1236643444 P * dwery 
1236643570 M * Bertl sladen: I'm off to bed now ... drop me a note if you obtained the id (thanks in advance)
1236643579 M * Bertl off to bed ... enjoy!
1236643584 N * Bertl Bertl_zZ
1236643973 M * sladen Bertl_zZ: 23:24 < s1aden> Bertl: okay, try that
1236643979 M * sladen Bertl_zZ: ('sladen')
1236644004 M * Plutotheplanet :)
1236644012 M * Plutotheplanet *:Pees*
1236644083 Q * duckx Remote host closed the connection
1236644238 J * ghislainocfs21 ~Ghislain@adsl2.aqueos.com
1236644671 Q * ghislainocfs2 Ping timeout: 480 seconds
1236644838 Q * bono Read error: Connection reset by peer
1236645097 J * bono bono@118-160-169-235.dynamic.hinet.net
1236647008 Q * Piet Quit: Piet
1236649403 J * ghislainocfs2 ~Ghislain@adsl2.aqueos.com
1236649730 Q * ghislainocfs21 Ping timeout: 480 seconds
1236650066 Q * Plutotheplanet Quit: http://www.mibbit.com ajax IRC Client
1236650422 Q * oo Read error: Connection reset by peer
1236650682 J * oo ~oo@64.28.12.166
1236653602 M * glen does vserver offer something for netlink security (opposed to /proc/net/tcp reading)?
1236653678 J * saulus_ ~saulus@c193004.adsl.hansenet.de
1236654059 J * derjohn_foo ~aj@e180207124.adsl.alicedsl.de
1236654070 Q * saulus Ping timeout: 480 seconds
1236654478 Q * aj__ Ping timeout: 480 seconds
1236655550 Q * balbir_ Read error: Connection reset by peer
1236656249 J * balbir_ ~balbir@122.172.110.122
1236659639 J * DreamerC_ ~DreamerC@122-116-181-118.HINET-IP.hinet.net
1236659639 Q * DreamerC Read error: Connection reset by peer
1236660044 Q * balbir_ Ping timeout: 480 seconds
1236660706 J * balbir_ ~balbir@122.172.108.1
1236661545 Q * balbir_ Ping timeout: 480 seconds
1236662145 J * Slydder1 ~chuck@dslb-088-074-053-171.pools.arcor-ip.net
1236662230 J * balbir_ ~balbir@122.172.151.62
1236662796 Q * Slydder1 Remote host closed the connection
1236662849 Q * balbir_ Ping timeout: 480 seconds
1236663552 J * balbir_ ~balbir@122.172.149.35
1236665922 Q * pmenier Quit: Konversation terminated!
1236668155 Q * balbir_ Read error: Connection reset by peer
1236668571 Q * thierryp Remote host closed the connection
1236668826 Q * hparker Quit: Read error: 104 (Peer reset by connection)
1236669187 J * balbir_ ~balbir@122.172.150.81
1236669512 J * pmenier ~pme@LNeuilly-152-22-72-5.w193-251.abo.wanadoo.fr
1236670089 J * yarihm ~yarihm@guest-docking-nat-1-110.ethz.ch
1236671329 J * duckx ~Duck@81.57.39.234
1236671484 Q * yarihm Quit: This computer has gone to sleep
1236671694 Q * balbir_ Read error: Operation timed out
1236672001 Q * duckx Quit: Client exiting
1236672171 J * dna ~dna@53-215-103-86.dynamic.dsl.tng.de
1236672357 J * davidkarban ~david@88.86.104.103
1236672415 J * balbir_ ~balbir@122.172.49.188
1236672440 J * sharkjaw ~gab@64.28.12.166
1236672601 Q * derjohn_foo Ping timeout: 480 seconds
1236672724 J * cga ~weechat@62.196.2.6
1236672997 J * thierryp ~thierry@zircon.inria.fr
