1330474292 J * notau ~notau@gw-1.mel1.paranode.id.au
1330475688 Q * notau Quit: Computer has gone to sleep.
1330476559 J * notau ~notau@gw-1.mel1.paranode.id.au
1330476679 Q * notau 
1330477415 Q * andrew Quit: http://www.mibbit.com ajax IRC Client
1330483342 Q * hparker Remote host closed the connection
1330492909 N * Bertl_zZ Bertl
1330492913 M * Bertl back now ...
1330499179 J * ncopa ~ncopa@3.203.202.84.customer.cdi.no
1330499945 J * ghislain ~AQUEOS@adsl2.aqueos.com
1330499945 Q * jeroen__ Read error: Connection reset by peer
1330501556 M * pmjdebruijn @/win 14
1330501558 M * pmjdebruijn sorry
1330502881 J * derjohn_mob ~aj@87.253.171.219
1330503991 M * ncopa hi
1330503997 M * ncopa the automake error came back
1330504006 M * ncopa i wonder how i made it work last time
1330504041 M * ncopa automake-1.11.3 will not work with current util-linux
1330504089 Q * ensc|w Remote host closed the connection
1330504097 J * ensc|w ~ensc@www.sigma-chemnitz.de
1330505635 J * jeroen__ ~jeroen@095-097-051-172.static.chello.nl
1330508125 M * Mr_Smoke Bertl: hi. What would be the least expensive way (resource wise) to get the vserver's name (not NodeName) from its XID ?
1330508250 M * Bertl the syscall switch command to retrieve the uts information
1330508286 M * Mr_Smoke Heh :) ok let me rephrase
1330508304 M * Mr_Smoke Bertl: hi. What would be the least expensive way (resource wise) to get the vserver's name (not NodeName) from its XID using bash/perl/userland stuff ?
1330508316 M * Mr_Smoke it's for a munin plugin
1330508407 M * daniel_hozac python -c 'import libvserver; print libvserver.get_vhi_name('$XID', "CONTEXT")'
1330508438 M * Mr_Smoke is there no way to get it from the proc filesystem directly ?
1330508480 M * daniel_hozac no.
1330508508 M * Mr_Smoke Well I guess I could parse all /etc/vservers
1330508574 M * Bertl so, if you plan to do fancy stuff, why do you ask for the most efficient way in the first place? :)
1330508595 M * Mr_Smoke I didn't know it was *that* fancy a thing to do :)
1330508600 M * Mr_Smoke Sorry about the phrasing
1330508608 M * daniel_hozac seems trivial to me...
1330508628 M * daniel_hozac but if you don't care how accurate it is, you could always just check /var/run/vservers.rev
1330508634 M * Mr_Smoke ah
1330508660 M * Mr_Smoke Why would it not be accurate ?
1330508687 M * daniel_hozac there are various reasons why it could be stale.
1330508702 M * Mr_Smoke owkay
1330508720 M * Mr_Smoke I'll probably stick to the NodeName then
1330508724 M * Mr_Smoke Thanks :)
1330508766 M * daniel_hozac not sure how parsing /proc is better than a syscall, but...
1330508793 M * Mr_Smoke daniel_hozac: I'm using debian's vserver munin plugin
1330508823 M * Bertl big mistake I guess :)
1330508828 M * Mr_Smoke Aha ?
1330508832 M * Mr_Smoke how so ?
1330508845 M * Mr_Smoke They work pretty well so far
1330508845 M * Bertl well, it contains the mistake keyword 'debian' :)
1330508848 M * Mr_Smoke Hahahaha
1330508855 M * Mr_Smoke And it's not even friday yet ;)
1330508876 M * Mr_Smoke is friday trollday elsewhere in the world ? In France it is, that's why :p
1330508893 M * Bertl but seriously, the debian bugs got less recently ...
1330508913 M * Mr_Smoke Well the plugins are not debian specific really
1330508925 M * Bertl either the debian users get more used to things not working, or the bugs are actually fixed or not relevant anymore
1330508928 M * Mr_Smoke all they do is get data from proc in order to do cpu and memory accounting
1330508956 M * Bertl you could always add a simple patch to show the information in /proc, you know?
1330508958 M * Mr_Smoke I was just curious as to why they'd use each vserver's nodename instead of context and whether it'd be easy to change that
