M * Bertl then they will run a time T1, which consists of the sum of both guest times 1171949524 M * Bertl but, the problem here is that we cannot simply add them up, as the total time will affect the accumulated tokens 1171949555 M * Bertl so we have a time T1 = T1a+T1b with the following formular 1171949633 M * Bertl T1*(R1a/I1a) = X1a tokens for a, plus the time from M1a 1171949652 M * Bertl T1*(R1b/I1b) = X1b tokens for b, plus the time from M1b 1171949699 M * Bertl gives T1*(R1a/I1a)+M1a = T1a, and T1*(R1b/I1b)+M1b = T1b 1171949709 M * Bertl replacing that gives 1171949756 M * Bertl T1*(R1a/I1a + R1b/I1b) + M1a + M1b = T1 1171949797 M * Bertl solving that for the interval T1 gives us? 1171949864 M * Bertl T1 = (M1a+M1b)/(1-(R1a/I1a + R1b/I1b)) 1171949903 M * Bertl after that period we will reach a point where no tokens are left 1171949930 M * Bertl and if enabled, the idle skip will kick in 1171949938 Q * Johnsie Quit: used jmIrc 1171949946 M * Bertl lylix: still with me? 1171949952 M * lylix yep 1171949966 M * lylix refilled coffee, so far so good 1171949976 M * Bertl okay, so here is the whole thing getting somewhat tricky 1171950006 M * quiksilv hmm ok trying a debian bootstrap.. 1171950013 M * quiksilv "E: Invalid Release file, no entry for main/binary-x86_64/Packages" 1171950029 M * Bertl ah, x86_64 :) 1171950036 M * quiksilv hehe yea... 1171950052 M * Bertl you either want to specify a repository for that, or build a 32bit guest for testing 1171950062 M * quiksilv how do i force a 32bit geust 1171950064 M * quiksilv guest 1171950079 M * Bertl IIRC, there is an --arch option to debootstrap 1171950093 M * Bertl specify that after a double dash -- at the end of the command 1171950098 M * quiksilv ideally, i would actually like a guest the same as the host (Redhat enterprise...) not fedora core or anything... 1171950122 M * Bertl what package dependancy resolver does RHEL use? 1171950127 M * Bertl yum? 1171950178 M * Bertl apt-rpm? 1171950260 M * quiksilv none unfortunately... this is my gripe 1171950273 M * quiksilv it has "up2date" which uses rpm 1171950290 M * quiksilv but doesnt deal with dependencies afaik 1171950306 M * Bertl hmm, okay, so how do you install RHEL into a directory? 1171950328 M * Bertl do you have a list of all rpms you need? 1171950349 M * quiksilv no but i could probably do a minimal install off a CD and see what rpm's it installs 1171950375 M * Bertl okay, if you do that, you can copy the result, clean it up a little and use it as guest (or template) 1171950380 M * quiksilv or i could just build a debian guest and be done with it :) 1171950391 M * Bertl which would probably be easier :) 1171950415 M * quiksilv ok next question... once i have the file system/directory structure... how do i create the /etc/vserver/ config files ? 1171950418 Q * pflanze Quit: [x]chat 1171950440 M * Bertl simple, you create it with the build command 1171950452 M * quiksilv using -t or something ? 1171950454 M * Bertl if you have all files somewhere, use the skeleton build method 1171950462 M * quiksilv ah skeleton build riteo 1171950476 M * Bertl (which will leave you with an empty dir where you can move the stuff) 1171950477 Q * mEDI_S Quit: mEDI_S 1171950491 M * quiksilv gotcha 1171950492 M * Bertl make sure not to copy the /dev though, as it might contain insecure device nodes 1171950523 M * quiksilv i only ask because i have a vserver guest image, and i cant get it to build with -t but ive extracted the contents so just need to build the conf files 1171950527 M * Bertl if you make a tar, you would use the template build method 1171950555 M * quiksilv can template method take a bz2 image ? i get application/x-bzip2 unsupported or something 1171950577 M * Bertl that is a (somewhat known?) bug with file 1171950589 M * quiksilv ok ill bunzip it first 1171950599 M * Bertl i.e. the older file tools do not recognize the contents properly 1171950606 M * Bertl and/or cannot unpack it properly 1171950622 M * Bertl (you might also check the 0.30.213-rc* releases) 1171950637 M * quiksilv ah i might do thanks...rc's always scare me :) 1171950641 M * Bertl make sure that your build method contains at least a 1171950653 M * Bertl --context and some --ip .... 1171950670 M * Bertl (otherwise you might get into trouble later ...) 1171950679 M * Bertl see vserver - build --help for details 1171950701 M * quiksilv hmm ive seen context around....what does it actually do/mean ? 1171950715 M * Bertl it is an unique identifier for your guest 1171950728 M * Bertl it will be used to identify the context the processes will belong to 1171950751 M * quiksilv gotcha 1171950777 M * quiksilv well, with that great info i will go an give it all a whirl 1171950783 M * Bertl lylix: okay, we will continue on that tomorrow, I'm quite tired and have to get up early today ... 1171950785 M * quiksilv thanks Bertl, great help ! 1171950789 M * Bertl you're welcome! 1171950802 M * Bertl off to bed now ... have a good one everyone! cya! 1171950807 M * quiksilv cya ! 1171950809 N * Bertl Bertl_zZ 1171950816 M * lylix n/p... getting to the interesting stuff though, so def pickup later 1171950819 M * lylix g/n 1171950829 M * lylix tnx for the thorough explaination 1171951265 Q * ard Ping timeout: 480 seconds 1171951278 J * ard ~ard@82-197-200-127.dsl.cambrium.nl 1171952188 Q * tokkee Server closed connection 1171952195 J * tokkee tokkee@casella.verplant.org 1171952323 J * Aiken ~james@ppp126-23.lns2.bne4.internode.on.net 1171953457 N * DoberMann_ DoberMann 1171955477 J * hardwire ~hardwire@rdbck-4676.wasilla.mtaonline.net 1171955596 M * daniel_hozac quiksilv: to build RHEL4 guests, use -d centos4. 