1236673074 Q * balbir_ Read error: Connection reset by peer
1236673218 J * harobed_ ~harobed@pda57-1-82-231-115-1.fbx.proxad.net
1236673552 J * yarihm ~yarihm@whitehead2.nine.ch
1236673584 J * derjohn_foo ~aj@51.42.69.80.in-addr.net-lab.net
1236673942 Q * thierryp Read error: Connection timed out
1236674141 J * balbir_ ~balbir@122.172.110.138
1236674231 J * thierryp ~thierry@zircon.inria.fr
1236674541 Q * qyve Ping timeout: 480 seconds
1236674939 J * geb ~geb@79.82.4.157
1236675032 M * geb hi
1236675079 M * ghislainocfs2 hi
1236675176 Q * thierryp Read error: Connection timed out
1236675465 J * thierryp ~thierry@zircon.inria.fr
1236675590 J * kir ~kir@swsoft-msk-nat.sw.ru
1236675616 Q * ncopa Quit: Ex-Chat
1236675697 Q * harry Ping timeout: 480 seconds
1236675808 N * DoberMann[ZZZzzz] DoberMann
1236675836 N * Bertl_zZ Bertl_oO
1236675840 M * Bertl_oO morning folks!
1236676122 J * esa ~esa@ip-87-238-2-45.static.adsl.cheapnet.it
1236676128 M * fb morning Bertl!
1236676166 J * davidkarban_ ~david@193.85.217.71
1236676267 J * ktwilight_ ~ktwilight@91.178.146.65
1236676307 M * ghislainocfs2 morning bertl :)
1236676379 M * hijacker morning
1236676415 Q * thierryp Read error: Connection timed out
1236676475 Q * scientes Ping timeout: 480 seconds
1236676550 Q * ktwilight__ Ping timeout: 480 seconds
1236676565 Q * davidkarban Ping timeout: 480 seconds
1236676699 J * thierryp ~thierry@zircon.inria.fr
1236676713 Q * davidkarban_ Ping timeout: 480 seconds
1236676720 J * harry ~harry@d51A461B4.access.telenet.be
1236676955 J * davidkarban_ ~david@88.86.104.103
1236677199 J * davidkarban__ ~david@193.85.217.71
1236677587 Q * davidkarban_ Ping timeout: 480 seconds
1236677598 J * larsivi__ ~larsivi@70.84-48-63.nextgentel.com
1236677653 Q * thierryp Read error: Connection timed out
1236677806 Q * larsivi_ Ping timeout: 480 seconds
1236677933 J * thierryp ~thierry@zircon.inria.fr
1236678101 M * derjohn Bertl_oO, Oops: 0000 [#1] PREEMPT SMP <-- that one exists with and without the vserver patch. So, the vserver-patch is not the root cause of the problem.
1236678123 J * gnuk ~F404ror@pla93-3-82-240-11-251.fbx.proxad.net
1236678662 J * doener_ ~doener@i577BBE83.versanet.de
1236678765 Q * doener Ping timeout: 480 seconds
1236678883 Q * thierryp Read error: Connection timed out
1236679167 J * thierryp ~thierry@zircon.inria.fr
1236679441 J * Pazzo ~ugelt@reserved-225136.rol.raiffeisen.net
1236680159 J * Piet ~piet@asteria.debian.or.at
1236680278 Q * davidkarban__ Ping timeout: 480 seconds
1236680403 J * davidkarban__ ~david@88.86.104.103
1236680564 M * Bertl_oO derjohn: good to know, thanks for testing!
1236681123 M * Guy- I'm having problems with mounting in guests again... this time, I'm attempting this: vnamespace -e 78 mount -n /dev/encrypted/shared_ef2k_incoming /vservers/stop32/var/spool/ef2k_incoming -o nodev,nosuid,noexec,noatime