1330508972 M * Mr_Smoke everything else is fie
1330508973 M * Mr_Smoke fine
1330508996 M * daniel_hozac or maybe we finally got the message across to not bother with the Debian packages...
1330509000 M * Mr_Smoke :D
1330509059 M * daniel_hozac so it's easier to get the guest's name, but you're sticking to nodename anyways?
1330509102 M * Mr_Smoke you said it's easy but inaccurate
1330509124 M * daniel_hozac to check the symlink.
1330509132 M * daniel_hozac since you for some reason don't believe the kernel?
1330509138 M * Mr_Smoke I never said I didn't
1330509188 M * Mr_Smoke It's just that if for some reason the vserver user decides to change their hostname, all the munin rrd data goes away
1330509201 M * Mr_Smoke so I'd rather index them by context, since that is less likely to change
1330509238 M * Bertl sounds reasonable, especially as the context _is_ the real identifier
1330509241 M * Mr_Smoke So the question really is, how stale can vservers.rev actuall be ?
1330509251 M * Bertl why do you care at all?
1330509257 M * Mr_Smoke Bertl: yeah, it strikes me as odd that they'd use the nodename at all in the first place
1330509262 M * Mr_Smoke Well
1330509271 M * Bertl all information you gather from /proc is indexed _by_ the xid
1330509285 M * Mr_Smoke I'm after the context name in that case
1330509296 M * daniel_hozac and what's wrong with vuname/python?
1330509298 M * Bertl there is no 'context name'
1330509312 M * Mr_Smoke Bertl: vserver <stuff> start/stop/blah
1330509323 M * Mr_Smoke "stuff" is what I need :)
1330509361 M * Bertl but not for the proc stuff
1330509375 M * Mr_Smoke daniel_hozac: didn't know about vuname. Turns out basename `vuname -g --xid 182 context` is just what I need :)
1330509378 M * Mr_Smoke thanks !
1330509398 M * Bertl anyway, for mapping the xid to the guest path, the given commands work fine
1330509409 M * Mr_Smoke Bertl: true, the name is only for display in munin, and because it is used to name the rrds, I need it to be explicit AND rather stable through time
1330509421 M * Bertl bad mistake
1330509427 M * Mr_Smoke How so ?
1330509436 M * Bertl use the xids for the rrds :)
1330509471 M * Mr_Smoke Bertl: yeah but when checking munin it's not that easy to wonder why context 1742 is using su much more memory than 337
1330509474 M * Mr_Smoke :)
1330509521 M * Bertl there is nothing wrong with displaying all uts information (and the guest path) there
1330509551 M * Mr_Smoke How do you mean?
1330509556 M * Mr_Smoke I don't follo
1330509561 M * Mr_Smoke w
1330509580 M * Bertl there is nothing wrong with _displaying_ the information for the user
1330509603 M * Bertl but you shouldn't use it for database names etc unless you have a good reason to do so
1330509613 M * Mr_Smoke That's not my choice
1330509644 M * Mr_Smoke Or do you mean it's possible to force munin to name the rrds differently and still display the name properly ?
1330509671 M * Bertl no idea, I don't use munin and I don't have the details on the plugin in question
1330509672 M * Mr_Smoke the naming scheme is just the result of enabling the plugin, I didn't chose it to be that way
1330509698 M * Mr_Smoke In other words, what the plugin does is look into /proc/virtual, get the data AND the NodeName from there, and report it
1330509711 M * Mr_Smoke Munin then stores that into NodeName-whateverdata.rrd
1330509719 M * Mr_Smoke I don't know if I have a say in that
1330509748 M * Bertl I presume if the 'NodeName' were the 'XID' it would use that, no?
1330509773 M * Mr_Smoke It would, but the NodeName is what the guest reports as its hostname, so that can change any minute
1330509788 M * Bertl but the xid won't
1330509806 M * Mr_Smoke Sure, so if I could use the XID for naming and the context for display, that'd be ideal
1330509816 M * Mr_Smoke Not sure that's possible though
1330509819 M * Bertl we are going in circles here
1330509824 M * Mr_Smoke Yeah
1330509838 M * Mr_Smoke I'd rather use the context "name" and be done with it
1330509845 M * Bertl you said: the plugin looks up the NodeName and reports it back to munin
1330509846 M * Mr_Smoke So I'll probably use vuname
1330509848 M * Mr_Smoke Yeah
1330509862 M * Bertl so, change the plugin to report the _XID_ as NodeName
1330509872 M * Mr_Smoke whn you say XID, you mean the numeric id right?
1330509882 M * Bertl xid = context id
1330509887 M * Bertl and yes, that's a number
1330509890 M * Mr_Smoke yes
1330509891 M * Mr_Smoke BUT
1330509910 M * Mr_Smoke munin will use whatever the plugin reports for BOTH the rrd name AND the data identifier on the graphs
1330509924 J * petzsch ~markus@dslb-188-103-057-014.pools.arcor-ip.net
1330509938 M * Mr_Smoke so if I use the XID, my rrds will be named after the XIDs, but the graphs will be much harder to read
