1317254718 Q * fisted Read error: Operation timed out
1317254738 J * fisted ~fisted@xdsl-87-78-218-145.netcologne.de
1317255039 J * Rockj rockj@rockj.net
1317257889 Q * kir Quit: Leaving.
1317259851 M * Rockj Bertl: http://comments.gmane.org/gmane.linux.vserver/19269 , this seems to be the issues I really have
1317259863 M * Rockj at least I get it squashed just like this guy gets, hmmhm
1317260202 M * Bertl that could be related to ext4 attribute caching
1317260245 M * Bertl try to mount the nfs with noac first, if that works, please try this patch: http://vserver.13thfloor.at/ExperimentalT/delta-nfs-fix01.diff
1317260424 M * Rockj mount -oremount,noac /home ? right?
1317260452 M * Rockj mount -oremount,noac /home
1317260454 M * Rockj mount.nfs4: an incorrect mount option was specified
1317261014 M * Rockj hmf, have a feeling they dont do noac in 3.0.4 kernel 
1317261333 M * Rockj ill reboot it tomorrow I guess (4am atm) and take a dig
1317261344 M * Rockj try the old kernel to make sure it is the newest kernel with patch that does it etc
1317262045 M * Bertl not sure it is kernel related, I'd opt for an older/outdated mount tool
1317270563 J * sannes ~ace@cm-84.209.106.118.getinternet.no
1317273928 J * click_ click@ice.vcon.no
1317274182 Q * click Read error: Connection reset by peer
1317274182 Q * FireEgl Read error: Connection reset by peer
1317274194 M * Bertl off to bed now ... have a good one everyone!
1317274199 N * Bertl Bertl_zZ
1317274774 J * FireEgl ~FireEgl@173-16-9-169.client.mchsi.com
1317275780 J * ghislain ~AQUEOS@adsl2.aqueos.com
1317275801 Q * ghislain 
1317276039 J * derjohn_mob ~aj@213.238.45.2
1317278653 J * ghislain ~AQUEOS@adsl2.aqueos.com
1317281772 N * zbyniu_ zbyniu
1317283639 J * thierryp ~thierry@zankai.inria.fr
1317288524 M * ghislain ohoh i see the 127.x.x.x address in my guest instead of 127.0.0.1 :)
1317288579 M * ghislain i put only quota_ctl in capacilities so seems strange
1317288645 M * ghislain hum i see a difference in bcaps
1317289075 M * ghislain 0.30.216-pre2987, does it put bcaps by default ? i have BCaps:  ffffffffb44c04ff
1317289087 M * ghislain my old tools sets: BCaps:  00000000344c04ff
1317289215 M * ghislain i have those bcaps set that are not in the old ones:MAC_OVERRIDE,MAC_ADMIN,^34,^35,^36,^37,^38,^39,^40,^41,^42,^43,^44,^45,^46,^47,^48,^49,^50,^51,^52,^53,^54,^55,^56,^57,^58,^59,^60,^61,^62,^63
1317289242 M * ghislain do any of you use a 3.0.4 patch and see the 127.0.0.1 replaced by 127.X.X.X in the guest ?
1317289317 M * ghislain i do not see the MAC_OVERRIDE,MAC_ADMIN in http://linux-vserver.org/Capabilities_and_Flags so i don't know what it does
1317289911 M * ghislain http://paste.linux-vserver.org/20609 for details
1317289919 Q * zbyniu Read error: Connection reset by peer
1317291771 M * daniel_hozac and nattribute --nid <guest> --get?
1317293330 N * Bertl_zZ Bertl
1317293335 M * Bertl morning folks!
1317294189 Q * thierryp Remote host closed the connection
1317296840 J * zbyniu ~zbyniu@ip-62.181.188.13.static.crowley.pl
1317297103 J * uxbod ~uxbod@r-42.cust-3650.ip.static.uno.uk.net
1317297119 M * uxbod Hello all
1317297228 M * Bertl hello uxbod!
1317297254 M * uxbod We have migrated our vserver guests from one machine to another
1317297267 M * uxbod and have probably taken across the .hash incorrectly
1317297277 M * uxbod what is the correct procedure please for re-hashification
1317297280 M * uxbod ?
1317297339 M * Bertl basically remove all the existing files in the hash and rerun vhashify
1317297362 M * Bertl (the cleanup is not strictly necessary but helps to get rid of wrong/useless data)
1317297395 M * uxbod so go into the .hash direcory and do a rm -rf * ?
1317297488 Q * Aiken Remote host closed the connection
1317297545 M * uxbod Bertl: our hasher script does have at the end /usr/bin/find /vservers/.hash -type f -links 1 -print0 | xargs -0 rm -f 
1317297559 M * uxbod so I guess I can hashify each vserver again and then let that run
1317297568 M * uxbod is that correct please ?
1317297588 M * Bertl assuming that they are not cross linked to files outside, that will work
1317297628 M * Bertl i.e. what could go wrong (and why I suggest that you clean it up once) is for example if you have .hash and .hash2 both containing the same/linked files