1236681141 M * Guy- it finishes and in vnamespace -e 78 cat /proc/mounts I can see the mount
1236681159 M * Guy- however, if I vserver stop32 enter, I can't, not even in /proc/mounts (and it is context 78)
1236681172 M * Guy- kernel 2.6.28.7-vs2.3.0.36.7
1236681250 M * Bertl_oO then it is a different namespace
1236681274 M * Guy- how can I find out which?
1236681287 M * Guy- I thought the context and namespace numbers normally matched
1236681308 M * Bertl_oO they are supposed to match
1236681332 M * Bertl_oO check /proc/self/nsproxy in both
1236681350 M * Bertl_oO the interesting space is the mnt space
1236681382 M * Guy- yes, they are different
1236681399 M * Guy- vnamespace -e 78 cat /proc/self/nsproxy has mnt:    ffff88006b1c3840(-)
1236681411 M * Guy- when I vserver enter, I get mnt:    ffff88006b1c3420(-)
1236681425 M * Bertl_oO what does vinfo say inside the guest?
1236681436 M * Bertl_oO cat /proc/self/vinfo that is
1236681446 M * Guy- XID:    78
1236681454 M * Guy- which line are you interested in?
1236681462 M * Bertl_oO are there more?
1236681464 M * Guy- yes
1236681471 M * Guy- BCaps:  00000000344c04ff
1236681471 M * Guy- this
1236681473 M * Guy- CCaps:  0000000000000101
1236681476 M * Guy- CFlags: 0000001402020010
1236681478 M * Guy- CIPid:  21487
1236681494 M * Bertl_oO okay
1236681507 M * Bertl_oO what util-vserver version?
1236681575 M * Guy- Debian sid's 0.30.216~r2772-6
1236681619 M * Bertl_oO should be fine, but please contact Daniel when he is around (or via email)
1236681645 M * Bertl_oO could be related to the double mnt namespace introduced in recent utils (utilizing pivot root)
1236681681 M * Guy- OK, so until I get a hold of him, I can't mount stuff in the vserver without restarting it, right?
1236681723 M * Bertl_oO if it is what I suspect, you need to make that mount shared
1236681736 M * Guy- what does that mean?
1236681741 M * Bertl_oO i.e. mount --make-shared /path/to/mount
1236681756 M * Bertl_oO (inside the namespace you enter with vspace
1236681778 M * Guy- let me try that
1236681816 M * Guy- no discernable effect
1236681924 M * Bertl_oO try to do a mount --make-slave /path inside the guest
1236682124 M * Guy- it complains about the path not being in fstab or mtab
1236682137 M * Guy- I can enter it in mtab manually, I suppose
1236682197 M * Guy- OK, now it says permission denied
1236682249 M * Bertl_oO for now, you probably need to either give more permissions or restart the guest with a proper fstab entry
1236682327 M * Guy- I gave it SYS_ADMIN
1236682340 M * Guy- now it says bad superblock on <the block device>
1236682342 M * Bertl_oO could be a bug/missing feature in the util-vserver version debian uses
1236682357 M * Bertl_oO (that it doesn't propagate by default)
1236682360 M * Guy- ioctl32(mount:11380): Unknown cmd fd(3) cmd(801c0204){t:02;sz:28} arg(ff8b9d4c) on /var/spool/ef2k_incoming
1236682365 M * Guy- I got this in dmesg
1236682391 M * Bertl_oO not from Linux-VServer 
1236682738 M * Guy- OK, thanks anyway
1236682825 M * Bertl_oO np
1236683079 J * mrfree ~mrfree@host1-89-static.40-88-b.business.telecomitalia.it
1236683131 Q * DreamerC_ Quit: leaving
1236683151 J * DreamerC ~DreamerC@122-116-181-118.HINET-IP.hinet.net
1236683226 M * ghislainocfs2 is there any debian packages of the last  vserver-tools ?
1236683242 M * ghislainocfs2 i try to create mine but i fail i am a loosy package maker
1236683708 Q * davidkarban__ Ping timeout: 480 seconds
1236684248 Q * aaronyy Quit: http://www.mibbit.com ajax IRC Client
1236684277 J * cluk ~cluk@p5B17F92C.dip.t-dialin.net
1236684298 Q * geb Remote host closed the connection
1236686035 Q * yarihm Quit: Leaving
1236686888 J * ntrs__ ~ntrs@77.29.9.55
1236686895 M * _Shiva_ i got a kernel module in the host that provides entries in /dev/this_module/* - i've copied the device nodes to the guest's /dev directory and started the guest... it is able to use them.. great! ;-)
1236686952 M * _Shiva_ but this module also provides an entry under /proc/ .. which is not visible to the guest however.. how do I make this visible to the guest? read-only would be enough