1330509941 M * Bertl that's a bad design and should be fixed
1330509965 M * Bertl what if you want your graph to be labeled with something like 'Test/123' ?
1330509979 M * Mr_Smoke I dunno, I haven't looked it up yet
1330509980 M * Bertl it won't be able to use that for the database name
1330510027 M * Mr_Smoke True
1330510030 Q * petzsch 
1330510040 M * Mr_Smoke The plugin must be able to handle that
1330510048 M * Mr_Smoke It's just that the default behaviour sucks then
1330510065 J * petzsch ~markus@dslb-188-103-057-014.pools.arcor-ip.net
1330510146 J * BenG ~bengreen@cpc10-aztw24-2-0-cust114.aztw.cable.virginmedia.com
1330510442 M * ncopa i am sending some autotools cleanup patches
1330511825 M * ncopa \o/
1330511833 M * ncopa i found the issue with autmake 1.11.2 and newer
1330511845 M * ncopa this one causes it:
1330511846 M * Bertl congrats!
1330511848 M * ncopa http://git.savannah.gnu.org/cgit/automake.git/commit/?h=maint&id=9ca632642b006ac6b0fc4ce0ae5b34023faa8cbf
1330512066 M * ncopa I have a patch for it
1330512167 M * ncopa did I send the "silence 'non-POSIX variable name' warnings in automake" patch?
1330512171 M * ncopa i think i didnt
1330512745 Q * Mr_Smoke Read error: Connection reset by peer
1330512747 J * Mr_Smoke ~smokey@layla.lecoyote.org
1330513310 Q * pmjdebruijn Remote host closed the connection
1330513312 J * pmjdebruijn ~pascal@overlord.pcode.nl
1330514239 J * dna ~dna@hmbg-4d06df8d.pool.mediaWays.net
1330514413 J * notau ~notau@gw-1.mel1.paranode.id.au
1330515036 M * Bertl off for a nap ... bbl
1330515040 N * Bertl Bertl_zZ
1330515319 J * ghislain1 ~AQUEOS@adsl2.aqueos.com
1330515357 M * daniel_hozac ncopa: pkglibexec isn't the same.
1330515403 M * daniel_hozac i.e. that patch will break everything.
1330515481 M * daniel_hozac and neither is pkgdata.
1330515524 J * ghislain2 ~AQUEOS@adsl2.aqueos.com
1330515608 Q * ghislain Ping timeout: 480 seconds
1330515791 J * ghislain ~AQUEOS@adsl2.aqueos.com
1330515803 Q * ghislain1 Ping timeout: 480 seconds
1330516006 Q * ghislain2 Ping timeout: 480 seconds
1330516392 J * ghislain1 ~AQUEOS@adsl2.aqueos.com
1330516663 Q * ghislain Ping timeout: 480 seconds
1330516949 M * ncopa ok?
1330516956 J * ghislain ~AQUEOS@adsl2.aqueos.com
1330516990 M * ncopa daniel_hozac: better suggestion how to solve it?
1330517156 M * BenG heya all
1330517169 M * BenG I'm upgrading some servers to Debian Wheezy, guest servers that is
1330517175 M * BenG I get this when upgrading:
1330517177 M * BenG Setting up sysvinit (2.88dsf-22) ...
1330517177 M * BenG sysvinit: restarting...init: timeout opening/writing control channel /run/initctl
1330517177 M * BenG .init: timeout opening/writing control channel /run/initctl
1330517193 M * BenG the last line is repeated indeffinatley
1330517228 M * BenG ah, no, I lied, only 7 times
1330517233 M * BenG and then I get:
1330517234 M * BenG failed.
1330517245 M * BenG after that loads more packages installed
1330517259 M * BenG I'm just reporting that anyhow
1330517265 Q * ghislain1 Ping timeout: 480 seconds
1330517274 M * daniel_hozac sounds like as it should be
1330517314 M * daniel_hozac ncopa: stick with automake 1.11.1, i guess.
1330517486 J * ghislain1 ~AQUEOS@adsl2.aqueos.com
1330517734 J * deathtje ~deathtje@216.67.225.106
1330517741 Q * ghislain Ping timeout: 480 seconds
1330517752 J * ghislain ~AQUEOS@adsl2.aqueos.com
1330517761 M * Mr_Smoke heh, /proc/meminfo inside a guest is fun
1330517799 M * Mr_Smoke basides the fact that active, inaactive and committed value are from the host, the default munin memory plugin computes "apps" to be negative :)
1330517804 M * BenG upgrading to Wheezy went fine BTW all
1330517816 M * deathtje just wondering how good is vserver as a security / application permission application ( the basic functionality should be the ability to limit interaction between applications )
1330517889 M * daniel_hozac should be perfect.
1330517981 J * ghislain2 ~AQUEOS@adsl2.aqueos.com
1330518058 Q * ghislain1 Ping timeout: 480 seconds
1330518201 Q * ghislain Read error: Operation timed out
1330518356 J * ghislain ~AQUEOS@adsl2.aqueos.com
1330518395 M * ncopa err... unfortunally is not util-vserver the only application we provide with alpine linux so sticking to old version of automake will not be possible in the long run