1317297648 M * Bertl in this case, the link count will be 2, despite the fact that they are not used by any guest
1317297689 M * Bertl note: in any case, it doesn't hurt if you remove the .hash contents completely
1317297704 M * uxbod ok. so completely clear out the .hash and then re-hashify
1317297725 M * Bertl that's usually the simplest way
1317297747 M * uxbod no issues with a  rm -rf .hash ?
1317297763 M * Bertl you might need to re-create the dir, not sure
1317297797 M * uxbod ok, will go into the .hash directory and clear the contents from there leaving .hash in place
1317297800 M * Bertl but also note: it's not necessary to remove it on a regular basis, just when you think you got a bad copy or similar
1317297834 M * Bertl i.e. if everything is fine, the find -links 1 will successfully remove unused files
1317298023 M * uxbod thank you Bertl
1317298033 M * Bertl you're welcome!
1317298077 Q * uxbod Quit: Ex-Chat
1317302631 M * markc hey Bertl, you have got to be one of the nicest and most helpful project leads I've ever come across... just want to say thanks for all your effort
1317302832 M * Bertl thank you for the flowers, appreciated! and thanks for using? Linux-VServer!
1317303072 M * markc enjoying vserver so much I am determined to get a web interface for it, make it easier to deploy etc
1317303227 M * mikez there is one openvcp
1317303503 J * clopez ~clopez@155.99.117.91.static.mundo-r.com
1317303885 M * markc mikez, I'm not sure it'll work with the latest verser, I couldn't get it to work
1317303935 M * markc vserver*
1317304077 M * markc I want something lighter, even html5 only, single page ajax app etc
1317304120 M * Bertl sounds good, maybe collect some ideas here on the channel or via the ML for it?
1317304271 M * markc eventually, yes, but "show me the code" first... my first cut should be on Github in a week, I hope
1317304335 M * Bertl nice, let me know if there is something to try out (would probably speed up things if there were some kind of test server one can play with
1317304339 M * markc it'll be really simple to start with and absolutely no attempt to be IE friendly
1317304373 M * daniel_hozac does it work with links?
1317304389 M * markc no chance, html5 only
1317304446 M * daniel_hozac my point exactly... :)
1317304470 M * Bertl well, if it conforms to standards (e.g. html5) it doesn't matter IMHO ... i.e. conforming browsers will work ...
1317304506 M * markc I was going to do a websocket frontend with a nodejs backend but my node-fu has been fully stretched, so back to PHP
1317304554 M * Bertl designing for links or lynx is the same as designing for IE, the important part is to use standards and to ensure that impaired people can use it as well
1317304582 M * Bertl of course, testing with a text based browser is a good thing IMHO
1317304620 M * Bertl (shows a lot of design flaws, also note: e-links can do a lot of things :)
1317305014 M * markc I've started with and refactoring https://github.com/freezmeinster/vmin.git by removing the codeigniter dependency, but keeping the PHP shell_exec(sudo something) execution concept