1236687160 M * fb vprocunhide?
1236687202 M * _Shiva_ fb: too easy :-) tnx :)
1236687511 M * _Shiva_ ah - so that would unhide this entry to any guest, right..? 
1236687560 M * _Shiva_ not particularly ony one guest specified  
1236687998 J * geb ~geb@79.82.4.157
1236688923 J * nou Chaton@causse.larzac.fr.eu.org
1236689006 J * BenG ~bengreen@94-169-110-10.cable.ubr22.aztw.blueyonder.co.uk
1236690043 Q * pmenier Read error: Connection reset by peer
1236690049 Q * ruskie Quit: Caught sigterm, terminating...
1236690050 J * pmenier ~pme@LNeuilly-152-22-72-5.w193-251.abo.wanadoo.fr
1236690114 Q * nou Read error: Connection reset by peer
1236690240 J * ruskie ruskie@ruskie.user.oftc.net
1236691907 J * uva ~bono@118-160-169-235.dynamic.hinet.net
1236692198 J * davidkarban ~david@199.123.broadband11.iol.cz
1236692277 Q * bono Ping timeout: 480 seconds
1236693886 Q * sharkjaw Quit: Leaving
1236695135 Q * DreamerC Ping timeout: 480 seconds
1236695191 J * DreamerC ~DreamerC@122-116-181-118.HINET-IP.hinet.net
1236695324 Q * Piet Remote host closed the connection
1236695444 J * Piet ~piet@asteria.debian.or.at
1236696331 J * ntrs_ ~ntrs@77.29.1.132
1236696499 M * Bertl_oO nap attack .. bbl
1236696503 N * Bertl_oO Bertl_zZ
1236696755 Q * ntrs__ Ping timeout: 480 seconds
1236697000 J * Piet_ ~piet@asteria.debian.or.at
1236697232 Q * Piet Ping timeout: 480 seconds
1236698757 Q * larsivi__ Remote host closed the connection
1236698809 Q * BenG Quit: I Leave
1236699282 J * hparker ~hparker@2001:470:1f0f:32c:290:96ff:fe50:40fa
1236699296 J * dowdle ~dowdle@scott.coe.montana.edu
1236699332 Q * thierryp Remote host closed the connection
1236701577 Q * davidkarban Ping timeout: 480 seconds
1236701587 J * davidkarban ~david@193.85.217.71
1236701812 Q * nenolod Quit: -> 2.6.29-rc7
1236702871 Q * davidkarban Quit: Ex-Chat
1236703551 Q * cga Quit: WeeChat 0.2.6
1236704048 M * nox any grsecurity + vserver startuphowto out there?
1236704069 M * sid3windr startuphowto?
1236704161 M * nox first steps and good hints :)
1236704165 J * nenolod nenolod@petrie.dereferenced.org
1236704893 J * bonbons ~bonbons@2001:960:7ab:0:2c0:9fff:fe2d:39d
1236705074 J * jpablo ~chatzilla@186.81.86.253
1236705123 J * esa` ~esa@ip-87-238-2-45.static.adsl.cheapnet.it
1236705416 Q * esa Ping timeout: 480 seconds
1236705745 Q * esa` Quit: Coyote finally caught me
1236705761 Q * kir Quit: Leaving.
1236705814 N * DoberMann DoberMann[PullA]
1236706236 J * Slydder ~chuck@dslb-088-074-053-171.pools.arcor-ip.net
1236706706 Q * Piet_ Quit: Piet_
1236707146 Q * Slydder Quit: Leaving.
1236707889 N * pmenier pmenier_off
1236708547 J * scientes ~scientes@75-165-95-28.tukw.qwest.net
1236709286 N * DoberMann[PullA] DoberMann
1236709608 N * Bertl_zZ Bertl
1236709614 M * Bertl back now ...
1236709638 M * Bertl nox: there should be a wiki page somewhere
1236709651 Q * harobed_ Quit: Ex-Chat
1236710326 J * dna_ ~dna@53-215-103-86.dynamic.dsl.tng.de
1236710571 J * cga ~weechat@94.36.110.25
1236710696 Q * dna Ping timeout: 480 seconds
1236711109 M * harry readme...
1236711262 M * nox Bertl: thx got queryhint from harry 
1236711816 Q * mrfree Ping timeout: 480 seconds
1236711895 J * hparker|laptop ~hparker@2001:470:1f0f:32c:290:96ff:fe50:40fa
1236711899 Q * hparker|laptop Read error: Connection reset by peer
1236712168 Q * jpablo Quit: ChatZilla 0.9.84 [Iceweasel 3.0.6/2009020407]
1236712588 Q * balbir_ Ping timeout: 480 seconds
1236713252 Q * gnuk Quit: NoFeature
1236713782 J * jmcaricand jm@121.10.199-77.rev.gaoland.net
1236714216 J * balbir_ ~balbir@122.172.63.167
1236714408 P * jmcaricand 
1236714508 J * Piet ~piet@asteria.debian.or.at
1236714598 J * puck ~puck@leibniz.catalyst.net.nz
1236714812 M * puck Hey, I'm running 2.3.0.36.4 how I enable loopback in the vservers?  I've added HIDE_LBACK and LBACK_REMAP into nflags, but there is no loopback, what else is needed?
1236715051 M * Bertl what is your guest config?
1236715108 M * puck I'm sorry, I don't understand the question.
1236715130 M * Bertl single ip, special casing enabled?
1236715178 M * puck Each vserver gets its own IP which is nat'ed via dummy0
1236715194 M * puck But this case I was hoping to have an IPv6 only vserver with a routable IPv6 address.
1236715200 M * Bertl upload /proc/virtual/<xid>/{info,status} and the output of 'vattribute --xid <xid> --get' to the pastebin please :)