1171955662 M * daniel_hozac (obviously won't be actual RHEL guests unless you change the configuration, but essentially it's the same) 1171955844 J * micah_ ~micah@micah.riseup.net 1171955956 Q * micah Ping timeout: 480 seconds 1171956440 N * DoberMann DoberMann[PullA] 1171956537 J * gab ~gab@158.36.45.236 1171956732 J * kir ~kir@swsoft-mipt-nat.sw.ru 1171959015 N * micah_ micah 1171959130 J * ema ~ema@lart.galliera.it 1171959751 N * DoberMann[PullA] DoberMann 1171959852 J * dna ~naucki@137-221-dsl.kielnet.net 1171960616 Q * tamitall Ping timeout: 480 seconds 1171961966 J * meandtheshell ~markus@85-124-175-19.dynamic.xdsl-line.inode.at 1171962416 J * dlezcano ~dlezcano@blueice2n1.uk.ibm.com 1171963413 Q * shedi Remote host closed the connection 1171964401 Q * baldy Server closed connection 1171965007 P * muuhDBXi 1171965905 J * baldy baldy@baldy.biz 1171965950 Q * Aiken Read error: Connection reset by peer 1171965962 J * Aiken ~james@ppp126-23.lns2.bne4.internode.on.net 1171968655 Q * Aiken Quit: Leaving 1171969404 N * DoberMann DoberMann[PullA] 1171969594 M * matti :) 1171971231 J * mEDI_S ~medi@snipah.com 1171971887 Q * derjohn Ping timeout: 480 seconds 1171971929 J * derjohn ~derjohn@dslb-084-059-027-208.pools.arcor-ip.net 1171976790 J * Aldo ~sebastian@pD957FC7D.dip.t-dialin.net 1171976795 M * Aldo HI 1171977001 M * daniel_hozac hello 1171977007 Q * ema Quit: leaving 1171977066 M * Aldo Can you help me with "Disk Limits" ? 1171977086 M * daniel_hozac what about them? 1171977117 M * Aldo Very easy: How does it work and how can I aktivate this? 1171977235 M * daniel_hozac http://oldwiki.linux-vserver.org/Disk+Limits 1171977273 M * Aldo One Problem ... my Host does not knows tagxid 1171977297 M * daniel_hozac is it running a Linux-VServer kernel? 1171977323 M * Aldo yes 1171977331 M * daniel_hozac which one? 1171977344 M * daniel_hozac and what makes you think it doesn't know it? 1171978523 Q * derjohn Ping timeout: 480 seconds 1171978524 J * derjohn ~derjohn@80.69.41.2 1171978756 M * Aldo when I reboot the System it says: Unknown tagxid 1171978794 M * daniel_hozac at what point? which filesystem are you trying to mount with tagxid? 1171978841 M * Aldo ext3 the root file System (/) 1171979098 M * Aldo any idea? 1171979362 M * daniel_hozac well, tagxid is not allowed on remount. 1171979395 M * daniel_hozac you really should put your guests on another filesystem, as adding tagxid to the root filesystem requires booting with rootflags=tagxid, as per the note at the bottom of that page. 1171979456 M * daniel_hozac (which really is not recommended, as it might cause problems) 1171979565 N * Bertl_zZ Bertl 1171979574 M * daniel_hozac morning Bertl! 1171979583 M * Bertl good morning folks! 1171979759 M * Aldo morning? 1171979786 M * Aldo Afternoon ^^ 1171980013 M * Bertl Aldo: just got up ... now heading for breakfast ... 1171980020 N * Bertl Bertl_oO 1171980024 M * Aldo Oo^^ 1171980038 M * Aldo Where come you from? 1171980044 M * Aldo *came 1171980059 M * Bertl_oO Aldo: Austria ... :) 1171980066 M * Aldo oh ^^ 1171980089 M * Aldo here is 15:01 o'clock ^^ (Germany) 1171980342 J * tamitall ~tam@gw.nettam.com 1171980732 M * mjt # vserver guest0 build ... 1171980733 M * mjt vserver-topdirectory '/etc/vservers/.defaults/vdirbase/guest0/etc' and/or configuration at '/etc/vservers/guest0' 1171980736 M * mjt exist already; please try to use '--force', or remove them manually. 1171980740 M * mjt # vserver guest0 build --force ... 1171980744 M * mjt vserver-build: unrecognized option `--force' 1171980781 M * mjt every time i try vserver, it looks so much umm... raw, so to say. 1171980806 M * mjt every script has one or another error 1171980831 Q * mnemoc Ping timeout: 480 seconds 1171980898 M * mjt i tried to collect/submit patches at least twice. But gave up due to just too many errors. 1171981098 M * daniel_hozac uh, vserver guest0 build --force should be fine. 1171981129 M * daniel_hozac unless your getopt is broken. 1171981175 M * mjt well it should, but it doens't ;) 1171981218 M * daniel_hozac using --force doesn't make much sense anyhow, as it leaves the old build around. 1171981232 M * mjt it's in fact even more... funny 1171981282 M * mjt i used newvserver initially (from vserver-debiantools if i recall correctly). It creates initial .deb apt "repository" in $vserver_root/$newserver/var/cache/apt/ 1171981294 M * mjt so that packages will not be downloaded again 1171981312 M * mjt but vserver-build complains about that directory being here 1171981332 M * mjt and doesn't understand --force, too 1171981372 M * mjt more - there's no stuff in the paths mentioned (in /etc/vservers) 1171981407 M * daniel_hozac the error says there's either stuff in /etc/vservers/ or /vservers/. 1171981450 M * mjt it's /var/lib/vservers not /vservers here... but doesn't matter. 1171981483 M * daniel_hozac well, Debian is odd. 1171981511 M * mjt it just keeps everything in places where it should be 1171981521 M * mjt at least by default 1171981549 M * daniel_hozac http://paste.linux-vserver.org/1185 works fine for me. 1171981633 M * daniel_hozac so most likely your getopt is just confused. 1171981678 M * mjt i wonder why 1171981787 M * mjt getopt -o +sv --long nonamespace,--nonamespace,--insecure,defaulttty,help,debug,version,sync,verbose,silent ... 1171981805 M * mjt there's no `force' option 1171981822 M * mjt and just getopt says so 1171981858 M * mjt s/just getopt/getopt just/ 1171981945 M * mjt wug. 1171981968 J * mnemoc ~amery@kilo105.server4you.de 1171982037 N * Bertl_oO Bertl 1171982043 M * Bertl morning mnemoc! 1171982058 M * Bertl daniel_hozac: I'm more for 'Debian is even' :) 1171982077 M * daniel_hozac haha 1171982102 M * daniel_hozac mjt: that's the wrong getopt invocation you're looking at. 1171982108 A * Bertl bertl recently read about malapropisms ... after hearing the old Beatles 'Tommorow Never Knows' again ... 1171982139 M * mjt daniel_hozac: i'm looking at bash -x output - the last line after which it fails. So it is the right one. 1171982189 M * daniel_hozac mjt: then your getopt _is_ broken. 1171982206 M * daniel_hozac the + should make it stop processing as soon as it finds a non-argument. 1171982215 M * daniel_hozac obviously, yours just keeps on going. 