1317305193 M * mikez markc: would be great to see something new
1317305221 M * mikez also think about limits quotas etc
1317305423 M * ghislain daniel_hozack: ncapabilities:
1317305423 M * ghislain raw_icmp
1317305423 M * ghislain nflags:
1317305423 M * ghislain lback_remap,hide_netif,hide_lback,state_admin
1317305430 M * ghislain oups sorry
1317305466 M * Bertl looks good, and you see 127.x.y.z inside the guest?
1317305473 M * ghislain yes i am
1317305478 M * Bertl did you explicitely assign that address to the guest?
1317305497 M * ghislain nope
1317305509 M * Bertl what is the kernel/patch version?
1317305524 M * ghislain i only have one ip and it it the one of the host whit nodev
1317305563 M * ghislain patch-3.0.4-vs2.3.1-pre10.1.diff on 3.0.4
1317305578 M * ghislain with delta-dev-fix01.diff too
1317305596 M * Bertl could you upload the contents of /proc/virtnet/<nid>/info to some pastebin?
1317305696 M * ghislain http://paste.linux-vserver.org/20610
1317305817 M * Bertl and what is the tool you use to show the lback address?
1317305831 M * Bertl i.e. what command shows the 127.x.y.z?
1317306034 M * ghislain netstat -nlp
1317306074 M * Bertl ah, 'ip a l' doesn't show it, yes?
1317306113 M * ghislain no i got 127.0.0.1
1317306124 M * Bertl okay, let me check that
1317306642 M * Bertl okay, I can confirm this (i.e. I can reproduce it here) let me see where the mapping is missing 
1317306835 M * Bertl seems to leak from /proc/net/* checking now ...
1317307105 M * ghislain oky ^^
1317307145 M * ghislain when i use dig on 127.0.0.1 it works so this is limited to output
1317307189 M * Bertl yeah, it's cosmetical, but easy to fix
1317307537 M * Bertl test compiling now ...
1317307736 M * Bertl http://vserver.13thfloor.at/ExperimentalT/delta-lback-fix05.diff
1317307742 M * Bertl that should fix the issue :)
1317308538 Q * ncopa Quit: Leaving
1317310046 J * dowdle ~dowdle@scott.coe.montana.edu
1317311772 M * ghislain ok i need to revert any patch ot just add this one ?
1317311876 M * ghislain ok patch applied, compilation is running
1317312011 M * Bertl just apply, shouldn't conflict with anything you have
1317313378 M * Rockj Bertl: I booted up an older kernel
1317313404 M * Rockj not sure if it's vserver patched, but then my ssh keys worked flawless
1317313415 M * Rockj 2.6.35.4-UAC
1317313424 M * Rockj so I guess it is releated to the vserver patches and nfs? 
1317313453 M * Rockj Bertl: ill try to see if I can apply this patch at the 3.0.4 kernel
1317313455 M * Rockj http://vserver.13thfloor.at/ExperimentalT/delta-nfs-fix01.diff
1317313656 M * Bertl yes, please do so, it should apply just fine
1317313680 M * Rockj compiling kernel now :-)
1317313875 J * bonbons ~bonbons@2001:960:7ab:0:ecb0:a84f:8ce1:b0ec
1317314389 M * Rockj Bertl: patch didn't work :/
1317314458 M * Bertl okay, first let's check what the mount actually is, i.e. upload the output of 'cat /proc/mounts' to a pastebin
1317314480 M * Bertl (on the host or in the guest namespace if it is mounted via fstab.remote)
1317314829 M * Rockj ok, give me a sec - rebooted server again to fix some local mount binds that obiously were wrong
1317315076 M * Rockj Bertl: http://pastie.org/private/hesuhek0zjayzi2dvnyrw
1317315083 M * Rockj (it's on the host)
1317315144 Q * fisted Read error: Connection reset by peer
1317315175 M * Bertl okay, can you - just for a test - try with vers=3 ?
1317315188 M * Bertl i.e. mount it as nfs v3 instead of v4?
1317315215 M * Rockj Bertl: I guess, 2 sec
1317315231 M * Bertl np, take your time
1317315256 J * petzsch ~markus@p57B647D6.dip.t-dialin.net
1317315396 M * Rockj hm, access denied while monuting
1317315405 M * Rockj probably have to enable nfs3 on server or something
1317315457 M * Rockj mounting with nfs3 works Bertl 
1317315459 M * Rockj :O
1317315490 M * Bertl okay, i.e. that makes your ssh stuff work as well?
1317315503 M * Rockj yes . 
1317315518 M * Bertl good, so it is, as I suspected, an nfsv4 related issue
1317315561 M * Bertl I would have suspected the attribute caching we fixed with that patch (or at least assumed to have fixed with that patch) 
1317315591 M * Bertl so it might really be worth trying to disable the attribute caching for nfsv4 with the 'noac' mount option
1317315632 M * Rockj but enabling noac makes horrible speed on the /home mount 
1317315648 M * Rockj also I didnt manage to mount with noac with latest kernel :O
1317315652 M * Rockj said unknown option
1317315657 M * Rockj which I found a bit odd
1317315659 M * Bertl it's not intended as solution, I'm just trying to find out where to look
1317315662 M * Rockj (as ive done it before)
1317315679 M * Bertl IIRC, you tried with a remount, which probably fails
1317315683 M * Rockj mount -onoac,remount /home should do it with nfs4 
1317315685 M * Rockj oh
1317315690 M * Rockj remount doesnt work?
1317315694 M * Bertl i.e. try to _mount_ it with noac in the first place
1317315700 M * Rockj let me just set noac then
1317315702 M * Rockj in fstab
1317315704 M * Rockj umount here and try
1317315706 M * Rockj with nfs4
1317315742 M * Rockj fernet.int:/home   /home   nfs4    _netdev,auto,noac    0   0
1317315745 M * Rockj in fstab I guess
1317315764 M * Bertl what's the _netdev?
1317315789 M * Rockj makes sure it tries to mount the drive after the ethernet devices are up
1317315792 M * Rockj so it doesn't hang ;)
1317315814 J * fisted ~fisted@xdsl-87-78-210-32.netcologne.de
1317315817 M * Bertl hum, guess that is something distro specific
1317315833 M * Bertl anyway, doesn't really matter here, rest is fine
1317315960 M * Rockj Bertl: noac works btw
1317315984 M * Bertl so it soleves your issue as well, yes?
1317315988 M * Bertl *solves
1317316020 Q * derjohn_mob Ping timeout: 480 seconds
1317316022 M * Rockj yes, it didnd't squash (nfs_nobody) on my authorized_files while sshing in
1317316029 M * Rockj but its not acceptable to run it with noac tho
1317316053 M * Bertl okay, could you please verify that you applied the nfs-fix patch, rebuilt _and_ booted the correct kernel?