1236715221 M * puck It turns out some apps I want to use are IPv4 only so I was hoping to add a IPv4 lookpack
1236715317 M * puck Bertl:  http://paste.dollyfish.net.nz/79479b
1236715531 M * puck Interestingly that doesn't show the contents of nflags
1236715551 M * Bertl yep, thus I also need the /proc/virtnet/<nid> contents
1236715579 M * Bertl and nattribute --nid <nid> --get
1236715654 M * puck http://paste.dollyfish.net.nz/dc717e
1236715673 M * puck So, it has assigned a loopback.  Do I need to create some files in interfaces/ ?
1236715712 M * Bertl no, you need to put a ~single_ip in your guest configs nflags
1236715717 Q * nenolod Quit: xen 3.4-unstable-dd489125a2e7 -> xen 3.4-unstable-b249f3e979a5
1236715735 A * puck tries
1236715752 M * puck didn't work
1236715763 M * Bertl and it might be necessary to assign the 127.0.7.1 IP as ipv4 too
1236715792 M * Bertl (the check for ipv4/ipv6 doesn't take the lback into account yet)
1236715812 M * Bertl so add an interface entry, with nodev and that ip
1236715841 M * puck Is that a file called nodev?
1236715868 M * Bertl touch a file 'nodev' and write the ip into the 'ip' file
1236715885 M * puck cool, that works.  Thank you!
1236715890 M * Bertl you're welcome!
1236716076 J * harobed ~harobed@arl57-1-82-231-110-14.fbx.proxad.net
1236717125 M * nox how can i ascertain the concurrent used AS?
1236717141 M * nox ... of a guest
1236717165 M * Bertl what is the meaning of this question?
1236717202 M * nox just reading http://linux-vserver.org/Resource_Limits and plays around with it
1236717212 M * nox rss is told by vserver-stat
1236717216 J * hijacker_ ~hijacker@87-126-142-51.btc-net.bg
1236717219 M * Bertl you want to get the current usage, yes?
1236717229 M * nox yes
1236717234 M * Bertl all the info is in /proc/virtual/<xid>/limits
1236717249 M * Bertl current/min/max
1236717268 M * nox ah that easy :)
1236717272 M * nox thx again
1236717400 M * Bertl np
1236717403 M * nox FILES mean max open filehandles?
1236717425 M * Bertl kernel side, yes
1236717469 Q * pflanze Quit: Leaving
1236717491 J * nenolod nenolod@petrie.dereferenced.org
1236717606 M * nox and DENT?
1236717711 M * Bertl http://linux-vserver.org/Resource_Limits
1236717753 M * nox aehm *blush*
1236717952 J * ntrs__ ~ntrs@77.29.2.28
1236718089 Q * hijacker_ Quit: Leaving
1236718287 Q * Pazzo Quit: ...
1236718389 Q * ntrs_ Ping timeout: 480 seconds
1236718701 Q * derjohn_foo Ping timeout: 480 seconds
1236719715 Q * DreamerC Ping timeout: 480 seconds
1236719763 Q * xdr Ping timeout: 480 seconds
1236719780 P * cluk Ex-Chat
1236719810 Q * cga Quit: WeeChat 0.2.6
1236719816 J * derjohn_foo ~aj@e180207124.adsl.alicedsl.de
1236719819 M * derjohn_foo Bertl, IIRC you did implement disk-io-usage based on the idea of atop into vserver. is that patch still in? I cannot find the IO data in /proc/virtual/....
1236719897 M * Bertl hmm? today I get the questions I cannot parse, it seems :)
1236720018 M * nox sry bertl bad rtfm error on my side :)
1236720204 J * DreamerC ~DreamerC@122-116-181-118.HINET-IP.hinet.net
1236720287 M * Bertl nox: no worries, happens ...
1236721076 J * cga ~weechat@94.36.110.25
1236721152 J * pippo ~pippo@host196-192-dynamic.16-79-r.retail.telecomitalia.it
1236721195 Q * cga 
1236721311 Q * pippo 
1236722182 J * hijacker_ ~hijacker@213.91.163.5
1236722182 Q * hijacker Read error: Connection reset by peer
1236723037 M * derjohn_foo Bertl, in short: where can I see the disk activitiy per guest?
1236723066 M * Bertl disk activity isn't accounted
1236723106 M * derjohn_foo But it was.... there was an patch intregrated IIRC ... was that lost or am I wrong?
1236723156 M * Bertl yep, there was a patch (this one: http://vserver.13thfloor.at/Experimental/delta-ioacc-feat01.diff)
1236723173 M * derjohn_foo ah, I never made it into devel ?
1236723216 M * Bertl wasn't tested very thoroughly
1236723357 M * derjohn_foo i'll check if it can be applied to current devel
1236723395 M * derjohn_foo  block/ll_rw_blk.c: No such file or directory
1236723402 M * _Shiva_ do you want to monitor or limit I/O?
1236723410 M * derjohn_foo _Shiva_, monitor
1236723427 M * derjohn_foo (to see which customer is the badass ;))
1236723457 M * _Shiva_ 'caus limiting could be done using CFQ and io[re]nice
1236723465 M * Bertl the problem is, you can only reliably pin down active/sync I/O
1236723486 M * derjohn_foo CFQ is default I think, but i iorenice is new to me. 
1236723499 M * derjohn_foo is that a recent kernel development ?
1236723512 M * mugwump sure, 2.6.16 or something
1236723515 M * Bertl there is no way to account/limit background and async I/O properly (yet)
1236723524 M * _Shiva_ the "re" was a wild guess :-) ionice exists though
1236723526 M * derjohn_foo Bertl, pin down = limit or account ?
1236723542 M * Bertl tie to the process :)
1236723586 M * derjohn_foo mugwump, well, 2.6.16 is around when Stonehendge was built ;)
1236723605 M * Bertl so brush up your kernel then :)
1236723617 M * derjohn_foo Bertl, if so, how does ionice work then ?
1236723621 M * _Shiva_ ionice is in sys-apps/util-linux
1236723648 M * Bertl ionice controls the active I/O only
1236723657 M * derjohn_foo Bertl, ionice also takes a pid as argument
1236723664 M * Bertl yep
1236723670 M * derjohn_foo ah, that way it works, but ex-post it doesnt ?
1236723697 M * derjohn_foo _Shiva_, but how I I ionice a whole guest (process group) ?
1236723705 M * _Shiva_ "Linux supports io scheduling priorities and classes since 2.6.13 with the CFQ io scheduler."
1236723745 M * derjohn_foo _Shiva_, ok,, but is there a context-aware ionice ?
1236723755 M * _Shiva_ derjohn_foo: i'm lookig it this is inherited.. if so, ionice to the guest's init would be it
1236723791 M * mugwump there's the new controller abstraction too
1236723806 M * Bertl which doesn't work yet
1236723839 M * Bertl but yes, I was hoping for that to get working/included so that I do not have to mess with the i/o queues
1236723846 M * derjohn_foo ah, but light at the end of the tunnel? (I dont mean the train comminh nearer :))
1236723890 M * mugwump there are some sticky issues with io scheduling ... especially when you look at trying to account all paging
1236723950 M * derjohn_foo is the paging activity to be accounted to the process? It's not the process fault that the machines doesnt have enough ram ?