1171982230 M * daniel_hozac (or you just added the option in the wrong place) 1171982290 M * mjt ok. it was the wrong place. 1171982306 M * mjt which isn't quite.. obvious, isn't it? :) 1171982326 M * Bertl mjt: nothing is obvious .. until you know it :) 1171982344 M * mjt now it successefully renamed that dir and created new one. How very helpful - d/loading all the packages again. 1171982351 M * mjt ugh! 1171982363 J * prae ~Benjamin@host.187.57.23.62.rev.coltfrance.com 1171982368 M * prae hi everybody 1171982372 M * daniel_hozac hello 1171982373 M * Bertl mjt: no debootstrap cache? 1171982399 M * mjt and ran out of disk space. because i symlinked that dir to other place, but it tries to create it in root filesystem. 1171982423 M * mjt Bertl: i do have it. but vserver build happily removes the cache. 1171982433 M * mjt or, rather, starts from scratch 1171982455 M * prae I have a small problem with latest patch 1171982465 M * daniel_hozac what does "latest" mean? 1171982472 M * Bertl mjt: really? quite unusual, as it just wraps the debootstrap IIRC 1171982486 M * mjt so how to stop it from renaming that directory which already has all the packages? 1171982489 M * prae "vs2.3.0.10" 1171982500 M * prae 2.6.20-vs2.3.0.10 1171982530 M * daniel_hozac prae: and, what's the problem? 1171982581 M * prae I try to start my vs 1171982583 M * prae # /usr/local/util-vserver-0.30.212/sbin/vserver unittest start 1171982600 M * prae secure-mount: mount(): No such file or directory 1171982600 M * prae /usr/local/util-vserver-0.30.212/lib/util-vserver/vserver.functions: line 682: pushd: /etc/vservers/unittest/vdir/: No such file or directory 1171982605 M * prae Failed to start vserver 'unittest' 1171982621 M * prae # ls -lahd /etc/vservers/unittest/vdir/ 1171982622 M * prae drwxr-xr-x 21 root root 4,0K 2006-06-19 17:47 /etc/vservers/unittest/vdir/ 1171982689 M * daniel_hozac and what does /etc/vservers/unittest/fstab contain? 1171982702 M * prae # cat /etc/vservers/unittest/fstab 1171982703 M * prae none /proc proc defaults 0 0 1171982703 M * prae none /tmp tmpfs size=16m,mode=1777 0 0 1171982703 M * prae none /dev/pts devpts gid=5,mode=620 0 0 1171982705 M * prae --eof 1171982735 M * prae (in previous kernel, all of my vs works perfectly) 1171982746 M * daniel_hozac what does a --debug run show? 1171982771 Q * dlezcano Read error: Connection reset by peer 1171982780 M * prae daniel_hozac: result at pastebin ? 1171982785 M * daniel_hozac sure. 1171982870 M * prae http://paste.linux-vserver.org/1187 1171982887 M * Bertl prae: ls -la /etc/vservers/unittest/vdir 1171982900 M * prae # ls -la /etc/vservers/unittest/vdir 1171982900 M * prae lrwxrwxrwx 1 root root 41 Jul 20 2006 /etc/vservers/unittest/vdir -> /etc/vservers/.defaults/vdirbase/unittest 1171982909 M * prae # ls -la /etc/vservers/.defaults/vdirbase/unittest 1171982909 M * prae lrwxrwxrwx 1 root root 22 Jan 5 13:22 /etc/vservers/.defaults/vdirbase/unittest -> /opt/vservers/unittest 1171982914 M * Bertl and ls -la /etc/vservers/.defaults/vdirbase 1171982957 M * prae # ls -la /etc/vservers/.defaults/vdirbase 1171982957 M * prae lrwxrwxrwx 1 root root 17 Jan 5 12:03 /etc/vservers/.defaults/vdirbase -> /var/lib/vservers 1171983117 M * prae any idea ? 1171983135 M * Bertl and doing cd /etc/vservers/unittest/vdir works just fine, I guess? 1171983151 M * Bertl I assume the tools unmount part of your filesystem 1171983157 M * prae # cd /etc/vservers/unittest/vdir 1171983157 M * daniel_hozac yep. 1171983157 M * prae thot:/etc/vservers/unittest/vdir (dev) # ls 1171983157 M * prae bin boot dev etc home initrd lib media mnt opt proc root sbin srv sys tmp usr var 1171983160 M * Bertl probably the /opt partition, daniel_hozac? 1171983164 M * daniel_hozac nah, it's /var. 1171983179 M * daniel_hozac that's what you get for having symlinks pointing to symlinks ;) 1171983207 M * Bertl hey, at least linux has working symlinks :) 1171983239 M * prae it's strange, see : 1171983241 M * prae ++ /usr/local/util-vserver-0.30.212/lib/util-vserver/secure-mount --rbind -n /etc/vservers/unittest/vdir / 1171983242 M * prae secure-mount: mount(): No such file or directory 1171983247 M * prae line 1600 1171983248 M * prae but 1171983270 M * daniel_hozac prae: just change /etc/vservers/unittest/vdir to point to the final destination, i.e. /opt/vservers/unittest 1171983272 M * prae I do this myself: # /usr/local/util-vserver-0.30.212/lib/util-vserver/secure-mount --rbind -n /etc/vservers/unittest/vdir / 1171983326 M * daniel_hozac prae: your namespace still has the /var mount. 1171983342 M * mjt ok, this is just brilliant. In order to stop it from d/loading the damn packages again, i have to run vserver build, catch it after it created the directories but before it starts debootstrap, suspend it, and copy the package files into var/cache/apt/archives/. And after that, resume vserver build. 1171983348 M * daniel_hozac prae: do vnamespace -n, and then umount -l -n /var, and then try it. 1171983355 M * prae arg 1171983356 M * prae ok 1171983359 M * prae /usr/local/util-vserver-0.30.212/sbin/vserver unittest start 1171983359 M * prae ncontext: vc_net_create(): Invalid argument 1171983361 M * prae so ... 1171983366 M * mjt for that, i added a `sleep 10' line at proper place in vserver script. 1171983370 M * daniel_hozac mjt: or, you know, implement that feature in the debootstrap build method. 1171983370 M * harry don't use dynamic contexts 1171983405 M * Bertl mjt: yeah, guess a patch doing that 'properly' will be quite welcome! 1171983420 M * prae hmmm ... I don't remember who I set the CTX id for vs 1171983425 A * prae search 1171983426 M * mjt daniel_hozac: the only 'implementation' needed is to remove that idiotic directory renaming in vserver script 1171983466 M * daniel_hozac mjt: so you'd prefer it to just overwrite any guest that you might have in the directory already? 