1317316115 M * Rockj easy way to check it?
1317316138 M * Bertl yes, we had a test case, sec
1317316201 M * Rockj at least can find a line from the diff in my kernel source
1317316207 M * Rockj let me do a make clean and fully recompile it
1317316213 M * Rockj maybe it screwerd it up that it was just a partly build
1317316214 M * Bertl basically pick a file you haven't used yet, on nfs, inside a guest
1317316230 M * Bertl then do 'cat /path/to/file; ls -la /path/to/file'
1317316233 M * Rockj so gotto make a guest? :p havent made those yet - heh
1317316240 M * Rockj it happens on the host
1317316253 M * Bertl it should work on the host as well, just not verified
1317316267 M * Bertl basically the ls -la should give wrong uid/gid
1317316294 M * Rockj it doesnt. 
1317316313 M * Rockj norangsh@absint:~$  ls -la .ssh/authorized_keys 
1317316315 M * Rockj -rw------- 1 4294967294 4294967294 1773 Sep 12 13:27 .ssh/authorized_keys
1317316327 M * Rockj if I leave an ssh open in another terminal and prompting for its password
1317316349 M * Rockj if I logout / cancel/ login or whatever, it fixes it permissions again afterwards
1317316351 M * Rockj which is kinda off
1317316364 M * Rockj (sometimes takes little time, sometimes as much as 8 -10 seconds) 
1317316368 M * Rockj from what Ive seen
1317316495 M * Rockj makes any sense? 
1317316962 M * Rockj *waiting for a clean kernel compile*
1317317298 M * Rockj ok, rebooting now
1317317303 M * Rockj clean kernel boot - patch applied
1317317307 M * Rockj double checked
1317317313 M * Rockj by reading source :P
1317317648 A * Rockj wonders if Bertl died :p
1317317695 M * Rockj I assume I'll have to use nfs3 for now until it is fixed? its better then nfs4 noac I would assume? 
1317317783 Q * petzsch Quit: Leaving.
1317318618 M * Bertl not dead (yet)
1317318656 M * Bertl well, the question is what actually happens and why the proposed patch doesn't fix it (for you)
1317318693 M * Bertl 4294967294 = -2, so it is definitely the caching default we saw yesterday
1317318735 M * Bertl I don't have an nfs v4 server to test with (yet), as I do not consider nfs v4 production ready
1317319099 M * Bertl you might want to enable nfsv4 debugging, I see a bunch of useful dprintk() outputs in fs/nfs/nfs4*.c
1317319123 M * Bertl specifically the fs/nfs/nfs4xdr.c seems to be a good place to start
1317319138 J * sweil ~stefan@p54ADB514.dip.t-dialin.net
1317319194 M * Bertl have to take a nap now ... bbl
1317319201 N * Bertl Bertl_zZ
1317319608 J * petzsch ~markus@p57B647D6.dip.t-dialin.net
1317321632 Q * petzsch Quit: Leaving.
1317321660 M * Rockj Bertl_zZ: this is interessting. Currentl using noac, but im simply writing a 16GB file over the nfs4 with noac ... and I mass spam ls -la on the file, owner rights is right most of the time, without a few times it seems to reset it self to nfs-nnodbody (getting squashed)
1317321664 M * Rockj and then im only doing :
1317321684 M * Rockj first cmd lined out in http://it.toolbox.com/blogs/database-soup/testing-disk-speed-the-dd-test-31069
1317321687 J * petzsch ~markus@p57B647D6.dip.t-dialin.net
1317322601 J * ghislain1 ~AQUEOS@adsl2.aqueos.com
1317322999 Q * ghislain Ping timeout: 480 seconds
1317323160 Q * petzsch Quit: Leaving.
1317323416 J * petzsch ~markus@p57B647D6.dip.t-dialin.net
1317324247 Q * sweil Ping timeout: 480 seconds
1317324471 Q * clopez Ping timeout: 480 seconds
1317324888 J * kshannon_ ~kris@122.252.14.166
1317324999 Q * kshannon Ping timeout: 480 seconds
1317325308 J * sweil ~stefan@p5086EF3D.dip.t-dialin.net
1317326260 Q * _nono_ resistance.oftc.net charm.oftc.net
1317326260 Q * jrayhawk resistance.oftc.net charm.oftc.net
1317326322 J * _nono_ ~gomes@licencieux.ircam.fr
1317326322 J * jrayhawk ~jrayhawk@nursie.omgwallhack.org
1317326666 Q * sannes Remote host closed the connection
1317326846 J * jrayhawk_ ~jrayhawk@nursie.omgwallhack.org
1317326871 Q * jrayhawk resistance.oftc.net charm.oftc.net
1317326871 Q * _nono_ resistance.oftc.net charm.oftc.net
1317326997 J * _nono_ ~gomes@licencieux.ircam.fr
1317327080 J * Aiken ~Aiken@2001:44b8:2168:1000:21f:d0ff:fed6:d63f
1317327628 Q * ntrs Ping timeout: 480 seconds
1317328103 J * ntrs ~ntrs@vault08.rosehosting.com
1317328225 Q * ghislain1 Quit: Leaving.
1317328510 N * Bertl_zZ Bertl
1317328513 M * Bertl back now ...
1317328565 M * Bertl Rockj: interesting indeed
1317329053 Q * markc Ping timeout: 480 seconds
1317329073 Q * bonbons Quit: Leaving
1317329186 J * markc ~markc@203.25.132.186
1317329363 M * Rockj Bertl: but do you think vserver is the cause of it? since it works with a 2.6.35 kernel I have laying around? 
1317329368 M * Rockj or could it be 3.x issue? 
1317329378 M * Rockj tho, 3.x shouldn't be more then 2.6.38 or 39 or what it is
1317329453 M * Bertl well, probably it is Linux-VServer related, but not necessarily, so it could as well be a completely unrealted 3.x issue ...
1317329476 M * Bertl 2.6.38 is more like 2.6.35, where 2.6.39 is more like 3.x
1317329834 M * Bertl I'm pretty confident we can (and will) track this down, but atm I do not have an nfsv4 test setup
1317329933 M * Rockj can I give you a guest to play with?
1317329935 M * Rockj does that help?
1317329936 M * petzsch currently moving from 2.6.32 vserver kernel to 3.0.4: any idea why running reboot -f within the vserver causes it to enter "stopped" state while the vservers ip stays pingable opose running vserver <name> stop also leaving the ip unpingable
1317329978 M * Bertl Rockj: no, guest doesn't help, but if you have a test system (to test with) you could do most of the tests for me I guess ...