1236723970 M * mugwump which process?  :)
1236723987 M * Bertl actually it's a lot more complicated
1236724026 M * derjohn_foo ok, the controller abstraction is not process related.
1236724039 M * derjohn_foo Bertl, well, yes, it sounds like ;)
1236724068 M * derjohn_foo what about _Shiva_'s idea to to ionice on the init process of a guest ?
1236724072 M * _Shiva_ "Now, one trick you might find handy is rescheduling an entire parent shell to idle class, so that all commands executed from that shell are idle priority."
1236724096 M * Bertl just consider disk read ahead, which might not even be used by any process :)
1236724097 M * _Shiva_ reads to me: it's inherited
1236724116 M * Bertl yep, it is, and yes, you can iorenice all processes in a guest
1236724180 M * derjohn_foo Bertl, what is the hand shell command for that? like ionice -c idle -p$(vps ...) ?
1236724181 M * Bertl it will work for the idle class, but it is not guest wide for the scheduling classes
1236724184 Q * geb Quit: Quitte
1236724252 M * _Shiva_ . o 0 ( # ionice -c3 -p$(pidof [kswapd]) ... swapping thrashes my disks... *ggg* )
1236724260 M * _Shiva_ scnr :-)
1236724311 M * cehteh hehe
1236724341 M * cehteh we have a 'devel' vserver which is started nice 11 and ionice -c3 ... works like a charm
1236724351 M * _Shiva_ ok - kernel threads don't have pids :-)
1236724371 M * cehteh but if it starts swapping you might consider adding more ram
1236724397 M * derjohn_foo cehteh, or limit memory for that guest.
1236724410 M * Bertl _Shiva_: sure about that? :)
1236724416 M * cehteh i also run 'swapspace' as swap daemon as insurance to against memory problems
1236724442 M * cehteh derjohn_foo: yes .. but unfortunally that 'devel' vserver is also the one where some very memory intensive tests are run
1236724447 M * _Shiva_ btw: /proc/kmem looks b0rked.. ls -lh /proc/kmem shows 4.7G.. with 4.0G memory in the machine..  
1236724460 M * cehteh and limiting memory in linux is not that trivial .. 
1236724473 M * _Shiva_ Bertl: not really - it shows a pit in ps/top ... ;-)
1236724485 M * _Shiva_ s/pit/pid/
1236724506 M * cehteh _Shiva_: PCI / IO address space, graphics card and other non existant addressable memory?
1236724518 M * cehteh DMA etc
1236724541 M * _Shiva_ cehteh: x86_64 with "memory hole remapping" ... may be th case
1236724541 M * Bertl not necessary, it's virtual address space
1236724562 M * cehteh Bertl: you mean limiting?
1236724575 M * Bertl no, I mean the kmem :)
1236724579 M * cehteh ah yes
1236724592 M * cehteh well any idea for 'useable' memory limiting ... 
1236724601 M * _Shiva_ graphic card is on-board - 128MB ded. memory and 128MB shared with ram..
1236724621 M * _Shiva_ still 500mb to find :-) 
1236724631 M * derjohn_foo cehteh, I use vservers approch to limit the guests memory (rss and as)
1236724642 M * _Shiva_ maybe my network card has 512MB buffers... *gg*
1236724674 Q * bonbons Quit: Leaving
1236724687 M * derjohn_foo _Shiva_, maybe that AGP aperture space ? (usually larger in BIOS than the memoy in the card is)
1236724692 M * cehteh derjohn: i dont know if vserver does some magic there .. but with vanilla kernel that wotn work very well and have some unexpected results
1236724710 M * _Shiva_ derjohn_foo: pci/pcie only - do not have agp
1236724725 M * cehteh limiting rss for example only accounts the brk point and not mmaped regions .. but glibc mmaps huge chunks
1236724729 M * _Shiva_ not even a slot on board :-)
1236724760 M * cehteh also limiting rss when there is still plenty of memory would provoke unnecessary swapping which is the most evil part
1236724784 M * _Shiva_ cehteh: set vm_swappiness=0 ? :-)
1236724796 M * mugwump not really.  leaves that memory free for other vservers when they need it
1236724816 M * cehteh on the other hand limiting as for applications which map huge address spaces but only use it infrequently if ever and then scan serially through it is far too limiting