1171983469 M * daniel_hozac yeah, that seems like a great idea! 1171983487 M * mjt ..which i just did, and no more jumping around it. 1171983491 M * daniel_hozac prae: echo > /etc/vservers//context 1171983498 M * prae yep, I found :) 1171983500 M * mjt if i said --force, i really meant it 1171983603 M * daniel_hozac mjt: would be better to add an option that would let you point it at your deb repo, no? 1171983612 M * daniel_hozac mjt: and then have the build method bind mount it into place. 1171983622 M * prae groumpf 1171983630 M * prae "save_ctxinfo: open("/var/run/vservers/unittest"): No such file or directory" 1171983631 M * prae *arg* 1171983635 M * daniel_hozac prae: doh. 1171983650 M * daniel_hozac prae: add /var to /etc/vservers/.defaults/namespace-cleanup-skip 1171983653 M * mjt well, for the start, would be better to actually treat --force as force, not as --rename 1171983671 M * prae ouaahahh 1171983673 M * prae daniel ! 1171983678 M * prae congrats' :) 1171983694 M * prae it's works :) 1171983809 M * prae thx daniel_hozac & Bertl 1171983872 J * Piet hiddenserv@tor.noreply.org 1171983893 M * daniel_hozac prae: np, i'll make sure we exclude more paths in 0.30.213... 1171983907 M * prae :) 1171983934 J * chand ~chand@m167.net81-64-156.noos.fr 1171983989 M * Aldo Any body here? 1171983999 M * daniel_hozac more or less. 1171984033 M * prae body not, virtual yes 1171984133 M * Bertl Aldo: I'm off to take a shower ... back later though ... 1171984150 M * Aldo I have a big Problem 1171984168 M * Aldo an amusingly problem ... 1171984178 M * Bertl well, amuse us then :) 1171984190 M * Aldo I had make a loop device for the vServer's 1171984209 M * Bertl you need a capability for that to work inside a guest 1171984217 M * Aldo what? 1171984242 M * Bertl and depending on the patch version, you want to set it up in guest context 1171984255 M * Bertl otherwise it will not be seen inside the guest :) 1171984259 M * Aldo Now to the error: http://www.timsta.net/ 1171984275 M * Bertl that is a feature (security) :) 1171984278 M * Aldo That error comes alway, ftp, aso 1171984291 M * Bertl okay, off to shower now ... I'll explain afterwards ... 1171984320 M * yang heya Bertl 1171984439 M * mjt hmm. processes running in guests aren't visible on host, right? 1171984456 M * daniel_hozac nope. 1171984937 Q * baldy Ping timeout: 480 seconds 1171985166 M * Aldo shice when I copied the files the owner / group is now root on all files :( 1171985477 M * mjt cp -p is your friend 1171986759 J * dlezcano ~dlezcano@blueice1n1.uk.ibm.com 1171986973 M * Bertl wb dlezcano! 1171986983 M * dlezcano hi 1171986988 M * Bertl Aldo: okay, what Linux-VServer version (kernel) do you use? 1171987015 M * Aldo The problem is solved 1171987020 M * Bertl excellent! 1171987045 M * Aldo The Ower / group was wrong ... 1171987051 M * Aldo on all files root:root 1171987132 M * Aldo Now can I change the size of /tmp ? 1171987149 M * Bertl sure, just ccheck the fstab 1171987160 M * Bertl (in /etc/vservers//fstab) 1171987194 M * Aldo I mean How not Now sorry :) 1171987206 M * Bertl well, answer is the same :) 1171987224 M * Bertl should contain an entry with size=16M or so 1171987229 M * Aldo yes 1171987241 M * Bertl well, adjust it according to your likings 1171987271 M * Aldo That I like on linux-vserver.org. It's easy and open source :) 1171987299 M * Bertl that's what we like too :) 1171987563 M * Bertl Aldo: but thanks for the positive feedback! 1171987579 M * Aldo np it is true :) 1171987634 M * derjohn Hello! http://vserver.13thfloor.at/Experimental/patch-2.6.20-vs2.2.0-pre4.diff ... what does pre4 mean? what is 2.2.0-rc13 ? Is rc13 2.6.20-ready? 1171987644 M * Bertl it is in the human nature to complain ... 1171987658 M * derjohn really? :) 1171987664 M * Bertl derjohn: simple, 2.2.0 is not well tested on 2.6.20 1171987685 M * Bertl thus, we didn't label it rc13.1 (as for 2.6.19) but pre4 :) 1171987695 M * derjohn so, pre4 means ... a pre-release before even rc1 is out ? 1171987771 M * Bertl a prerelease for testing, not more and not less 1171987792 M * Bertl if we get positive feedback for that version, it might even become a release 1171987793 M * derjohn well, I think i'll try that out on my own machine and let customer machines stay with 2.6.18.5 ... 1171987820 M * derjohn yup. I'll compile now. to pre4 is the "newest" thing for 2.6.20, rightß 1171987821 M * derjohn ? 1171987863 M * Bertl IIRC, yes .. might still contain the missing include for jfs though 1171987876 M * Bertl (see mailing list) 1171988164 Q * Aldo Read error: Operation timed out 1171988180 M * derjohn daniel_hozac, told me there is a ipv6 patch out for 2.2.0 .. on his personal page? or elsewhere ? 1171988248 J * muuhDBXi ~foo@a213-22-7-216.cpe.netcabo.pt 1171988349 M * Bertl derjohn: no idea ... daniel might know 1171988383 M * derjohn ok, i'll have a look 1171988683 J * stefani ~stefani@flute.radonc.washington.edu 1171989000 P * muuhDBXi 1171989046 Q * mnemoc Ping timeout: 480 seconds 1171989215 J * mnemoc ~amery@kilo105.server4you.de 1171989428 N * DoberMann[PullA] DoberMann 1171989827 M * Bertl hey stefani! 1171989836 M * stefani salut. 1171990062 M * cehteh so now i can generate a crashdump for the stucked processes ... what to do exactly? 1171990093 M * Bertl well, check _where_ it is stuck, with e.g. gdb 1171990120 M * cehteh daniel_hozac saied something about writing to /proc/sysreq* 1171990211 M * cehteh less ./kdump/kdump.txt 1171990211 M * cehteh .. checking 1171990287 M * cehteh mhm 1171990624 M * Bertl daniel_hozac: any reason why the tools (util-vserver) are in the patches (p) path on people.