1330518593 J * ghislain1 ~AQUEOS@adsl2.aqueos.com
1330518616 Q * ghislain2 Ping timeout: 480 seconds
1330518661 M * daniel_hozac you also don't have to run automake though.
1330518678 Q * ghislain Read error: Operation timed out
1330518705 M * ncopa i dont run it for fun
1330518777 J * ghislain ~AQUEOS@adsl2.aqueos.com
1330518856 M * ncopa ok
1330518860 M * ncopa i removed the patches
1330518864 M * ncopa here is what i get:
1330518873 M * ncopa configure: creating ./config.status
1330518873 M * ncopa config.status: error: cannot find input file: `util-vserver.spec.in'
1330518929 M * daniel_hozac with a pristine tarball?
1330518944 M * daniel_hozac because that file is there.
1330518973 M * ncopa yup
1330518986 M * ncopa pristine tarball
1330518991 M * daniel_hozac so why did you remove it?
1330519010 M * ncopa i didnt
1330519026 M * ncopa its there
1330519037 M * ncopa the configure script is broke some how
1330519050 M * daniel_hozac yet it works for everyone else.
1330519070 M * ncopa everyone else does not use busybox
1330519078 Q * ghislain1 Ping timeout: 480 seconds
1330519083 M * ncopa look
1330519092 M * daniel_hozac so that sounds like a problem with autoconf and busybox.
1330519094 M * ncopa everyone else does not runn their vserver host frrom tmpfs
1330519154 M * ncopa what do you think autoconf devs tells me when i say util-vserver doesnt work?
1330519157 M * ncopa or busybox devs?
1330519168 M * ncopa i do send patches
1330519177 M * daniel_hozac maybe you should figure out _why_ it doesn't work.
1330519199 M * ncopa i suppose i did once for years ago
1330519218 M * ncopa but sure...
1330519240 M * ncopa the conclusion back then was that it was eaisest to patch som m4 macros
1330519245 M * ncopa and rerunn automake
1330519286 M * daniel_hozac well, there's always an easy and a correct way to fix things.
1330519312 M * daniel_hozac one fixes it for good, one fixes it for right now.
1330519345 M * ncopa i know
1330519357 M * ncopa thats why i'm here today
1330519362 M * ncopa i want fix it for good
1330519366 M * ncopa the "correct way"
1330519384 J * ghislain1 ~AQUEOS@adsl2.aqueos.com
1330519395 M * ncopa util-vserver is one of those packages that "always" break
1330519402 M * ncopa 1000 others just works
1330519426 M * ncopa so i'm interested in fixing things for good
1330519563 Q * ghislain Ping timeout: 480 seconds
1330519602 M * ncopa btw
1330519608 M * ncopa i dont need the .spec file
1330519703 M * daniel_hozac i doubt any of the others would work either.
1330519726 Q * Aiken Remote host closed the connection
1330519768 M * ncopa yup
1330519773 M * ncopa you are right
1330519777 J * ghislain ~AQUEOS@adsl2.aqueos.com
1330519845 M * ncopa i guess its the ensc_pathprog patch that fixes it
1330519858 M * ncopa but kinda difficult to know now that i cannot runn automake
1330519864 M * daniel_hozac how would that be related?
1330519894 M * ncopa im just guessing/intuition
1330519898 M * ncopa I'll check it
1330519971 J * ghislain2 ~AQUEOS@adsl2.aqueos.com
1330519986 Q * ghislain1 Ping timeout: 480 seconds
1330519988 M * ncopa i have a feeling that its sed that doesnt work as epxepcted
1330520058 M * daniel_hozac that all looks like pure bash.
1330520063 M * ncopa ENSC_PATHPROG_SED
1330520073 M * ncopa still just guessing though
1330520079 M * daniel_hozac try a sh -x config.status
1330520092 M * daniel_hozac might provide some more clues.
1330520182 M * ncopa no config.status
1330520196 M * ncopa iirc
1330520226 M * ncopa http://git.alpinelinux.org/cgit/aports/tree/main/util-vserver/ensc_pathprog.patch
1330520231 J * ghislain1 ~AQUEOS@adsl2.aqueos.com
1330520238 M * ncopa that one solves an issue with busybox
1330520257 M * daniel_hozac what issue, exactly?
1330520259 M * ncopa since sed is a symlink to /bin/busybox already
1330520260 Q * ghislain Ping timeout: 480 seconds
1330520273 M * ncopa it will think that the real sed command is /bin/busybox
1330520288 M * ncopa same with other commands
1330520299 M * daniel_hozac hmm.
1330520318 M * ncopa so the pach will check if the symlink is relative,
1330520320 M * ncopa ../
1330520330 M * ncopa and if so then resolve the symlink
1330520337 M * ncopa otherwise, just use it as is
1330520477 M * ncopa actually
1330520479 M * daniel_hozac that doesn't cover all cases though. e.g. /usr/bin/xyz linking to /bin/xyz which is a link to ../sbin/xyz
