1302140452 J * imcsk8 ~ichavero@148.229.9.250 1302141369 N * deever_ deever 1302141935 Q * FireEgl Ping timeout: 480 seconds 1302142292 Q * imcsk8 Quit: This computer has gone to sleep 1302143567 J * FireEgl FireEgl@Sebastian.FireEgl.Com 1302144488 J * AlexanderS ~Alexander@f052049044.adsl.alicedsl.de 1302144547 M * AlexanderS hi, anybody know any way to execute a script after completly stopping a vserver, so that I can "eject" its rootdisk? 1302144574 M * AlexanderS even postpost-stop has mounted the root 1302145292 M * MooingLe1ur kludgy, but you could spawn a background process at that point and wait for it to be unmounted :) 1302145323 M * MooingLe1ur loop until it succeeds, with a timeout of a few min, perhaps 1302145684 M * Bertl AlexanderS: define 'mounted' 1302145751 M * Bertl i.e. who mounted it, where and when? 1302145789 M * AlexanderS vserver-utils mounted it during the start of the vserver (it is listeted in /etc/vservers//fstab) 1302145842 M * Bertl and what do you want to do to 'eject' it? 1302145873 M * Bertl it will be unmounted once the guest's mount namespace is completely gone 1302145904 M * AlexanderS concret: i want to write scripts that handling iscsi. each vserver should login to the iscsi target and get its volumes during start and should logout from the target during stop 1302145917 M * Bertl alternatively you can probably leave the guest root and unmount it manually in postpost-stop 1302147354 M * AlexanderS tested, and no. the process /bin/bash /usr/sbin/vserver ----nonamespace stop has the cwd as current cwd 1302147367 M * AlexanderS *has the root as cwd 1302148073 M * Bertl and you did change the cwd in your script, yes? 1302148127 M * AlexanderS yes 1302148460 M * Bertl okay, please notify daniel_hozac, that might be a bug 1302148619 M * AlexanderS okay 1302153026 M * Bertl off to bed now .. have a good one everyone! 1302153034 N * Bertl Bertl_zZ 1302153745 Q * FireEgl Ping timeout: 480 seconds 1302154175 J * derjohn_mob ~aj@d025228.adsl.hansenet.de 1302154910 Q * FIChTe Quit: bashpipe 1302154985 J * FIChTe ~fichte@bashpipe.de 1302155140 Q * _nono_ synthon.oftc.net oxygen.oftc.net 1302155524 J * ncopa ~ncopa@3.203.202.84.customer.cdi.no 1302155675 Q * derjohn_mob Ping timeout: 480 seconds 1302155880 M * daniel_hozac AlexanderS: it's known, it's being fixed... 1302156156 J * _nono_ ~gomes@licencieux.ircam.fr 1302156418 J * fisted_ ~fisted@p50883869.dip.t-dialin.net 1302156425 Q * fisted Ping timeout: 480 seconds 1302157170 J * derjohn_mob ~aj@213.238.45.2 1302158517 Q * _nono_ synthon.oftc.net oxygen.oftc.net 1302158747 J * _nono_ ~gomes@licencieux.ircam.fr 1302159266 J * manana ~mayday090@84.17.25.149 1302159300 J * ghislain ~AQUEOS@adsl2.aqueos.com 1302160487 Q * nkukard Ping timeout: 480 seconds 1302161031 J * nkukard ~nkukard@41-133-112-179.dsl.mweb.co.za 1302162463 J * bsingh ~balbir@122.248.163.2 1302163784 J * petzsch ~markus@dslb-092-078-115-203.pools.arcor-ip.net 1302164544 Q * petzsch Quit: Leaving. 1302165560 J * imcsk8 ~ichavero@148.229.9.250 1302165596 Q * bsingh Ping timeout: 480 seconds 1302166277 N * manana krushik 1302168953 M * AlexanderS daniel_hozac; it's fixed? in what version? 1302170832 Q * AlexanderS Quit: AlexanderS 1302170845 J * AlexanderS ~alex@fob.spline.inf.fu-berlin.de 1302171086 Q * AlexanderS 1302171095 J * AlexanderS ~alex@2001:6f8:1c3c:f76::152:1 1302172124 M * daniel_hozac AlexanderS: not yet. it's being fixed. 1302172637 J * Hawq ~hawk@limanowa.net 1302174103 M * AlexanderS daniel_hozac: ah okay... let me know, if I could help 1302175105 Q * click Ping timeout: 480 seconds 1302175700 M * Hawq Hi. I've problem with mounting guest resources using UUIDs. There are two guests, on one mounting by UUID works, on second it fails 1302175719 M * daniel_hozac errors? configuration? 