1236724829 M * mugwump but I thought lowering the rss limit kills processes rather than makes them start to page
1236724830 M * cehteh _Shiva_: that doesnt disable swapping
1236724864 M * _Shiva_ cehteh: true - but it uses vfs-caches prior to swap
1236724882 M * Bertl swapping is the last thing you want in a low memory situation
1236724891 M * Bertl regardless of the guest or process
1236724898 M * cehteh mugwump: yes .. if a process needs more rss than granted by ulimits it gets killed ... 
1236724918 M * _Shiva_ Bertl: and the second thing you don't want to is having ooom-kill killing the "wrong" processes :-)
1236724924 M * mugwump ulimit doesn't have rss
1236724933 M * cehteh but the problem is it can still mmap more memory and glibc does so it circumvents rss limits in a clever way
1236724939 M * mugwump oh, yes it does
1236724944 M * _Shiva_ . o 0 ( like "postgres" - the big and evil memory hogger .. )
1236724964 A * mugwump uses Alt+SysRq+F to close Java apps
1236724985 M * cehteh while as limits somehow defeat the benefits you get from 64 bit machines ...
1236724991 M * Bertl cehteh, mugwump: depends on your setup, without overcommitment, rss will be correct
1236725025 M * cehteh some token-bucket memory limit might be useful to balance memory between all vservers
1236725053 M * cehteh Bertl: correct yes .. but even with overcommit it doesnt account for mmap .. which translates to swapping then
1236725060 M * cehteh without overcommit i meant
1236725081 M * cehteh only the memory below brk() point is accounted
1236725084 N * DoberMann DoberMann[ZZZzzz]
1236725114 M * Bertl mapped pages can be discarded anytime without a second thought (as long as they are read only) on write they will be accounted
1236725117 M * derjohn_foo cehteh, a token-bucket for memory? slow (freeze) the guest until enough memory get free?!?
1236725135 M * cehteh so an application cant brk() beyond the rss ... but it can still mmap ANON memory 
1236725137 M * mugwump no, free its mmap'd pages
1236725157 M * Bertl cehteh: it cannot mmap anon memory
1236725174 M * mugwump I think that keeping the totals of which rss is for which guest is hard though
1236725190 M * cehteh Bertl: is that vanilla or vserver?
1236725190 M * mugwump especially in the face of segments shared between guests
1236725215 M * Bertl cehteh: I do not remember vanilla having guest memory accounting
1236725225 M * cehteh some unixes send a SIGURG or so when memory gets low ... apps can then react on that
1236725231 M * cehteh not POSIX tough
1236725261 M * cehteh Bertl: i was speaking about vanilla kernel memory limits (ulimit)
1236725311 M * _Shiva_ would vunify'ing the guests help in any (little) way? it would at least limit rss to the guest's private mmap and share the common (application code) right?
1236725313 M * Bertl process ulimit (rss) should be accounted correctly with overcommitment disabled. i.e. you get it accounted when you map anon memory
1236725337 M * Bertl _Shiva_: unifying is a big memory saver
1236725347 M * cehteh and in our case where the 'devel' vserver rarely but intentionally needs much memory static limiting doesnt help much
1236725353 M * Bertl especially if the guests are similar
1236725375 M * cehteh anyways .. i saied 'works like a charm' ... even without any memory limiting here
1236725433 M * cehteh btw .. i have good experiences with:
1236725435 M * cehteh $ cat /proc/sys/vm/vfs_cache_pressure 
1236725435 M * cehteh 0
1236725440 M * _Shiva_ Bertl: yeah  - mostly they are all alike.. like 20-30 guests all Apache webserver containers with JeOS
1236725457 M * cehteh that saves a lot disk io 
1236725481 M * cehteh _Shiva_: you want unification then .. 
1236725487 M * Bertl _Shiva_: so unification would probably be a big gain, just be careful if you are using debian
1236725510 M * _Shiva_ i am.. :-) at least in guests - host is Gentoo :)
1236725519 M * cehteh Bertl: how that? .. i just use it without problems :)
1236725525 M * Bertl guest doesn't matter, I was referring to the kernel
1236725529 M * cehteh ok
1236725540 M * Bertl problem is that they messed up with the 2.6.26 kernel
1236725542 M * cehteh vserver utils and kernel self build :)
1236725545 M * _Shiva_ oh .. Kernel is vanilla+experimental :-)
1236725557 M * Bertl good, then you're fine :)
1236725608 M * cehteh _Shiva_: i have a some script to measure the disk space benefit from unifying .. which is really impressive