*? I always start to search for them again :) 1171991106 J * muuhDBXi ~foo@a213-22-7-229.cpe.netcabo.pt 1171991172 M * Hollow Bertl: we figured out the sched problem meanwhile.. idle time was set, so the only context that was running the cpuhog of course got 100% 1171991202 M * Bertl sounds reasonable ... I thought about that yesterday, when lylix and I went through the scheduler mathematics 1171991230 M * Bertl btw, you might want to go through the irc log and copy/paste (~or at least double check with the wiki) 1171991238 M * Hollow but i have another problem now.. it try to set different buckets per cpu.. and it somehow does not work.. it always updates all buckets.. tested on quad system, and checked the mask.. it contains VXSM_CPU_ID .. *shrrug* 1171991250 M * Hollow ah ok.. will do 1171991260 M * Hollow i just fixed the fair scheduling formula.. was missing some brackets 1171991261 M * Bertl you need to specify the cpu/bucket 1171991275 M * Hollow yeah, it is specified, together with VXSM_CPU_ID 1171991288 M * Bertl what kernel/patch? 1171991302 M * Hollow 2.6.19-vs2.2.0-rc12-gentoo 1171991322 J * bonbons ~bonbons@83.222.38.57 1171991337 M * Bertl okay, please make sure to update to rc13.1 (if not done already), sec checking for interface now 1171991422 J * baldy baldy@baldy.biz 1171991446 M * Bertl Hollow: what cpu_id do you specify? 1171991478 M * Hollow i tried 1 and 2 1171991479 M * Bertl ah, no, looks like a bug :) 1171991490 M * Bertl you ahve force enabled, yes? 1171991517 M * Bertl VXSM_FORCE that is 1171991543 M * Hollow yep 1171991554 M * Bertl that currently updates all cpus to the new value 1171991559 M * Hollow ah, i see 1171991568 M * Bertl have to think about proper behaviour there ... 1171991585 M * Bertl will do a patch for that today 1171991600 M * Hollow ok, thanks :) 1171991660 M * Bertl thanks for spotting that! 1171991911 M * Hollow Bertl: the scheduler formulas you gave are way to confusing for my math-hating brain ;) 1171991953 M * Hollow *shrug* imo the formulas are correct, they are just somewhat simplified 1171991968 J * jmcaricand ~jmcarican@d90-144-56-52.cust.tele2.fr 1171991975 M * Hollow at least they seem to produce reasonable results :) 1171991993 M * Bertl which can be seen as an advantage :) 1171992080 M * Hollow well, basically the idle time is distributed among context that would run if they had tokens according to R2/I2 of all context that would run 1171992083 M * Hollow right? 1171992230 M * Bertl the idle time is distributed among the context being on hold and having idle time enabled according to the ratios i.e. R2a/I2a : R2b/I2b : R2c/I2c ... 1171992249 M * Hollow yep 1171992305 M * Hollow should be correct in my eyes, but would be really nice if you could verify that 1171992333 M * Bertl http://paste.linux-vserver.org/1188 (unrelated) 1171992509 M * daniel_hozac derjohn: yes, http://people.linux-vserver.org/~dhozac/p/k/patch-2.6.19.2-vs2.2.0-rc9.ipv6.diff is the most recent one. 1171992542 M * derjohn daniel_hozac, yep. I found it and adapted it to 2.2.0-pre4 on 2.6.20.1 1171992557 M * derjohn just one hopefully trivial reject 1171992642 M * Bertl daniel_hozac: can we get an rc3 for util-vserver with the shiny16 and the &>> fixes? for testing? 1171992687 M * daniel_hozac i've been meaning to. 1171992702 M * Bertl okay, please double check that your shiny16 is the latest 1171992712 M * Bertl IIRC, I updated it in place at some point 1171992722 M * daniel_hozac yeah, i refetched it after you told me. 1171992729 M * Bertl okay, great! 1171992735 M * daniel_hozac http://svn.linux-vserver.org/projects/util-vserver/changeset/2499 is the refetch diff, FWIW. 1171992766 M * Bertl yeah, looks food 1171992771 M * Bertl hmm, good :) 1171992772 M * Hollow yummi indeed 1171992773 M * Hollow :) 1171992796 M * Hollow have to update too.. it's really shiny the last weeks ;) 1171992855 M * daniel_hozac Bertl: as for the location of the utils... i don't like it either, but i'm not sure where to put it. 1171992865 M * Bertl Hollow: yes, hopefully it will shine for a long time ... 1171992881 M * Bertl daniel_hozac: you have a dir called 't' (Tools) 1171992896 M * Bertl it would at least make sense for me to find it there 1171992898 M * daniel_hozac yeah, but that's more for small test things/scripts. 1171992901 M * Hollow any reason why you don't implement it directly in assembler instead of workaround gcc inline assembly bugs? 1171992905 M * Bertl e.g. /t/uv/testing 1171992919 M * Bertl Hollow: yes, several reasons actually 1171992942 M * daniel_hozac i guess getting errno right would be the hardest problem. 1171992962 M * Bertl 1) i386 has 8 registers, 3 of them have special purpose 1171992975 M * Bertl we need to load 6 registers with syscall values 1171993001 M * Bertl 2) assembler code cannot be inlined 1171993012 M * Bertl (unless you use inline assembler :) 1171993040 M * Bertl 3) most archs actually get it right, so i386 is just the ugly duck :) 1171993050 M * Hollow (as always :) 1171993072 M * Bertl 4) this way we can avoid excessive register reloading 1171993085 M * Bertl 5) I like challenges :) 1171993097 M * Hollow :) 1171993252 Q * baldy Ping timeout: 480 seconds 1171993309 J * muuhDBX ~foo@a213-22-7-229.cpe.netcabo.pt 1171993348 P * muuhDBX 1171993506 M * Bertl okay, off for now ... back later 1171993509 N * Bertl Bertl_oO 1171993532 M * daniel_hozac http://people.linux-vserver.org/~dhozac/t/uv-testing/util-vserver-0.30.213-rc3.tar.bz2 1171993591 Q * muuhDBXi Ping timeout: 480 seconds 1171993682 Q * jmcaricand Remote host closed the connection 1171994333 J * baldy baldy@baldy.biz 1171994642 Q * baldy Read error: Connection reset by peer 1171994665 Q * dlezcano Read error: Connection reset by peer 1171994891 Q * mnemoc Ping timeout: 480 seconds 1171995056 M * daniel_hozac Hollow: have you had a chance to test the Gentoo initscripts in 0.30.213*? 1171995387 J * mnemoc ~amery@kilo105.server4you.de 1171995500 J * dlezcano ~dlezcano@blueice4n2.uk.ibm.com 1171996170 N * Bertl_oO Bertl 1171996200 M * Bertl daniel_hozac: did blino provide some clues how to install mandriva via yum/urpmi/rpm/whatever zet? 1171996223 A * Bertl is using a different kezboard now .. as you can see :) 1171996234 M * daniel_hozac hehe 1171996249 M * daniel_hozac well, urpmi looked like a dead-end. 1171996262 M * Bertl okay, any viable alternatives? 