1302175756 M * Hawq using strace I found that on working guest secure-mount forks /bin/mount which grabs real device name from /dev/disk/by-uuid and does mount() syscall 1302175795 M * Hawq on second one secure-mount tries mount() syscall directly and obviously it fails because uuid is given as device 1302175843 M * Hawq why wtih second guest /bin/mount isn't called? probably some simple mistake on my side, but I can't figure it out 1302175902 M * daniel_hozac /dev/disk/by-uuid should always work, regardless of which of those you use 1302175996 M * Hawq the only error returned is "no such file or directory" from which I think that syscall tries to find file UUID=something and fails 1302176053 M * Hawq /bin/mount checks /dev/disk/by-uuid and provides proper /dev/something to syscall... 1302176124 M * Hawq is there any way to force secure-mount to call /bin/mount instead of using syscall directly? 1302176314 M * Hawq or what can be the reason that it choose direct syscall instead of calling /bin/mount? 1302176327 M * daniel_hozac well 1302176332 M * daniel_hozac you can either just use /dev/disk/by-uuid 1302176338 M * daniel_hozac or, set the filesystem type to auto. 1302176347 M * daniel_hozac which will force /bin/mount to be used. 1302176704 M * Hawq yup. it works with both workarounds. thanks. yet I still must dig it out why on one guest secure-mount prefers /bin/mount :) 1302176849 M * daniel_hozac what was its configuration? 1302176860 M * daniel_hozac (notice how that was my second question, after the error) 1302177006 M * Hawq in both cases its lvm volume to be mounted as /var inside guest. I already have an answer... in working guest fs is xfs, in non-working its ext3. reformatting to xfs helped (/bin/mount is preferred over direct syscall) 1302177046 Q * ser Remote host closed the connection 1302177049 J * ser ~ser@host1.tldp.ibiblio.org 1302177116 M * Hawq fstab entry is same on both guests except for fs type 1302177123 M * Hawq UUID=ba0aa13a-0316-43c5-8143-91804bab6cd2 /var xfs defaults,noatime,nodiratime,data=writeback,commit=30 1 1 1302177242 M * arekm noxxatime makes little sense with recent kernels (since relatime is default for some time) 1302177322 M * daniel_hozac i'm not sure why you'd prefer mount over the syscall, but... 1302177483 M * AlexanderS daniel_hozac: could it be solved by something like that: http://p.animux.de/FifQ2Aw8AE.html (with teardown as a new script as counter part to initialize) 1302177540 M * Hawq daniel_hozac: syscall() doesn't parse UUID=something and ends with ENOENT (No such file or directory) 1302177556 M * Hawq arekm: not so recent kernel, 2.6.32.x :) 1302178013 M * daniel_hozac Hawq: right, but /dev/disk/by-uuid works all the time. 1302178016 M * daniel_hozac AlexanderS: umm, i guess. 1302178027 Q * ser Remote host closed the connection 1302178031 M * daniel_hozac aside from the syntax error. 1302178067 N * Bertl_zZ Bertl 1302178070 M * Bertl morning folks! 1302178583 M * Hawq daniel_hozac: true... but old habits die hard :) (using UUID=) 1302178713 M * Bertl daniel_hozac: the teardown script approach sounds like a good idea to me ... 1302178729 M * daniel_hozac sure, it's needed for other things as well. 1302178732 M * Bertl btw, I have a 'feature request' for the guests fstab 1302178743 M * Bertl i.e. guest configs fstab 1302178751 M * daniel_hozac vserver.stop and vserver.start need overhauls. 1302178812 M * Bertl I'd like to see a placeholder/variable/whatever to refer to the guest name or guest path there, specifically for mounting files from /etc/vservers// to /proc or similar 1302178864 M * daniel_hozac okay 1302178881 M * Bertl so either something like %name, $name ${NAME} or a 'special' marker like '%' '$' or whatever referring to /etc/vservers/ 1302178904 M * daniel_hozac shouldn't be too hard. 1302178924 M * Bertl that'd be great! thanks in advance! 1302179001 M * Bertl btw, a friend of mine reported a new debian issue (at least new for me) with guest quote support 1302179013 M * Bertl *quota 1302179048 M * daniel_hozac oh? 1302179066 M * Bertl debian squeeze seems to use /proc/mounts instead of /etc/mtab now 1302179079 M * Bertl and of course, that shows /dev/root and the true filesystem 1302179091 M * daniel_hozac right... 