1236725609 M * _Shiva_ ... still util-vserver is Gentoo portage stable... (Hello Hollow, waiting for an overlay..  *g*)
1236725637 M * cehteh not done for maped files yet, if someone is bored one may try that :)
1236725712 M * _Shiva_ anyone attending LinuxTag this year? or even giving a talk..?
1236725755 M * nox cehteh: vfs_cache_pressure minimize the prefetch from disk?
1236725807 M * cehteh nox: no, prioritizing to keep inodes (dentry?) cached
1236725821 M * nox ic
1236725834 M * _Shiva_ cehteh: what's your FS? 
1236725859 A * nox would like to meat Bertl again on linuxtag :)
1236725862 M * nox meet
1236725863 M * cehteh so at the expense of some memory (up to few hundred MB, but still its a cache which gets eventually purged on need) your disk structure is in ram
1236725906 M * cehteh means on my laptop with a 1.8" 4200rpm disk i can do a du -sh / in 4 seconds
1236725922 M * cehteh (as long that cache is hot)
1236725940 M * cehteh _Shiva_: i just migrated to ext4 few weeks ago
1236725950 M * cehteh laptop and a server .. 
1236725959 M * _Shiva_ cehteh: ah nice :)
1236725977 M * cehteh otherwise ext3, xfs and maybe some reiser3 may lying around
1236725985 Q * ntrs__ Ping timeout: 480 seconds
1236725993 M * cehteh ext4 gives a very good feeling
1236725997 M * _Shiva_ cehteh: i think i'll stick to XFS for a while and the directly go over using BtrFS :-)
1236726032 M * cehteh xfs feels little un-gnuish/un-linuxish .. and some metadata operations are notoriously slow on it
1236726037 M * nox yeah butter sounds promising
1236726048 M * cehteh when you use git then xfs sux hard
1236726059 M * _Shiva_ cehteh: is growing on-line mainline for ext4? 
1236726075 M * cehteh i dont know .. dont need that
1236726114 M * _Shiva_ oh.. i use it all the time for my guests :-)
1236726127 M * cehteh unfortunally the 'no journal' patch didnt even made it to mainline .. 
1236726139 M * _Shiva_ every guest has it's own LV..
1236726139 M * cehteh i wanted to try it for dvd-ram :)
1236726169 M * cehteh then you cant unify them
1236726189 M * _Shiva_ but i have to rethink about that when vunify'ing... :-)
1236726202 M * _Shiva_ 2l8 :-)
1236726265 J * dna__ ~dna@53-215-103-86.dynamic.dsl.tng.de
1236726277 M * cehteh the times i made many filesystems are over .. i now just try a bare minimum in most cases
1236726338 M * cehteh one for thev root .. /boot isnt needed anymore on modern hardware .. one for /data which contains /var and /vserver and thats enough in most cases
1236726365 M * _Shiva_ i use reiserfs only on partitions that hold un-valuable data that have to be fast on metadata changes
1236726409 M * cehteh well reiser3 is reliable since ages when you dont try to do some raid/lvm stunts which involve moving images around
1236726461 M * _Shiva_ Quote: "Hans Reiser's wife is missing - mus be in his journal somewhere" *gg*
1236726466 M * cehteh i mostly gone away from it since support/future for it is somehow uncertain :)
1236726511 M * _Shiva_ or another quote: "reiserfs is fast, it's really fast, it's so fast it forgets about it's files"
1236726544 M * cehteh it really had some bad reputation at the beginning .. but i never lost files on one
1236726567 M * _Shiva_ and most customers complaining about crashed filesystems used standard SUSE-setups on reiserfs...
1236726579 M * cehteh and when i asked people why/how they lost files it was in most cases obvious that they did something they better shouldnt do
1236726618 M * cehteh crashed filesystems .. or bad recovery attempts?
1236726634 M * cehteh my guess the later is the main reason for data loss
1236726644 M * _Shiva_ like writing on reiserfs beyond 95%..? then files or directories simply vanish..
1236726661 Q * dna_ Ping timeout: 480 seconds
1236726676 M * cehteh when was that .. that must be fixed since ages
1236726686 M * mugwump I found a nice bug in reiserfs once
1236726690 M * _Shiva_ last time i looked reiserfsck had been a link to /bin/true...
1236726697 M * cehteh as saied in the beginning it was buggy 
1236726704 M * mugwump I was just running the 'cddump' program
1236726717 M * mugwump it makes a link to every file on the filesystem to 'spread' the backups over several CDs
1236726739 M * mugwump then unlinks them once the mkisofs run has finished, makes links for the next one...
1236726758 M * mugwump anyway, all that rapid linking and unlinking discovered some nasty reiserfs races
1236726774 M * mugwump you'd end up with dentry's you couldn't stat() & suchlike
1236726796 M * cehteh no doubts .. hans has always great plans and ideas .. but his implementations are flawed
1236726806 M * cehteh see reiserfs
1236726807 M * _Shiva_ mugwump: maybe it wasn't thread safe an you were doing this on a SMP machine? ;-)
1236726819 M * cehteh .. see how to get rid of a costly wife 
1236726835 M * mugwump yes, must be time for my William Tell routine
1236726860 M * cehteh reiserfs got eventually fixed .. the was fucked beyond repair
1236726889 M * mugwump the what?
1236726910 M * mugwump the wife?  being killed usually has that effect :)
1236727662 J * lurobep ~lurobep@193.43.249.169
1236727925 Q * harobed Ping timeout: 480 seconds
1236728067 J * dna ~dna@53-215-103-86.dynamic.dsl.tng.de
1236728386 J * Wonka_ produziert@chaos.in-kiel.de
1236728430 Q * dna__ Ping timeout: 480 seconds
1236728435 Q * Wonka Ping timeout: 480 seconds
1236728435 N * Wonka_ Wonka
1236729402 Q * dna Quit: Verlassend