1306886741 Q * lodan Ping timeout: 480 seconds
1306886970 Q * PowerKe Ping timeout: 480 seconds
1306887279 Q * imcsk8 Quit: Leaving
1306887503 J * PowerKe ~tom@94-226-105-27.access.telenet.be
1306887565 Q * ryker Quit: Leaving.
1306888750 Q * Piet_ Quit: Piet_
1306890561 J * ryker ~Adium@c-76-16-115-27.hsd1.in.comcast.net
1306890629 J * Piet ~Piet__@659AABYBP.tor-irc.dnsbl.oftc.net
1306891179 Q * ksn Quit: leaving
1306893506 J * lodan ~lodan@rb178-1-88-163-25-248.fbx.proxad.net
1306896655 Q * tty234 Remote host closed the connection
1306899631 M * Bertl off to bed now ... have a good one everyone!
1306899637 N * Bertl Bertl_zZ
1306902585 J * sannes ~ace@cm-84.209.81.224.getinternet.no
1306903790 J * bzed_ ~bzed@devel.recluse.de
1306903857 Q * bzed Ping timeout: 480 seconds
1306903861 N * bzed_ bzed
1306905009 J * petzsch ~markus@dslb-088-075-121-229.pools.arcor-ip.net
1306906635 J * ghislain ~AQUEOS@adsl2.aqueos.com
1306906805 Q * FireEgl Ping timeout: 480 seconds
1306908477 J * thierryp ~thierry@82.226.190.44
1306908881 J * ncopa ~ncopa@3.203.202.84.customer.cdi.no
1306909438 J * derjohn_mob aj@88.128.176.250
1306909999 M * sannes1 daniel_hozac: In the _generateCgroupOptions (vserver.functions) is the outer parenthesis in CGROUP_SUBSYS=( $($_SED '/^#/d;/^ns[[:space:]]/d;s/[[:space:]].*//' /proc/cgroups) ) intended?
1306910188 M * sannes1 In most setups (of my setups) that would end up only mounting with the cpu cgroup and not the memory one.
1306910254 Q * thierryp Remote host closed the connection
1306910379 Q * derjohn_mob Ping timeout: 480 seconds
1306913645 M * daniel_hozac sannes1: sounds like you have initscripts that don't match your util-vserer then.
1306913734 J * thierryp ~thierry@zanzibar.inria.fr
1306914016 J * FireEgl ~FireEgl@173-16-9-3.client.mchsi.com
1306914108 J * derjohn_mob ~aj@213.238.45.2
1306914933 M * daniel_hozac sannes1: are you using gentoo?
1306914947 M * sannes1 daniel_hozac: ah, yes, was too old version of util-vserver messing with me
1306915028 M * sannes1 daniel_hozac: I misinterpreted the vserver-stat messages where memory usage was not shown because of an already fixed bug in vserver-stat
1306915249 M * sannes1 daniel_hozac: And, yes, gentoo
1306915286 M * daniel_hozac using gentoo initscripts then? because those are lacking the CGROUP_SUBSYS fixes since... well, gentoo is difficult.
1306915341 M * sannes1 Actually I just put it into fstab, for some reason I thought vserver would mount cgroup filesystem if it was not mounted :P
1306915349 M * daniel_hozac it does
1306915362 M * daniel_hozac that is the point of the util-vserver initscript.
1306915464 M * sannes1 ah, hm, there is a util-vserver initscript .. hm
1306915626 M * sannes1 okay, in that case what does CGROUP_SUBSYS=( $(sed '/^#/d;/^ns[[:space:]]/d;s/[[:space:]].*//' /proc/cgroups) ) give you versus CGROUP_SUBSYS=$(sed '/^#/d;/^ns[[:space:]]/d;s/[[:space:]].*//' /proc/cgroups) 