1330520503 M * ncopa thats the only patch left that needs get upstrreamed some place
1330520507 M * ncopa yeah
1330520519 M * ncopa i think it was more a quick and dirty thing
1330520523 M * daniel_hozac which would hit the problem it's aimed to circumvent.
1330520557 M * ncopa i suppose there is are reason it is done in first place
1330520569 J * ghislain ~AQUEOS@adsl2.aqueos.com
1330520580 M * ncopa also
1330520593 M * ncopa i wonder why i didnt check explicit for /bin/busybox
1330520608 M * ncopa and im pretty sure there is a reason why i didnt
1330520625 M * daniel_hozac well, in theory, any link to /* is fine.
1330520635 M * ncopa exactly what problem is it aimed to circumvent?
1330520655 M * daniel_hozac .. to / is not the same as /
1330520666 Q * ghislain2 Ping timeout: 480 seconds
1330520677 M * ncopa so it is the ../ that is the problem
1330520709 J * ghislain2 ~AQUEOS@adsl2.aqueos.com
1330520713 Q * ghislain1 Ping timeout: 480 seconds
1330520875 J * ghislain1 ~AQUEOS@adsl2.aqueos.com
1330521039 Q * ghislain Read error: Operation timed out
1330521285 Q * ghislain2 Ping timeout: 480 seconds
1330521344 N * Bertl_zZ Bertl
1330521352 M * Bertl back again ...
1330521360 Q * ghislain1 Ping timeout: 480 seconds
1330521523 M * ncopa daniel_hozac: how about this: http://sprunge.us/YdRi
1330521588 M * daniel_hozac looks fine
1330521616 M * ncopa you want me to send a proper patch to mailing list with proper signed-off tag?
1330521630 M * daniel_hozac we don't use them.
1330521644 M * ncopa do you use git am?
1330521652 M * daniel_hozac no.
1330521659 M * ncopa its very handy
1330521667 M * daniel_hozac i never get patches. :)
1330521671 M * ncopa lol
1330521703 M * ncopa basically, its just to pipe the raw email to 'git am'
1330521714 M * daniel_hozac sure, i know how to use it.
1330521724 M * ncopa goody
1330521734 M * ncopa i really like git
1330522073 M * daniel_hozac try http://people.linux-vserver.org/~dhozac/t/uv-testing/util-vserver-0.30.216-pre3029.tar.bz2
1330522170 M * ncopa \o/
1330522178 M * ncopa works like a charm
1330522197 M * daniel_hozac interesting.
1330522211 M * ncopa tack som tusan!
1330522231 M * daniel_hozac well, you did all the work ;)
1330522235 M * daniel_hozac so thank you.
1330522540 M * ncopa np
1330522556 M * ncopa i'm really happy get those things upstreamed
1330522572 M * daniel_hozac it tends to make life easier.
1330522581 M * ncopa it does indeed
1330522864 M * ncopa i suppose you at some point will have to deal with the automake-1.11.2 business
1330522871 M * ncopa but thats not my problem anymore :)
1330522997 M * daniel_hozac yeah, eventually. it's a fairly major operation though, since all the files will be moving...
1330523023 M * ncopa right
1330523049 M * ncopa i wonder if the "lets move everything into /usr" affects you
1330523056 M * ncopa probaly not much
1330523070 M * daniel_hozac not more than needing to rebuild it.
1330523082 M * ncopa thats good
1330523135 M * daniel_hozac but that global insanity should only happen on major upgrades anyway.
1330523850 Q * BenG Quit: I Leave
1330526498 Q * Alex[fob] Remote host closed the connection
1330527377 Q * notau Quit: Computer has gone to sleep.
1330527775 J * notau ~notau@gw-1.mel1.paranode.id.au
1330528572 Q * notau Quit: Computer has gone to sleep.
1330531451 J * dowdle ~dowdle@scott.coe.montana.edu
1330531662 Q * ncopa Quit: Leaving
1330533801 J * ghislain ~AQUEOS@adsl2.aqueos.com
1330533829 M * ghislain hello there , is there anyone that know well the blkio controler of the cgroups ? I need some help and will write an how to on the wiki when sorted out but i am lost now
1330534173 M * Bertl Documentation/cgroups/blkio-controller.txt
1330534181 M * Bertl but you probably know that already
1330534456 J * dna_ ~dna@130-160-103-86.dynamic.dsl.tng.de
1330534643 M * ghislain yes
1330534649 M * ghislain this clear like british fog
1330534680 M * ghislain and look at the blkio.* some speak of the device (sda) some of the md0 one and others from the lvm
1330534710 M * Bertl so we can asume it works for all of them :)
1330534782 J * hparker ~hparker@2001:470:1f0f:32c:222:19ff:fe22:b7b5
1330534791 M * ghislain yes but you do not know if the limit is to be set on the lvm or on the md one or sdx i lose all bearing so i would want to see if someone know then write down somethign so we can all use it
1330534811 M * ghislain the issue is that i do not know really how to test it after setting the limits 
1330534831 M * ghislain perhaps doign a dd something to write then read it to dev/null
1330534851 M * daniel_hozac seems like it should be sufficient
1330534900 Q * dna Ping timeout: 480 seconds
1330535620 M * ghislain if murphy's law do not comes into play, then i need to find 2/3h to test this, would have prefered to even pay a specialist for that, dam cgroup. I had a complete host crashed because a guest  used all diskIO 6GO of swap...