1171996276 M * daniel_hozac not that i know of. 1171996290 M * daniel_hozac i was unable to use apt-rpm for the "apt" repo. 1171998018 M * nebuchadnezzar hello 1171998075 M * daniel_hozac hi 1171998136 Q * dlezcano Read error: Connection reset by peer 1171998266 Q * hardwire Ping timeout: 480 seconds 1171998626 Q * gerrit Ping timeout: 480 seconds 1171999190 J * gerrit ~gerrit@c-67-160-146-170.hsd1.or.comcast.net 1171999575 Q * dna Read error: Connection reset by peer 1171999575 J * dna ~naucki@137-221-dsl.kielnet.net 1171999618 J * jmcaricand ~jmcarican@d90-144-56-52.cust.tele2.fr 1172000955 Q * prae Quit: Quitte 1172002030 M * Bertl daniel_hozac: do you remember what the problem with the apt-rpm was? (with mandriva) 1172002120 M * daniel_hozac it seemed to require an apt that was capable of parsing the Mandriva repository format. 1172002131 M * daniel_hozac i.e. i couldn't find a typical apt-rpm repo. 1172002196 M * Bertl i.c. 1172002416 M * phedny Bertl: I can confirm grsec+vserver with ipv6 patch does compile and boot 1172002433 M * Bertl phedny: excellent! does it work too? 1172002443 M * phedny that I will test the upcoming days 1172002457 M * phedny at least testme.sh passes all tests 1172002458 M * Bertl okay, keep us posted then ... 1172002614 M * yang hey Bertl:) 1172002674 M * Bertl hey yang .. Ralf asked me to do some testing with the mips, when possible 1172002680 M * yang sure 1172002694 M * Bertl he suggested to try the 2.6.18-4? debian kernel on that 1172002701 M * yang ok 1172002702 M * mjt what people are using as syslog daemon in guests? Unless it's a complete installation... 1172002717 M * Bertl syslog-ng is quite nice 1172002754 M * mjt i mean, in a guest, only really minimal syslogd is needed - to forward everything to host 1172002775 M * daniel_hozac ... if that's what you want. 1172002775 M * mjt and syslog-ng to take care of all that 1172002798 M * daniel_hozac you could simply configure the host's syslog-ng to create sockets in guests as well. 1172002824 M * mjt i thought about bind-mounting /dev/log in guests. 1172002834 M * Bertl bad idea 1172002842 M * mjt but that way, host^Wguest name is lost 1172002883 M * mjt ie, all messages will look as they come from the host itself. 1172002935 M * mjt but configuring syslog-ng like that seems interesting. 1172003411 M * jmcaricand Hi alls. Does anyone use Oracle database in a guest ? 1172004041 J * VooDooMaster ~icechat5@p549C921C.dip0.t-ipconnect.de 1172004084 M * VooDooMaster Guten Abend! 1172004123 M * VooDooMaster Getreu dem Motto im Topic wollte ich mal ein paar Fragen los werden ;) 1172004172 M * VooDooMaster Ich hab mal vor einiger Zeit (~8 Monate) mit VServern unter Gentoo und jetzt heute wieder unter Debian experimentiert 1172004188 M * VooDooMaster ich hab gesehen, dass sich da EINIGES getan hat *freu* 1172004199 M * VooDooMaster was ist vunify? 1172004212 M * Bertl okay, sounds good, channel language is english though ... 1172004227 M * VooDooMaster sorry ... 1172004241 M * VooDooMaster *switching to my bad english* 1172004245 M * Bertl np, I can understand you, but others might not :) 1172004281 M * Bertl to anser your question, vunify is a tool to use (and apply) unification 1172004285 M * VooDooMaster Hi Bertl! You were a great help at the time I tried vserver for the first time! 1172004306 M * Bertl a method to reduce disk space and memory (mostly caching) resources 1172004315 M * VooDooMaster I think I have to start from the very beginning ... maybe advancing a bit faster ;) 1172004324 M * VooDooMaster I set up a base/host system ... 1172004332 M * VooDooMaster then a vserver ... 1172004340 M * VooDooMaster an another ... and so on .. 1172004358 M * VooDooMaster now with gentoo I have the problem that I need the portage tree in every vserver .. 1172004385 M * Bertl yes, depending on the distro, you can use prebuilt packages for kernels and tools, the guests are often kept separate 1172004416 M * Bertl we are aware of the gentoo portage issue, and IIRC, there is quite a nice workaround for that available 1172004424 M * VooDooMaster I think I'll stick with gentoo mainly and have a look at ubuntu-server/debian ... 1172004426 M * Bertl I guess Hollow will have more details on that 1172004434 M * bonbons VooDooMaster: you have a few options for that, e.g. bind-mounting the portage tree into the guests 1172004494 M * VooDooMaster hi bonbons! thaks - that was the way I did it when first trying vservers - is that still the way to go? fine! 1172004528 M * bonbons That's one option, and certainly the most efficient one if you have few hosts 1172004538 M * VooDooMaster for things i need in more than one vserver I used ccache ... but I read that some use portage overlays with binaries - how's that? 1172004570 M * bonbons what do you mean by using portage overlays with binaries? 1172004583 M * bonbons is it using binpkg feature of portage? 1172004587 M * VooDooMaster humm ... how much is view ... my test setup will be about 5 to 6 hosts ... ... production setup maybe ... 20 (max) 1172004661 M * VooDooMaster if I understood it right the guys know what all vservers need and build a binpkg ... using that one in all n-1 vservers 1172004694 M * Bertl 5-6 guests I assume? 1172004704 M * Bertl or 5-6 hosts with N guest_ 1172004706 M * VooDooMaster isn't there an emerge option to always use a binpkg when there ist one available? 1172004731 M * VooDooMaster Bertl: Oh - sorry - you're right ... guests of course! 1172004745 M * bonbons that's fairly possible, a good way to do this would be having the portage tree on a single master host (including distfiles and packages subfolders) 1172004773 J * Aiken ~james@ppp126-23.lns2.bne4.internode.on.net 1172004774 M * VooDooMaster this number of hosts would be nice ;) 1172004774 M * bonbons all other hosts would NFS-mount the portage tree and share it with the guest via bind-mount 1172004779 M * Bertl morning Aiken! 1172004854 M * VooDooMaster for my testsetup I use one host and 5-6 guests ... so I don't need NFS right? ... only when I share it on the network? 