1306915696 M * sannes1 For me the first one gives "cpu" and the second one "cpu cpuacct memory devices freezer"
1306915733 M * daniel_hozac it's an array
1306915778 M * sannes1 aha :P
1306916367 Q * SwenTjuln Read error: Connection reset by peer
1306916396 J * SwenTjuln ~SwenTjuln@193.164.140.194
1306916756 J * SwenTjuln_ ~SwenTjuln@77-111-2-36.ipv4.tusmobil.si
1306916786 Q * pmjdebruijn Quit: leaving
1306916879 Q * SwenTjuln Ping timeout: 480 seconds
1306916879 N * SwenTjuln_ SwenTjuln
1306917491 J * pmjdebruijn ~pascal@overlord.pcode.nl
1306918914 J * BenG ~bengreen@212.183.140.35
1306921567 Q * kwowt Ping timeout: 480 seconds
1306922163 N * Bertl_zZ Bertl
1306922167 M * Bertl morning folks!
1306922760 M * BenG morning
1306922784 M * BenG any plans for a 2.6.39 patch Bertl ?
1306922906 M * Bertl we'll see, definitely needs some work and proably even more testing
1306922939 M * Bertl changes affect CoW link breaking as well as most of the IP isolation
1306922958 M * Bertl (the rest I've already fixed up I think :)
1306923147 M * BenG top one
1306923147 M * BenG sounds like a big crop of changes though
1306923371 Q * BenG Quit: I Leave
1306923720 Q * thierryp Remote host closed the connection
1306923946 Q * petzsch Quit: Leaving.
1306927928 J * thierryp ~thierry@zanzibar.inria.fr
1306929343 N * zbyniu_ zbyniu
1306929824 M * thierryp hey there
1306929836 M * thierryp looking into fedora15 vservers...
1306929867 M * thierryp looks like a capability of some kind is needed for cap_set_file
1306929899 M * thierryp this is something I am running into e.g. when yum install'ing httpd
1306929905 M * thierryp and I guess it's all over the place
1306929962 M * Bertl yup, not implemented yet
1306930052 M * thierryp oh
1306930184 M * thierryp any known work around then ?
1306930229 M * Bertl do not use posix file capabilities inside a guest for now
1306930334 M * thierryp ok; any plans to support this ? 
1306930374 M * thierryp it's not as if I had the choice to use it or not, as rpm seems to use it
1306930464 M * Bertl it is on my todo list, feel free to implement and check it yourself and/or sponsor this particular feature
1306930759 M * thierryp fair enough
1306930943 M * Bertl but as I already stated, it shouldn't be too hard to add/fix
1306931042 M * thierryp any reflexion about which cap this should be based on ? some already existing one, or a new one ?
1306931071 M * Bertl CAP_SETFCAP IIRC
1306931447 M * daniel_hozac hmm
1306931457 M * daniel_hozac would there be anything bad by just letting guests have that?
1306931526 M * Bertl not sure about that, haven't investigated it yet
1306931573 M * daniel_hozac seems to just grant setting and removing the xattr.
1306931693 M * Bertl what about xattrs on unified files?
1306931707 M * daniel_hozac immutable should catch that, i would think...
1306932336 M * daniel_hozac yeah, setcap on a unified file gives me operation not permitted.
1306932530 M * daniel_hozac we could extend COW to cover that, i suppose.
1306932615 M * Bertl so no harm atm, but maybe rpm needs that as well
1306932633 M * Bertl although I doubt that it will change xattrs on exisiting files
1306932660 M * daniel_hozac right
1306932663 M * daniel_hozac it shouldn't.
1306932679 M * daniel_hozac dpkg is as far as i know the only package manager that changes existing files.
1306933364 M * thierryp well it sounds like rpm in f15 does try to mess with that, at least on some packages
1306933365 M * thierryp http://nopaste.info/95a8948863.html
1306933514 M * Bertl thierryp: so try to give CAP_SETFCAP to the guest(s) in question and let us know how it goes
1306934206 M * fback Bertl: ext3 is still preferred filesystem for guest (root) partition?
1306934229 M * Bertl ext2/3 yes
1306934362 M * fback Bertl: and no preference over something mounted inside a guest?
1306934442 M * Bertl there is no problem with other filesystems at all, it's just a question of stability and features supported
1306934492 M * Bertl i.e. reiserfs is fine if you do not care about the data, jfs is fine if you don't want quota ...
1306934869 M * fback Bertl: I'll move guests (root filesystems) between partitions (to unify them) soon, so I can set any type of fs for them
1306934942 M * fback Bertl: but there are other filesystems mounted inside guests (mail spool, /homes to name a few) and I'd rather not to touch them
1306935002 M * fback Bertl: you mentioned jfs -- is it preferred over xfs?
1306937148 J * ser_ ~ser@host1.tldp.ibiblio.org
1306937211 Q * ser Read error: Connection reset by peer
1306939358 J * dowdle ~dowdle@scott.coe.montana.edu
1306940272 Q * derjohn_mob Ping timeout: 480 seconds
1306940880 M * thierryp Bertl: sorry for the delay
1306940899 M * thierryp so SETFCAP actually works for me
1306941368 Q * pmjdebruijn Quit: Lost terminal
1306941593 M * ecapriolo I really like ext4 vs ext3. 
1306941643 M * ecapriolo I have found that with larger more full disks ext4 performs better. Also EXT4 handles delete operations in a nicer multi-user way.