1330535657 M * ghislain i wonder why linux is not robust enoug to survive to this, i saw sun in the old days stay alive with more load that this
1330535723 M * Bertl a crash sounds unusual, do you have traces?
1330535748 Q * derjohn_mob Ping timeout: 480 seconds
1330535768 M * ghislain no the server just stopped answering to anything, nothing on the screen nada screen  on but no answer at all, no more disk activity, just dead
1330535787 M * Bertl magic-sysrq?
1330535822 M * ghislain not tried it the guy in the DC just tried ctrl-alt-sup when nothing answered but no answer
1330535908 M * ghislain so as there was no disk IO we just hit it with a hammer and restarted it
1330535929 M * ghislain standard procedure we have big hammer near all of our racks
1330535947 M * Bertl what brand do you use? *G*
1330536055 M * ghislain  Dell PowerEdge R710
1330536075 M * Bertl nah, the brand of the hammer ...
1330536087 M * ghislain oh
1330536101 M * ghislain it is a noname one as ong as it weigth
1330536145 M * Bertl well, sounds like a fabricated story to me ... all datacenters I know use high quality hammers from well known manufacturers :)
1330536202 M * ghislain dam, you cannot fool a kernel veteran so easely  !
1330536558 M * ghislain first lesson, you need to have cgroup memory swap control, if not when it reach the limit it will just swap until it kill the whole host !
1330536608 M * ghislain second lesson, blkio.weight do not take swap into the accounting
1330537415 J * derjohn_mob ~aj@88.128.19.5
1330537464 M * ghislain bertl: would you share one of your template config you setup on vserver for highly hostile environement on the guest ? (read customer that do php without even knowing what an algorithm is...)
1330537489 M * ghislain or anyone else...
1330537903 Q * derjohn_mob Ping timeout: 480 seconds
1330537993 Q * nkukard Remote host closed the connection
1330538126 M * Bertl ghislain: I do not really have any highly hostile guest setups atm, so no special config to prevent that either
1330538304 J * nkukard ~nkukard@41-133-198-167.dsl.mweb.co.za
1330538583 M * ghislain :) i was pretty sure of that but i tried anyway :p
1330541120 Q * dna_ Quit: Verlassend
1330542103 M * ard heh...
1330542127 M * ard it will not kill our hosts... I've seen tomcat's dissappear, but the host stays :-)
1330542144 M * ard The processes in the vserver get's terminated one by one like any normal OOM ;-)
1330542188 M * ard ah, and php is the same thing...
1330542194 M * ard What might help though:
1330542216 M * ard root@skipper:/home/ard# cat /etc/sysctl.d/min_free_kbytes.conf 
1330542216 M * ard vm.min_free_kbytes = 524288
1330542258 M * ard And on the apachephp servers with mysql vservers on the same iron:
1330542267 M * ard apache2 invoked oom-killer: gfp_mask=0x0, order=0, oom_adj=0, oom_score_adj=0
1330542280 M * ard Killed process 5939:#3107 (cron) total-vm:22472kB, anon-rss:252kB, file-rss:800kB
1330542285 M * ard And stuff like that
1330542296 M * ard just because they use php-imagemagick
1330542331 M * ard the database server however is pretty solid, and never has memory problems
1330542440 M * ard These are servers that went down on that upstream bug in 2.6.38.something
1330542553 M * ard I only got one problem, and that's that fusion-mpt occasionally rejects the vertex-2 SSD, and then never signals upstream (raid1) that the jobs failed, and hence raid1 cannot recover.
1330542583 A * ard guesses that's fixed in 3.2 series
1330542713 M * ard BTW: if your client is doing php with imagemagick, do this kind of magic in your apache config:
1330542720 M * ard export OMP_NUM_THREADS=1
1330542720 M * ard export OMP_MAX_THREADS=1
1330542728 M * ard AndrewLee: most important:
1330542732 M * ard export MAGIC_MEMORY_LIMIT=$((16*1024*1024))
1330542767 M * ard imagemagick has a memory limit of 2GB, so it will happily consume all of your not available memory before deciding to switch to memory mapped files
1330542790 M * ard these settings are in our config in /etc/apache2/envvars ;-).
1330542901 M * ard Another setting we do is that we have an IP-stack per DMZ. So your client would get it's own network namespace, it's own vlan device, and that vlan device directly goes to the switch.
1330542907 M * ard Firewalling is done by other systems
1330542995 Q * petzsch Quit: Leaving.
1330542996 M * ard Limit cpuset of the client, or the cpu shares. With a 3.2 kernel you can also force a maximum limit on allowed cpu-cycles
1330543094 M * ard ghislain : if you want any more info, I am happy to share my knowledge, but I assume you will put it in the wiki :-)