1172004876 M * bonbons (read-only mount where needed), the master vserver/host would emerge generating binpkgs at same time, all other vservers would the install from binpkgs as available 1172004893 M * bonbons NFS is for when you have multiple hosts 1172004908 M * bonbons when you expand ;) 1172004924 M * Bertl okay, off for now ... back later ... 1172004925 M * VooDooMaster which distro do you guys use your vservers on? (don't want to start a "better-than...-because.."-discussion!) 1172004929 N * Bertl Bertl_oO 1172004949 M * VooDooMaster CU Bertl 1172004988 M * bonbons I do Gentoo only at home, but run some SuSE guest at work as servers there are running SuSE and I need to build RPMs for those 1172005086 M * VooDooMaster bonbons: how can the hist build binpkgs for the guests? .. the vservers can have installed different packages ... that are not on the host and not on (all) other vservers 1172005101 M * bonbons but I share the portage tree using SquashFS images 1172005144 M * bonbons you add --buildpkg option to portage and share the $PORTAGE/packages dir with the vservers 1172005170 M * bonbons the only thing you need to be careful about are packages that happen to be compiled with different use-flags depending on host/vserver 1172005213 M * bonbons binpkg does not support maintaining binary archive for multiple USE-flag combinations 1172005275 Q * jmcaricand Quit: KVIrc 3.2.4 Anomalies http://www.kvirc.net/ 1172005281 M * VooDooMaster I turned on my old gentoo-vserver-test machine today and tried to update ... a nightmare ... /etc/make.profile not valid anymore ... and nearly all packages new! ... aber vserver build does not work :/ 1172005309 M * bonbons (but that might change in the future), for different archs (e.g. x86 versus x86_64) you need to have to distinct packages directories 1172005330 M * VooDooMaster ah - so I have to keep the use flags in sync on all vservers - that should be possible ... I hope ;) 1172005391 M * bonbons for base packages that's easy, but e.g. if you have one vserver with mysql server and all others just mysql clients you either install server everywhere or do some handwork 1172005693 M * VooDooMaster ah - good point - what about dependencies? 1172005706 Q * gerrit Read error: Operation timed out 1172005717 M * VooDooMaster one of my plans was to test vservers by splitting up services 1172005737 M * VooDooMaster one samba server, one mail server, one web server, one ldap server, ... 1172005742 M * bonbons well most of it is use-flags based 1172005767 M * bonbons or do you mean dependencies to start vservers in the right order? 1172005772 M * VooDooMaster humm ... but then I can skip the binpkgs ... 1172005777 M * VooDooMaster damn 1172005804 M * VooDooMaster first the starting order and second the package dependencies ... 1172005817 Q * micah Read error: Operation timed out 1172005828 M * VooDooMaster say I want an apache with mysql support bit an extra database vserver ... 1172005846 M * bonbons you certainly want binpkg for everything in system, for any packages that's installed more than once 1172005851 M * VooDooMaster bit sould be but ;) 1172005895 M * VooDooMaster ah - ok ... then I have to select the binpkgs to be built "by hand"? 1172005928 M * bonbons the way I did it is having binpkg for everything on vservers, just manually replacing the tbz2 in /usr/portage/packages/All/.bzip2 to mach the correct use-flags 1172005990 M * bonbons emerge will either ignore your use-flags or offer to build the package from ebuild (for those you would have to check what binpkg is there and eventually back it up for some other vserver) 1172006119 J * dlezcano ~dlezcano@AToulouse-252-1-85-27.w86-201.abo.wanadoo.fr 1172006413 M * VooDooMaster I didn't get that ... (rereading the last two lines) ... one moment ;) 1172006502 J * micah ~micah@micah.riseup.net 1172006593 Q * Hollow Quit: Disconnecting from stoned server. 1172006733 M * VooDooMaster How do you do that "use-flag-matching" in /All/ ? 1172006802 J * Hollow ~hollow@styx.xnull.de 1172006854 M * bonbons I emerge ... --pretend --verbose and check what portage would do. If it wants to rebuild something I already have a bin-pkg for (different use-flags) I move /usr/portage/packages/All/${P}.bz2 to /usr/portage/packages/All/${P}_XXX.bz2 and start over again 1172006870 M * mjt what `vserver start' executes inside guest? 1172006889 M * bonbons when I already have a matching binpkg (not active on) I move it in place (or copy it in place) 1172006925 M * bonbons mjt: that depends on init style... either some rc script or the vserver's init 1172006955 M * VooDooMaster ah - ok ... now I get it --- sorry - am a bit slow today ... was partying till 5am ;) 1172007002 M * bonbons no problem :) 1172007003 M * VooDooMaster but without manual intervention it's not possible? ... vupdateworld -a 1172007027 M * VooDooMaster and nothing more would be great ;) 1172007053 M * VooDooMaster gentoo sometimes is sooo damn slow if you need something fast :/ 1172007092 M * bonbons no, that's not possible yet, but I hope portage guys will fix that restriction on binpkgs so portage can do it transparently 1172007199 M * VooDooMaster you mean by handling the same program compiled with different use flags? 1172007205 M * VooDooMaster that woul be great ... 1172007227 M * bonbons yes 1172007232 M * VooDooMaster but a hard piece of work considering the huge number of use flags 1172007236 J * hardwire ~hardwire@rdbck-6299.wasilla.mtaonline.net 1172007271 M * bonbons not really, it's just a matter of tagging the tbz2 files and choosing the right one if it exists 1172007325 M * bonbons binpkgs built by gentoo will never exist anyhow, but a single binpkg repository should be able to hold packages compiled with different use-flags 1172007393 M * mjt bonbons: found it by using `bash -x'.. but how it knows that it should run `/etc/init.d/rc 3' ? 