1306941683 M * Bertl it also causes weird kernel hangs and sometimes panics
1306941729 M * Bertl I'm still hoping for btrfs advances, but it doesn't look like the developers take it seriously
1306941757 M * Bertl zfs on linux is making progress too :)
1306942777 N * Mr_Smoke_ Mr_Smoke
1306944253 J * pmjdebruijn ~pascal@overlord.pcode.nl
1306944392 M * ryker I see I can set shmmax for a guest on the fly with the command: vspace -e CONTEXTID --ipc sysctl -w kernel.shmmax=134217728
1306944392 M * ryker How can I check the shmmax value for a guest from the host?
1306944408 Q * ncopa Quit: Leaving
1306944682 M * Bertl by entering the ipc space
1306944803 M * ryker sry, i'm an idiot about that stuff.  That's what vspace does, right?  so I can use something like: vspace -e CONTEXTID —ipc sysctl -n kernel.shmmax ?
1306944835 M * Bertl yep
1306944836 Q * FireEgl Ping timeout: 480 seconds
1306944897 M * ryker thanks Bertl.
1306944942 M * ghislain i really hope btrfs will become stable, could be a real pleasure to have this one working
1306944944 M * Bertl you're welcome!
1306944993 J * bonbons ~bonbons@2001:960:7ab:0:e912:adae:e23a:bf5f
1306945014 M * ghislain i heard some zfs designer said that the design of btrfs was really simpler then zfs making it less prone to problem. seems features wise that it is not far zfs anyway
1306945023 M * ghislain s/then/than/
1306945033 Q * pmjdebruijn Quit: leaving
1306945067 J * pmjdebruijn ~pascal@overlord.pcode.nl
1306945088 Q * pmjdebruijn 
1306945121 J * pmjdebruijn ~pascal@overlord.pcode.nl
1306945136 M * fback Bertl: ext3 is the only fs I had disaster with
1306945149 M * fback Bertl: and data loss...
1306945202 M * Bertl really?
1306945212 Q * thierryp Remote host closed the connection
1306945254 M * Bertl it's interesting, because over the years I testd many filesystems for performance and stability, and almost always returned to ext3 because of the excellent stability
1306945285 M * fback after this I experimented with jfs, then turned to xfs -- I needed fs that stores lots of small files under single directory
1306945285 M * Bertl i.e. I never lost any data with ext3 and never had any 'real' problems with the filesystem on upgrades or similar
1306945334 M * fback and don't hit performance issues 
1306945380 M * Bertl are we talking about 10^3 or 10^6 or even more files?
1306945384 M * fback Bertl: otoh that was quite some time ago, and on SMP system, not common back then
1306945431 M * Bertl because ext3 with dir_index plays quite nicely up to 10^4 or so, then it gets ugly
1306945436 M * fback Bertl: definitely more than 10^3, maybe even 10^6 at some points
1306945466 M * Bertl but I'm not sure what I would need even 10^3 files in a single directory
1306945492 M * fback Bertl: mail spool with lots of spam comming in :)
1306945514 M * Bertl that's what hash and hash subdirectories are for
1306945565 M * fback Bertl: imapd of choice have to support it first :)
1306945585 M * Bertl what imapd do you use?
1306945618 M * fback cyrus
1306945644 M * fback Bertl: but filecount is not a problem anymore
1306945682 M * fback there's graylisting, spamassassin, and I'm not prowiding mail service to my friends anymore :)
1306945692 M * fback *providing
1306946053 M * fback ok, time to start the migration
1306946059 M * fback hope to back soon :)
1306946073 Q * fback Quit: leaving
1306946771 J * imcsk8 ~ichavero@148.229.1.11
1306946783 J * hijacker_ ~hijacker@87-126-142-51.btc-net.bg
1306947256 J * derjohn_mob aj@88.128.17.178
1306948920 J * tam ~tam@says.screwallofyoubitches.com
1306949693 M * ecapriolo We run Hadoop for distributed storage 1TB disks I notices performance looked bad past a certain point. Upgrading to EXT4 was less io and less ram used for VFS cache as well. 
1306949753 M * ecapriolo http://www.edwardcapriolo.com/roller/edwardcapriolo/entry/a_great_reason_to_use  
1306949793 M * ecapriolo I have found ext3 very reliable, its only problems for me were limits on files in a directory, and unlink performance being bad for large files.