1330543125 M * ard I tried starting documenting how we setup vservers, but I usually stop after typing an intro :-(
1330544603 Q * Hunger Ping timeout: 480 seconds
1330544703 J * petzsch ~markus@dslb-188-103-057-014.pools.arcor-ip.net
1330548823 Q * Chlorek Ping timeout: 480 seconds
1330548991 J * Chlorek ~chlorek@2001:470:1f09:ae6::3ffe
1330549088 J * Hunger hunger@proactivesec.com
1330549902 Q * petzsch Quit: Leaving.
1330550746 Q * Hunger Remote host closed the connection
1330550913 M * ghislain ard: thx, i just came back i am playing with cgroup but whatever i put in /blkio.throttle.write_bps_device my dd is still goign at 450mbps
1330550952 M * ghislain but yes when i find something i put it in the wiki, this is me that created the cgroup page or the util vserver cheat sheet as i wanted to share what i found
1330551019 M * ghislain dd if=/dev/zero of=/var/cache/testfile_10MB bs=10485760 count=1
1330551025 M * ghislain 10485760 bytes (10 MB) copied, 0.0196074 s, 535 MB/s
1330551090 M * ghislain tried a bigger one: 524288000 bytes (524 MB) copied, 0.855258 s, 613 MB/s
1330551113 M * ghislain i have throttled to 2mbps all the physical disk, the 3 raid, the lvm also
1330551258 M * ghislain inside a guest stat give me no device:
1330551258 M * ghislain (root)> stat -c '%t %T' /var/cache/testfile_10MB
1330551259 M * ghislain 0 0
1330551340 M * ghislain a chcontext 1 stat give also 0 0 
1330551421 M * ghislain bertl: how can i ask for the blck devicee major/minor of a file inside a guest ?
1330551422 M * daniel_hozac you want to make sure it doesn't get buffered.
1330551436 M * daniel_hozac writing to memory will always be fast.
1330551447 Q * dowdle 
1330551456 M * daniel_hozac IIRC dd has a direct option for that.
1330551462 M * ghislain but i have 32go ram and only 8Go of free disk for the test
1330551468 M * Bertl ghislain: files do not have major/minors, but if you want to find the device the file is on, then stat should do
1330551500 M * ghislain ok
1330551526 M * daniel_hozac try dd if=/dev/zero of=/var/cache/testfile bs=1M count=100 direct fdatasync
1330551527 M * ghislain so i get Device: fd0eh/64782d  as a device humm does not look like the number i know lol, i need to find a way to translate this. let me google
1330551559 M * ghislain dd: unrecognized operand `direct'
1330551619 M * Bertl conv=direct ?
1330551635 M * daniel_hozac ah yes, it's conv=direct,fdatasync
1330551685 M * ghislain dam it: dd: invalid conversion: `direct'
1330551695 J * notau ~notau@gw-1.mel1.paranode.id.au
1330551701 M * Bertl ghislain: debian?
1330551702 M * daniel_hozac hmm
1330551706 M * daniel_hozac my bad
1330551711 M * ghislain bertl: of course
1330551718 M * daniel_hozac i think you actually want oflag=direct,dsync
1330551741 M * ghislain oh
1330551752 M * ghislain this one is stalled right now, lets wait a little
1330551895 M * ghislain ohoh ! 104857600 bytes (105 MB) copied, 126.585 s, 828 kB/s
1330551918 M * ghislain now i setup 2mbps per disk i go 800kbps but at least i AM throttled
1330551940 M * ghislain thanks a lot, now lets test by removing device per device until it throlle not
1330552179 M * ghislain ok the limit work with only the lvm
1330552206 M * ghislain i put 2mbps but it gives me 2.6 
1330552239 M * ghislain and 117 without limit
1330552276 Q * Chlorek Quit: -
1330552281 J * Chlorek chlorek@2001:470:1f09:ae6::3ffe
1330552306 Q * Chlorek 
1330552306 J * Chlorek chlorek@2001:470:1f09:ae6::3ffe
1330552397 M * ghislain ok, seems it is not 1024 multpl but 1000 to have the right one
1330552465 M * ghislain i get the upper limit, now i need to find a way to allow for bursting
1330552623 M * ghislain hum, does not seems to exist..yet
1330552968 M * Bertl I don't think I/O bursting is that important actually
1330553080 M * ghislain bertl: well if IO is not used by the other guest i would be ok for the running one to get more bandwidth than normaly
1330553112 M * ghislain bertl: but i can put an high upper limit and hope blk.weigth does the job
1330553115 M * Bertl that's not burst, that is 'idle time'
1330553137 M * ghislain oh sorry this is what i wanted to say, allow idle time to be used
1330553175 M * Bertl could probably be done with a hierarchical setup
1330553199 M * Bertl i.e. upper limit for all guests, separate limit/share for each one
1330553233 Q * guerby Ping timeout: 480 seconds
1330553273 J * guerby ~guerby@nc10d.tetaneutral.net
1330553440 M * ghislain hum, you mean 100mbps for all guest and then 30 for each so that when the sum goes higher than 100 as a whole the blkio.weight should handle it but permit a idle to up to 30 individualy ?
1330553459 M * Bertl for example
1330553473 M * ghislain yes just taking numbers as exemples
1330553519 M * ghislain the problem being to have a good upper layer limit
1330553534 M * ghislain agot to see if blkio support hierachical, i am not sure of that
1330553582 M * ghislain - Currently none of the IO control policy supports hierarhical groups. But 98	  cgroup interface does allow creation of hierarhical cgroups and internally 99	  IO policies treat them as flat hierarchy.