1172007481 M * bonbons I think that's hardcoded for the default init style (well checking if /etc/rc.d/rc or /etc/init.d/rc exists) 1172007525 M * bonbons the other way would be using init style = plain which runs a full init inside the guest (automatically supports various distros without effort) 1172007526 M * daniel_hozac yes, the default initstyle is sysv. 1172007595 M * mjt wow. `vserver stop' generates tons of scary error messages!.. ;) 1172007613 J * DavidS ~david@chello062178045213.16.11.tuwien.teleweb.at 1172007624 M * mjt Unmounting temporary filesystems...umount: none: not found 1172007625 M * mjt umount: /tmp: must be superuser to umount 1172007630 M * mjt mount: permission denied 1172007637 M * mjt /usr/lib/util-vserver/vserver.stop: line 85: 26447 Killed 1172007668 M * mjt there's no reason to umount them as namespace cleanup will do that automatically, right? 1172007675 M * VooDooMaster one side question: a friend of mine deleted his /bin/login and can't log on ;) ... we booted with init=/bin/bash but how can we auto-login for ex. root to reemerge the right package ... (it's a notebook) 1172007720 M * mjt create a fake /bin/login that just executes shell 1172007746 M * mjt or try to bring the network up 1172007758 M * mjt from you bash-init ;) 1172007790 M * mjt i think that the first way is easier if you don't know exactly how it all works ;) 1172007846 M * daniel_hozac mjt: yep, that's what guests which have not been properly cleaned will output. 1172007872 M * hardwire sup fooz 1172007898 M * mjt daniel_hozac: "cleaned" = all hw-related stuff from init.d removed? 1172007913 M * daniel_hozac yep. 1172007948 M * VooDooMaster mjt ... icqed him that - thanks! 1172007961 M * mjt VooDooMaster: don't forget to chmod+x your new login :) 1172008021 M * VooDooMaster I hope he'll do that ... hmm .. I'll write him that, too .. just in case ;) 1172008042 J * gerrit ~gerrit@c-67-160-146-170.hsd1.or.comcast.net 1172008378 Q * bonbons Quit: of to bed 1172008433 M * VooDooMaster uhm .. bye bonbons - many thanks ;) 1172008633 M * mjt wug. I ran stripserver, it gave tons of errors, and now it doesn't mount /dev/pts (and has no devices in /dev/, either) 1172008673 M * daniel_hozac "stripserver"? 1172008691 M * mjt from vserver-debiantools 1172008703 M * mjt Description: Tools to manage debian virtual servers 1172008715 M * mjt ok-k. Doing it manually ;) 1172008715 M * daniel_hozac i guess you'll have to talk to Ola about that. 1172008735 M * mjt but i wonder what it did so that /dev is broken 1172008743 M * daniel_hozac probably removed it completely. 1172008747 M * mjt it doesn't mount /dev/pts 1172008797 M * mjt aha 1172008798 Q * meandtheshell Quit: Leaving. 1172008806 M * mjt it created /etc/vservers/guest0.conf 1172008815 N * DoberMann DoberMann[ZZZzzz] 1172008816 M * mjt like, old-style config -- right? 1172008831 M * daniel_hozac usually legacy, but sure. 1172008930 M * mjt no, removing guest0.conf didn't help 1172008988 M * daniel_hozac those are only used if there is no directory. 1172009033 M * mjt it's the host (vserver scripts) that mounts all that stuff for guest, right? Ie, it's none of guests business? 1172009052 M * mjt (mount in guest says 'permission denied') 1172009055 M * daniel_hozac yep. 1172009078 Q * eyck_ Server closed connection 1172009088 A * mjt does more `bash -x' dance... 1172009102 M * DavidS [2.0 series] can i workaround the fact that a guest without the src ip of a route cannot use (connect: Invalid argument) this route? 1172009104 J * eyck eyck@kuszelas.com 1172009109 M * VooDooMaster bye! and thanks to all! good night! 1172009130 Q * VooDooMaster Quit: Not that there is anything wrong with that 1172009168 M * daniel_hozac DavidS: that's news to me... 1172009192 Q * Hollow Remote host closed the connection 1172009194 M * daniel_hozac DavidS: works fine here. 1172009408 J * Hollow ~hollow@styx.xnull.de 1172009440 M * mjt can a vserver dir (the root of a guest) be a symlink instead, or better use bind-mount? 1172009497 Q * Hollow Remote host closed the connection 1172009548 Q * phreak`` Remote host closed the connection 1172009561 J * Hollow ~hollow@styx.xnull.de 1172009609 J * phreak`` ~phreak``@deimos.barfoo.org 1172009653 Q * Hollow Remote host closed the connection 1172009708 Q * [PUPPETS]Gonzo Server closed connection 1172009712 P * stefani I'm Parting (the water) 1172009721 J * [PUPPETS]Gonzo gonzo@langweiligneutral.deswahnsinns.de 1172009787 J * Hollow ~hollow@styx.xnull.de 1172009902 Q * Johnnie Server closed connection 1172009933 M * daniel_hozac mjt: why not just point /etc/vservers//vdir to the new location? 1172009933 J * Johnnie ~jdlewis@jdlewis.org 1172009992 M * DavidS the guest has only one ip from 127.0.0.0/8 and all tries to contact the outer world (via SNAT) result in one or other "Invalid argument" error... AFAICT the problem is that ip route get $something says "... src $external_IP ..." which is not available in the guest 1172010032 M * daniel_hozac you shouldn't use 127.0.0.0. 1172010091 M * daniel_hozac use a private IP address and i'm willing to bet it'll start working. 1172010140 M * DavidS i fear so ;) 1172010461 Q * gerrit Ping timeout: 480 seconds 1172010657 M * DavidS indeed, using 192.186.0.1 as guest ip, I can SNAT it ... 1172011228 Q * brcc Server closed connection 1172011234 J * brcc bruce@i.am.someasshole.com 1172011323 M * mjt hmm. so at least some process is required to be running in a guest? :) 1172011931 Q * cehteh Server closed connection 1172011948 J * cehteh ~ct@pipapo.org 1172012573 Q * Loki|muh Server closed connection 1172012574 J * Loki|muh loki@satanix.de 1172012876 Q * DavidS Quit: Leaving. 1172013816 J * shedi ~siggi@ftth-237-144.hive.is 1172013899 Q * ensc Killed (NickServ (GHOST command used by ensc_)) 1172013909 J * ensc ~irc-ensc@p54B4D816.dip.t-dialin.net 1172013956 Q * michal`_ Ping timeout: 480 seconds 1172014417 J * michal` ~michal@www.rsbac.org 1172015296 Q * Piet Quit: Piet 1172015422 J * click click@ti511110a080-0949.bb.online.no 1172015498 J * gerrit ~gerrit@c-24-22-119-159.hsd1.mn.comcast.net 1172015938 Q * dna Quit: Verlassend