1317330013 M * Bertl petzsch: sounds like you have no proper helper (util-vserver) configured?
1317330018 M * Rockj it's our private server which we plays at. the others get a bit unhappy since I reboot and kill their screen's with irssi/weechat on tho - but wth
1317330026 M * Rockj I'd rather help with solving a bug
1317330028 M * Rockj :D
1317330040 M * Rockj hm
1317330044 M * Rockj have another server, lets mount it up there
1317330048 M * Rockj and see if it maybe has the same issues
1317330053 M * Rockj if so, I can give you root on that one 
1317330098 M * Bertl as I said, access from my side is mostly unnecessary, the important part is that we can test small changes to the kernel rapidly
1317330121 M * Bertl i.e. I provide a patch/change, you build a kernel and run/test it
1317330148 M * Bertl debug output/etc goes to some web server or pastebin ...
1317330164 M * Rockj Bertl: aye, I can do . Ill just notify the others in techiegrp at student union that there will be some reboots in the incoming days while we sort out the nfs bug
1317330168 M * Rockj as im happy to solve it
1317330170 M * Rockj :)
1317330195 M * Bertl go ahead .. first step is to get the nfsv4 debugging up and running
1317330253 M * Bertl i.e. we want to see messages like the one on line 894 in fs/nfs/nfs4xdr.c
1317330265 M * petzsch Bertl: any documentation on that helper (util-vserver) you mentioned? haven't modified the halt/reboot guest binaries since i think vserver 1.2 or so
1317330287 M * Bertl is your util-vserver recent enough?
1317330305 M * petzsch 0.30.216-pre2990-1
1317330365 M * petzsch can i somehow monitor the vserver while issuing the reboot -f from within the vserver?
1317330395 M * Bertl to some degree, with the Linux-VServer debugging system enabled
1317330474 M * petzsch the guests are built with the debootstrap method, any modifications needed for reboots to work as intended?
1317330503 M * Bertl basically the reboot -f should invoke the /proc/sys/kernel/vshelper
1317330524 M * petzsch i'll look into the guest if the proc entry is present
1317330534 M * Bertl it's on the host
1317330550 M * Bertl you can, for example, wrap that with a shell script
1317330567 M * Bertl i.e. use a different helper in /usr/local or so
1317330567 M * petzsch seems /proc/sys/vshelper is prsent
1317330579 M * petzsch on the host that is
1317330684 M * petzsch could you verify that on your system a reboot -f works as expected?
1317330735 M * Bertl in a few minutes, yes
1317330770 M * petzsch i've tried it with a wheezy and a squeeze guest with the same result
1317330806 M * Rockj Bertl: btw, I have no clue on how to get nfsv4 debugging up, I study informatics but ive actaully never touched low level programming lanauge as C
1317330874 M * daniel_hozac reboot -f is supposed to bring it back up.
1317330879 M * Bertl well, C is not really low level :) but if you know how to code (regardless of the language) you know the basic principles
1317330911 M * petzsch daniel_hozac: that's what i expected and what works perfectly fine on the 2.6.32 hostmashines
1317330913 M * Bertl Rockj: but don't worry, I'll guide you in the right direction
1317330960 M * Rockj want me to fire up    make menuconfig on the kernel? 
1317331003 M * Bertl yep, let
1317331014 M * Bertl 's first enable all debugging stuff we can find (for nfs)
1317331026 M * daniel_hozac petzsch: so what is /proc/sys/kernel/vshelper set to?
1317331079 M * Bertl Rockj: so first make sure to enable CONFIG_DEBUG_KERNEL and CONFIG_DEBUG_INFO (so we get data),
1317331088 M * petzsch cat vshelper: /usr/lib/util-vserver/vshelper
1317331113 M * petzsch which seems to exist: -rwxr-xr-x 1 root root 6284 29. Sep 00:10 /usr/lib/util-vserver/vshelper
1317331153 M * Rockj Bertl: give me your ssh public key, I can share a screen with you :O
1317331210 M * daniel_hozac so mkdir /etc/vservers/.defaults/apps/vshelper; echo /tmp/vshelper.log > /etc/vservers/.defaults/apps/vshelper/logfile; touch /etc/vservers/.defaults/apps/vshelper/debug
1317331217 M * Bertl Rockj: guess that's not necessary at the moment ... you want to learn something as well, no?
1317331238 M * Bertl Rockj: after all, you then claim that you 'hacked' on the kernel
1317331248 M * Rockj haha, I guess
1317331253 M * Rockj thought it would just go faster for you 
1317331261 M * Rockj instead of wasting too much valuable time of yours
1317331261 M * petzsch Rockj: my the source be with you ;)
1317331279 M * Bertl Rockj: no worries, I'm doing a lot of things in parallel
1317331322 M * Rockj Bertl: aye  , DEBUG_FS wanted?
1317331332 M * Rockj in the Kernel hacking category in the config?
1317331333 M * Bertl yep, let's get that as well
1317331367 M * Rockj do I need debug shared IRQ handlers , detect hard and soft lockups under kernel debugging? I assume no?
1317331396 M * Bertl only nfs related debug features for now (although as far as I see, the general DEBUG should suffice) together with the 'debug' boot parameter and dmesg log level 9)
1317331435 M * Bertl daniel_hozac: what's new in 2990?
1317331452 M * Rockj ok, got CONFIG_DEBUG_KERNEL and CONFIG_DEBUG_INFO
1317331473 M * Bertl good, build the kernel, no need to 'make clean'
1317331494 M * Bertl i.e. whenever you change something, just do 'make'
1317331515 M * Bertl (speeds up things significantly)
1317331518 M * Rockj make bzImage -j5 :)
1317331565 M * petzsch daniel_hozac: last lines of the /etc/vservers/.defaults/apps/vshelper/logfile http://paste.linux-vserver.org/20611
1317331573 M * Bertl petzsch: bzImage is the default, so you can skip that
1317331579 M * Bertl *Rockj
1317331607 M * Rockj Bertl: old habbit, I learned to compile the kernel from som old linux geeks in 2005 that used slackware and hated me for using anything else
1317331609 M * Rockj :p
1317331631 M * Rockj it was my first task ever when touching the linux world, you gotto compile your own kernel :S
1317331633 M * Rockj heh
1317331635 M * Rockj :p
1317331646 M * Bertl sounds good to me :)
1317331743 M * petzsch daniel_hozac: the /usr/bin/find: `dirs': No such file or directory ... was there before your last util-vserver update as well... not quite sure what is means... the vservers start when using vserver start+stop or restart from the hosts commandline
1317331783 M * Rockj Bertl: hm, it already started compiling but .. sure I dont have to enable some kind of nfs4 debug under network file system?