1330553658 M * ghislain will be in the future perhaps but  for now this flat even if you build it as a hierarchy
1330553706 J * Hunger hunger@proactivesec.com
1330554321 J * vn ~sys6x@nostalgeek.net
1330554349 M * vn hi, is fsck and e2fsck supposed to run normally on vserver-based setups?
1330554366 M * Bertl on the host, yes
1330554391 M * vn should I do it on the virtual group?
1330554406 M * Bertl 'virtual group' being?
1330554429 M * vn /dev/mapper/vg0-vserver_name
1330554443 M * vn sorry I'm not fluent with vserver and this setup is kinda clunky
1330554481 M * Bertl well, to the nomenclature: host = physical machine, guest = isolated domain
1330554493 M * vn says theres a problem with the ext2 superblock
1330554574 M * Bertl first, what kernel and Linux-VServer patch?
1330554601 M * vn uh yea...quite very old
1330554625 M * vn 2.6.17.8vs2.0.2-rl1smp #1 SMP Mon Aug 21 08:14:35 EDT 2006 i686
1330554652 M * Bertl okay, well, and who is complaining about the superblock?
1330554674 M * vn e2fsck
1330554692 M * Bertl on the host?
1330554695 M * vn yes
1330554707 M * Bertl well, then the filesystem is probably damaged ...
1330554713 A * vn facepalms
1330554724 M * vn I'm too used to recent server
1330554738 M * vn lets see how it works with reiserfs..
1330554865 M * Bertl may I ask why you are sticking to that quite old kernel?
1330554901 M * Bertl it's probably full of bugs already fixed over the years (centuries?)
1330554944 M * daniel_hozac maybe they're aiming for a decade old kernel?
1330554949 M * daniel_hozac that's not something you see every day.
1330554966 M * vn haha
1330554987 M * vn I am trying to get rid of it, but management you know...
1330554995 M * Bertl yeah, could be an aniversary kernel :)
1330555007 M * daniel_hozac maybe it gets a gold watch.
1330555072 M * vn they went for vserver with a consultant, then the consultant dropped support for openVZ
1330555100 M * vn no more repos, no more this and that, systeme was managed weirdly..
1330555134 M * vn it got a gold watch for the no-touch rule from management at least
1330555158 M * Bertl interesting ...
1330555187 Q * Roomster Ping timeout: 480 seconds
1330555191 M * Bertl well, if your company is interested in Linux-VServer consulting ... there are a bunch of folks here doing that :)
1330555247 M * vn ...maybe when the disks are gonna fail ehe
1330555262 M * vn for now, the recent architecture I fully have control on is vmware/ubuntu
1330555306 M * daniel_hozac ouch
1330555309 M * daniel_hozac i'm sorry.
1330555314 M * vn it's such a reward for me to be able to do anything from these archaic systems
1330555333 M * vn how sorry?  its a charm
1330555357 M * vn its a mix of Windows, Ubuntu, BSD
1330555370 M * Bertl my condolences as well :)
1330555418 M * Bertl anyway, back to your 'issue' .. if you run fsck on the host, there is no difference to an unpatched kernel (same version)
1330555440 M * Bertl i.e. if it fails to check the filesystem, then the filesystem is most likely broken/damaged
1330555444 M * vn yeap thanks, I ran reiserfsck, its now fixed
1330555454 M * vn finally booted properly
1330555461 M * daniel_hozac you ran reiserfsck on an ext2 filesystem?
1330555481 M * Bertl nah, obviously it is/was reiser in the first place
1330555487 M * vn it was reiserfs
1330555487 M * Bertl explains the e2fsck fail :)
1330555491 M * vn yup
1330555499 M * vn I'm too used to ext*
1330555602 M * vn well if you are sorry for me, I have a home server to build soon and it has to run VMs windows-based and linux-based from CLI, vserver seems like an option but I doubt it supports windows?
1330555634 M * Bertl Linux-VServer is on a different layer than e.g. vmware
1330555671 M * Bertl that's why you can run 1000 guests on a system which can barely maintain 5 virtual machines
1330555727 M * vn that, I can tell you, works like a charm
1330556918 Q * Hunger Ping timeout: 480 seconds
1330557134 M * ghislain daniel: i again forgot how to mount all cgroups of the guest in one parent cgroup, do you remember the parameters's name ?
1330557166 M * ghislain name is to group all in one cgroup and cannot remember the other
1330557501 Q * notau Quit: Computer has gone to sleep.
1330557629 M * ghislain is it "base" ?
1330557665 M * ghislain as this is not in the flower page i do not find it lol i am searching int the source code ^^
1330557753 J * Romster ~romster@202.168.100.149.dynamic.rev.eftel.com
1330557840 M * ghislain ok that's it
1330558164 M * ghislain daniel_hozac: if i move a cgroup from /ev/cgroup/xxx to /dev/cgroup/vservers/xxx after changin the base parameter will the tool recognize this or will they fails ?