1306950748 J * fb fback@red.fback.net
1306950753 M * fb :-)
1306950980 M * Bertl time for a nap ... bbl
1306950992 N * Bertl Bertl_zZ
1306952898 J * FireEgl FireEgl@2001:470:e056:1:e10f:d07b:59b:cc89
1306954708 J * Piet_ ~Piet__@659AABZIC.tor-irc.dnsbl.oftc.net
1306955114 Q * Piet Ping timeout: 480 seconds
1306956577 J * nicholi ~nicholi@12.232.116.66
1306957922 Q * derjohn_mob Ping timeout: 480 seconds
1306958995 Q * arekm Quit: leaving
1306959123 J * arekm ~arekm@ixion.pld-linux.org
1306959125 N * Bertl_zZ Bertl
1306959130 M * Bertl back now ...
1306959234 M * fb Bertl: seems I managed to migrate from debian-supplied to custom compiled util-vserver and move guest root filesystems from separate to common partition
1306959286 M * fb and from /var/lib/vservers to /vservers....
1306959319 M * fb now that I have newer utils I have to decide on the kernel
1306959342 M * fb which one you'd suggest, 2.6.38?
1306959575 M * Bertl yep
1306959598 M * Bertl but make sure that you fix up flags/tags when 'upgrading' from debian
1306959736 M * fb Bertl: I have custom kernel here, just utils from the distro
1306959763 M * Bertl ah okay then
1306959814 M * fb Bertl: is there a way to "measure" the gain from hashify?
1306959879 Q * sannes Remote host closed the connection
1306959917 M * Bertl yeah, do a du on each guest and a du -s on all of them (or use df)
1306960023 M * fb and common libraries?
1306960116 M * Bertl will be part of the difference
1306960213 M * fb i did: find -links 2 | wc -l
1306960224 M * fb then 3, then 4, and so on
1306960238 M * Bertl hmm, you can simply check with the .hash store for that :)
1306960253 M * fb and seems most files are shared only between .hash and one guest
1306960281 M * fb which is kind of strange, they are mostly identical squeeze guests
1306960283 M * Bertl means they differ in some way or are excluded from unification
1306960298 M * daniel_hozac make sure you disable prelink
1306960349 M * fb daniel_hozac: is there any distro beside ubuntu, that uses this by default?
1306960403 M * daniel_hozac fedora has been doing prelink for years...
1306960463 M * fb I don't have it on any guest installed
1306960495 M * daniel_hozac so run diff on the binaries
1306960542 M * fb these are just 4 lines so I let myself to paste it here ;-)
1306960543 M * fb atlas:/vservers/.hash# find . -links 2 | wc -l
1306960543 M * fb 42023
1306960543 M * fb atlas:/vservers/.hash# find . -links 3 | wc -l
1306960543 M * fb 6948
1306960573 M * daniel_hozac -type f
1306960591 M * fb I expected those numbers to be oposite
1306960643 M * fb 19600 vs 6900
1306960748 M * daniel_hozac check the files then
1306963052 Q * hijacker_ Quit: Leaving
1306964901 J * derjohn_mob ~aj@p4FFD16E3.dip.t-dialin.net
1306965464 M * areq_ samba could have some problems in vserver guest ? 
1306965497 M * areq_ I got Failed to open nmb bcast socket on interface
1306965767 M * Bertl did you assign the bcast to the guest?
1306965831 M * areq_ 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP qlen 1000 link/ether 00:1a:a0:11:1e:f9 brd ff:ff:ff:ff:ff:ff inet 10.105.80.3/28 brd 10.105.80.15 scope global secondary eth0
1306965839 M * areq_ in guest
1306965849 M * areq_ bind(13, {sa_family=AF_INET, sin_port=htons(137), sin_addr=inet_addr("10.105.80.15")}, 16) = -1 EADDRNOTAVAIL (Cannot assign requested address)
1306965863 M * Bertl yeah, but did you assign the 10.105.80.15 to the guest?
1306965868 M * areq_ # cat /etc/vservers/samba/interfaces/0/bcast
1306965868 M * areq_ 10.105.80.15
1306965924 M * daniel_hozac bcast isn't really used.
1306965934 M * daniel_hozac you should assign like a normal address.
1306965980 M * areq_ how ? 
1306965991 M * areq_ interface 1 ?
1306966111 M * areq_ nodev ?
1306966118 M * daniel_hozac yes.
1306966130 M * areq_ with prefix ? 
1306966144 M * areq_ 10.105.80.15/28 or 10.105.80.15
1306966163 M * daniel_hozac whichever.
1306966176 M * areq_ works, thanks
1306966354 M * Bertl you're welcome!
1306968072 Q * bonbons Quit: Leaving
1306969209 Q * ghislain Quit: Leaving.
1306969544 Q * derjohn_mob Ping timeout: 480 seconds
1306970897 N * ensc Guest3105
1306970907 J * ensc ~irc-ensc@p5DF2F3D4.dip.t-dialin.net
1306971170 Q * dowdle Remote host closed the connection
1306971319 Q * Guest3105 Ping timeout: 480 seconds
1306972018 J * tty234 telex@anapnea.net