1317331800 M * Rockj *just wondering if I should abort and double check or if those two default debug constants was enough?*
1317331845 M * Bertl we'll see after the first test .. I don't see any debug related options specifically for nfs
1317331852 M * daniel_hozac petzsch: looks like a cgroup problem.
1317331924 M * daniel_hozac Bertl: should allow mountpoints as guest roots during build, building of SL6, and /var/run and /var/lock can be symlinks.
1317331932 M * daniel_hozac Bertl: although 2991 is preferred.
1317331951 M * daniel_hozac (which i now realize i haven't uploaded yet)
1317331966 M * Bertl ah, so that will be there (as tar) shortly then?
1317331988 M * daniel_hozac should be there now.
1317332045 M * Bertl excellent! thanks a bunch!
1317332208 M * petzsch daniel_hozac: just checked my cgroup config for the vserver in question no oddities afaik: cpu.shares 512 | memory.limit_in_bytes 2773833045 | memory.memsw.limit_in_bytes 8321499135
1317332272 M * daniel_hozac petzsch: try doing sleep 5 > /etc/vservers/.defaults/scripts/initialize
1317332371 Q * cuba33ci Read error: Connection reset by peer
1317332401 M * Bertl Rockj: once the kernel is built and installed, modify the bootloader command line to specify 'debug' on the kernel command line, and reboot with the new kernel
1317332476 J * cuba33ci ~cuba33ci@111-240-167-172.dynamic.hinet.net
1317332543 M * petzsch vserver start command taking 5 seconds longer, but reboot within the vserver still leads to a stopped vserver (this time ping response stays off for about 5 seconds)
1317332568 M * Rockj Bertl: ok, its built, im unsure of the bootloader tho
1317332573 M * Rockj I really hate the grub2 config
1317332582 M * Rockj grub1 was so much easier to just edit it and get things done
1317332593 M * daniel_hozac the log is what's interesting.
1317332610 M * Rockj GRUB_CMDLINE_LINUX_DEFAULT="quiet"
1317332613 M * Rockj on this line?
1317332623 M * Bertl yep, replace the quiet with debug
1317332637 M * Bertl you could also specify it per kernel
1317332641 M * petzsch daniel_hozac: the last lines on the screen, let me know if you need it all: http://paste.linux-vserver.org/20612
1317332657 M * Rockj Bertl: lets see what update-grub generates for me
1317332660 M * Rockj in the grub.cfg file
1317332777 M * Rockj http://pastie.org/2613546
1317332827 M * Rockj I guess here we go with a reboot? :)
1317332836 M * daniel_hozac petzsch: okay, let's try adding a cat /proc/self/cgroup to the initialize script as well.
1317332842 N * jrayhawk_ jrayhawk
1317332906 M * Rockj server rebooting, lets hope it boots up again ;D
1317332920 M * Bertl Rockj: looks good, boot and check with 'cat /proc/cmdline'
1317332981 M * Rockj btw, does booting a debug kernel take longer time then a normal one? ohhh
1317332983 M * Rockj ping reply
1317332996 M * Rockj just gotto wait for it to boot up propertly then and run sshd
1317332999 M * petzsch daniel_hozac: new log output: http://paste.linux-vserver.org/20613
1317333059 M * petzsch cgroup got mounted by the util-vserver init script i belive 
1317333128 M * Rockj cat /proc/cmdline 
1317333130 M * Rockj BOOT_IMAGE=/vmlinuz-3.0.4-vs2.3.1-pre10.1-UAC root=/dev/sda1 ro debug
1317333162 M * Bertl looks good, now do 'echo 9 >/proc/sysrq-trigger'
1317333164 M * petzsch not sure if it's of any relvance... when running vserver start after this 'unclean' reboot -f i get a: RTNETLINK answers: File exists
1317333198 M * Rockj sysrq-trigger same as sysrq button?
1317333206 M * Rockj so .. if you echo in b  , it reboots?
1317333220 M * Rockj SysRq : Changing Loglevel
1317333222 M * Rockj Loglevel set to 9
1317333302 M * Bertl good, now let's do the nfs mount and recreate the issue, let's see if we get some debugging
1317333324 M * Rockj can I mount it twice?
1317333326 M * Rockj or nvm
1317333340 M * Rockj default nfs4 mount then or?
1317333342 M * Rockj noac or?
1317333350 M * Bertl whatever
1317333366 M * petzsch Bertl: nice -n 19 echo have you been able to reproduce the reboot -f issue with 3.0.4 on your system?
1317333435 M * Bertl not yet
1317333466 M * Rockj Bertl: no logs :/
1317333468 M * Rockj in syslog
1317333470 M * Rockj nor in dmesg
1317333553 M * Rockj maybe have to start nfs in debug mode? or?
1317333573 M * Rockj Sep 29 23:57:13 absint rpc.idmapd[1225]: nss_getpwnam: name 'ruby' not found in domain 'localdomain'
1317333575 M * Rockj hm
1317333587 M * Rockj that's what causing it?
1317333616 M * Bertl that's idmapd (an nfsv4 feature) looking up an user
1317333630 M * Bertl try: echo 65535 > /proc/sys/sunrpc/nfs_debug
1317333660 Q * sweil Remote host closed the connection
1317333723 J * petzsch1 ~markus@p57B66E20.dip.t-dialin.net
1317333775 M * Rockj http://pastie.org/pastes
1317333779 M * Rockj hmf
1317333782 M * Rockj larger then 64kb
1317333784 M * Rockj bah
1317333790 N * petzsch Guest12022
1317333790 N * petzsch1 petzsch
1317333821 M * Rockj hah
1317333828 M * Rockj your pastebin think it is spam
1317333830 M * Rockj damnit
1317333841 M * Rockj https://gist.github.com/369f446a4cd48fb4d0e5
1317333843 M * Rockj there we go
1317333919 M * Rockj hmf, my username not even in that paste
1317333921 M * Rockj hm
1317333938 M * Bertl doesn't matter
1317333944 M * Rockj but it got generated when I ssh'ed in
1317333946 M * Rockj hm
1317333975 M * Bertl I see a lot of owner/group decodes with -2
1317333985 M * Bertl and some with uid=2220 and gid 1110
1317334015 M * Rockj uidnorangsh:x:2202:1110:Roy Sindre Norangshol,,,,:/home/dotkom/norangsh:/bin/bash
1317334017 M * Rockj looks like me
1317334019 M * Rockj :)
1317334093 M * Rockj you mean 2229 :O that's from those other files
1317334097 M * Rockj let me try again
1317334100 Q * Guest12022 Ping timeout: 480 seconds
1317334103 M * Rockj and get the lookup to authorized_keys
1317334155 M * Rockj https://gist.github.com/010c6af5a09bd1272c81
1317334157 M * Rockj that looks better
1317334259 M * Rockj did it make any sense?
1317334295 M * Bertl delta-nfs-fix01.diff is applied to that kernel, yes?
1317334334 M * Rockj yes
1317334391 M * Bertl okay, give me a few minutes
1317334458 M * Rockj take your time :-) 
1317335748 M * Bertl okay, let's try this one: http://vserver.13thfloor.at/ExperimentalT/delta-nfs-fix02.diff
1317335772 M * Bertl simply apply to the current kernel, then build/install/boot as previous
1317335832 M * Rockj absint:/usr/src/linux-3.0.4-vs2.3.1-pre10.1# cat ../delta-nfs-fix02.diff | patch -p1
1317335834 M * Rockj patching file fs/nfs/inode.c
1317335836 M * Rockj Hunk #1 succeeded at 1341 (offset 8 lines).
1317335838 M * Rockj Hunk #2 succeeded at 1386 (offset 8 lines).
1317335989 M * Rockj rebooting into new kernel 
1317336276 M * Rockj Bertl: seems to work :O
1317336299 M * Rockj at least takes my ssh key
1317336316 M * Rockj and it is using nfs4 normal mount
1317336317 M * Rockj \o/
1317336674 M * Bertl great! now let's try without the fix01
1317336684 M * Bertl i.e. simply revert that one
1317336710 M * Bertl (patch -R)
1317336790 M * Rockj cat ../delta-nfs-fix01.diff | patch -Rp1
1317336792 M * Rockj patching file fs/nfs/inode.c
1317336794 M * Rockj Hunk #1 succeeded at 363 (offset -6 lines).
1317336806 M * Rockj *compiling*
1317336896 M * meebey Bertl: nfs-fix01 didnt work for me, had no time today for further testing though. I can give fix02 a spin though, I planned to ask you if patching fhget was on purpose instead of update_inode where I was seeing the troublesome code always overwriting the uid/gid of the inode
1317336946 M * Bertl yeah, well, I thought that the fix01 was necessary, but missed the wrongly modified update_inode completely
1317336966 M * meebey but good to see that this issue has some progress and wrt nfsv4 not production ready, I have used v3 alot and I think it was never production ready with all the bugs it has in the different kernel versions :-P
1317336982 M * Bertl as Rockj is testing right now, we should know shortly, but you might give him your 'basic tests' so that we can test that as well
1317336985 M * Rockj hmf, forgot make modules and make modules_install . damnit
1317336996 M * Rockj shouldn't matter I guess?
1317337004 M * meebey the basic test is: cat $file; ls -l $file
1317337007 M * meebey it will be -2 then
1317337012 M * Bertl Rockj: unless your nfs is compiled as modules
1317337025 M * Bertl meebey: inside a guest or on the host as well?
1317337027 M * meebey after about 10 seconds when the cache expires, ls -l $file will show normal user/group again
1317337035 M * meebey Bertl: I only tested on host
1317337041 M * Bertl excellent then!
1317337050 M * Rockj Bertl: nah, everything is compiled into kernel
1317337057 M * Bertl meebey: we'll know in a few minutes I guess ....
1317337061 M * Rockj ill comppile the modules (there is 1 or 2) 
1317337064 M * Rockj and install them when it boots
1317337074 M * Rockj meebey: same issue I had
1317337085 M * Rockj which basically screwed up my authorized_keys file
1317337093 M * Rockj and sshd rejecting the handshake ;)
1317337102 M * meebey yeah I have seen such report in the debian bts
1317337134 M * meebey I hit the problem with apache and suexec rejecting to run the script because the file is not the same owner as the directory owner
1317337204 M * Rockj I just find it odd that I only have this issue on 1 out of 4 servers
1317337206 M * Rockj :O
1317337220 M * meebey it doesnt apply to older kernels
1317337237 M * meebey 2.6.26 it was ok, the nfs code looked different
1317337251 M * Rockj Bertl: accepted my authorized keyfile
1317337253 M * Rockj so looks nice
1317337311 M * Rockj http://pastie.org/2613879   ("modules" missing :p)
1317337317 M * Rockj hehe
1317337325 M * Rockj  INSTALL drivers/scsi/scsi_wait_scan.ko
1317337328 M * Rockj  INSTALL fs/isofs/isofs.ko
1317337331 M * Rockj that's it
1317337333 M * Rockj heh
1317337365 M * Rockj did I just help with solving a kernel bug? o_O
1317337366 M * Bertl cool, so while I still think that there is potential for problems without the fix01
1317337379 M * Bertl Rockj: yes, you did!
1317337396 M * Rockj btw, safe to run with debug stuff on the kernel? or should I recompile
1317337407 M * Rockj or enough to just remove the debug flag in boot and reboot?
1317337419 M * Bertl should be safe but adds some overhead
1317337434 M * Bertl (rather insignificant though if not enabled)
1317337493 M * Bertl meebey: I will update/provide patches for older kernels as well, it is a simple bug where I obviously got it backwards :)
1317337519 M * Rockj ok, I'll just leave the debug there
1317337524 M * Rockj incase something happens and I need to turn it on
1317337526 M * meebey yeah I will try to push that patch into debian's stable kernel
1317337526 M * Rockj :)
1317337531 M * meebey as its a serious regression
1317337545 M * meebey I am surprised not more people complained yet
1317337563 M * Bertl IMHO nobody uses nfsv4 :)
1317337564 M * meebey its not like linux has nfs alternatives
1317337575 M * meebey you think its nfsv4 specific?
1317337618 M * Bertl it seems to be
1317337625 M * Rockj hm, shoudln't nfs4 take over?
1317337632 M * Bertl well, at least the issue is unnoticeable with nfsv3
1317338602 Q * markc Ping timeout: 480 seconds
1317338652 M * Bertl Rockj: well, yes, unfortunately it doesn't seem that stable yet
1317338673 M * Bertl I tried it around 2.6.38, and it almost instantly crashed my server
1317338695 M * Rockj :O heh. runs fine here - but now ... I don't have an enterprise setup either 
1317338707 M * Rockj just some student hacking together some services for small amount of users
1317338709 M * Rockj :p
1317338721 J * markc ~markc@203.25.132.186
1317338789 M * Bertl petzsch: only a few more moments, 3.0.4 kernel is up and running on my test setup
1317338844 M * Bertl petzsch: do you have a build line for me so that we have the same guest?
1317338927 M * petzsch vserver vps50 build -m debootstrap --context 50 --hostname vps50.kunden-server.org --interface eth0:46.4.153.22/32 -- -d squeeze -m ftp://ftp.freenet.de/debian
1317338944 M * petzsch Bertl: hope that's what you meant
1317338980 M * Bertl yep, tx!
1317339218 M * Bertl daniel_hozac: two minor bug reports for 2991: 1) the debootstrap url is outdated again and 2) the debootstrap now seems to require a pgp key (at least by default)
1317339283 Q * markc Read error: No route to host
1317339360 Q * ensc|w Quit: Lost terminal
1317339381 M * Bertl petzsch: could you upload your /usr/share/keyrings/debian-archive-keyring.gpg somewhere for me to grab please?
1317339391 M * Bertl (or give me an url where I would find that)
1317339441 M * petzsch just a sec... i'll upload it
1317339524 M * petzsch here: http://uni.petzsch.eu/debian-archive-keyring.gpg
1317339528 M * Bertl tx
1317339581 M * Bertl you are building/using an x86 guest I presume?
1317339589 M * petzsch x86_64
1317339592 M * petzsch same arch as the host
1317339595 M * Bertl okay
1317339627 M * Bertl guest install running now ...
1317339687 M * petzsch @keyring problem: didn't run to it on my wheezy hostsystem using dpkg-buildpackage to turn daniel_hozac's work into a debian package
1317339742 M * Bertl well, I guess as the host system is debian as well
1317339769 M * petzsch debian wheezy, correct... could be the reason
1317339771 M * Bertl you already have that keyring installed, I'm using mageia as host system, which naturally doesn't have the debian keys preinstalled
1317339839 Q * Walex Read error: No route to host
1317340225 M * Bertl okay, after the build, when trying to start the guest, I get a bunch of chroot-sh error/warnings, is that the same you see? http://paste.linux-vserver.org/20614
1317340359 M * petzsch don't see any warnings at all
1317340394 M * Bertl daniel_hozac: any ideas to that output, especially as the guest seems to fail to startup after that?
1317340417 M * petzsch just the /usr/bin/find: "dirs": Datei oder Verzeichnis nicht gefunden line which doesn't seem to have an impact of the vserver starting
1317340689 M * petzsch i'll go to bed... have to get up kinda early
1317340726 M * petzsch have a good one everyone and thx for the help so far
1317340728 M * Bertl okay, I'll leave a message when I get it working
1317340744 M * Bertl seems to be a problem introduced with latest util-vserver 
1317340755 M * petzsch i'll catch on with the irc logs tomorrow
1317340757 M * petzsch good night
1317340763 Q * petzsch Quit: Leaving.
1317340763 M * Bertl sleep well