1302179096 M * Bertl /dev/root can be worked around with symlinks, the fs not 1302179106 M * Mr_Smoke Hm 1302179126 M * Mr_Smoke How is it that my /proc/mounts and /etc/mtab are exactly the same ? 1302179137 M * daniel_hozac probably a symlink. 1302179137 M * daniel_hozac :) 1302179138 M * Bertl maybe ln -s ? 1302179170 M * Mr_Smoke Nope 1302179197 M * Mr_Smoke Just saying, it happens in the guest 1302179203 M * Mr_Smoke Not an issue for me though :) 1302179215 M * Bertl of course, /proc/mounts or /proc/self/mounts cannot be faked with bind mounts either ... so I guess we need to address this somehow 1302179266 M * Hawq same here on my squezee guest. no issue. /proc/self/mounts has exactly same contents as mtab, /proc/mount is symlink to /proc/self/mounts 1302179269 M * Bertl I was thinking of a reverse vroot mapping for /proc/self/mounts which can be enabled via a flag 1302179311 M * Bertl Hawq, Mr_Smoke: the problem is that usually /etc/mtab is used to trick quota tools into not doing ugly stuff 1302179316 M * Mr_Smoke oh ok 1302179333 Q * imcsk8 Quit: This computer has gone to sleep 1302179341 M * Bertl but recent quota tools just ignore /etc/mtab and use the proc entry directly 1302179349 M * Bertl (in debian that is) 1302179457 M * Hawq ok. I'm not using quota on squezee yet, but planning to upgrade from lenny soon so will see if something fails 1302181439 M * Bertl daniel_hozac: what's the script to dump/restore filesystem attributes when migrating from 2.6.26? 1302181995 M * daniel_hozac http://people.linux-vserver.org/~dhozac/t/save-vsdata.sh 1302182491 M * Bertl ah, thanks! 1302184753 J * ktwilight__ ~keliew@91.176.78.48 1302185049 Q * ktwilight_ Ping timeout: 480 seconds 1302185664 J * mcp ~mcp@wolk-project.de 1302186104 J * click click@ti0127a340-0148.bb.online.no 1302186448 M * Bertl nap attack .. bbl 1302186452 N * Bertl Bertl_zZ 1302187495 J * dowdle ~dowdle@scott.coe.montana.edu 1302188454 M * AlexanderS daniel_hozac: now tested: http://p.animux.de/FR6jNurW8g.html 1302188470 M * AlexanderS should I send it to the vserver mailinglist? 1302188518 M * daniel_hozac that's not the right way to do it. 1302188655 M * AlexanderS not? 1302190078 M * AlexanderS if you describe me the right way, i try to do it the right way ;) 1302190795 J * imcsk8 ~ichavero@148.229.9.250 1302191470 Q * imcsk8 Quit: This computer has gone to sleep 1302192246 Q * ncopa Quit: Leaving 1302192620 Q * derjohn_mob Ping timeout: 480 seconds 1302192918 J * bonbons ~bonbons@2001:960:7ab:0:95e0:c230:5969:65e 1302194565 J * imcsk8 ~ichavero@148.229.1.11 1302196508 M * geb hi folks 1302196560 M * geb daniel_hozac, as bertl is not here, i got a question for you: do you know why /proc/virtual/$x/ccact can be wrong ? 1302196606 M * daniel_hozac on what? 1302196608 M * geb it show 120Mb/s of inbound and outbond traffic on a vserver that hasn't any kind of traffic (and on a host with a 100Mb/s nic) 1302196626 M * daniel_hozac lo goes there as well. 1302196651 M * daniel_hozac you should use iptables if you want accurate on-link data. 1302196655 M * geb it is an empty vserver just used as a template, it has nothing running on it except a syslog 1302196685 M * geb that's not the first time i see it, it does that every time i vserver clone it 1302196695 M * geb (i gave you the link to the graphs in private) 1302196770 M * daniel_hozac well, what are you grahping exactly? 1302196782 M * daniel_hozac and what is the raw data? 1302196831 M * geb http://paste.linux-vserver.org/19236 1302196876 M * daniel_hozac AlexanderS: basically rewriting vserver.stop to not keep any part of the context around post-init_cmd_stop. 1302196889 M * geb it uses /proc/virtual/$x/ccact , i checked it , the graphs looks to be correct, the value is constantly increasing 1302196895 M * daniel_hozac AlexanderS: and vserver.start to start the context early and do similar checks. 1302196911 M * daniel_hozac geb: which values? 1302196953 M * geb INET: 1302197728 N * Bertl_zZ Bertl 1302198916 M * Bertl back now ... 1302199192 M * Bertl geb: what kernel/patch and what does the proc entry contain? 1302200444 M * AlexanderS daniel_hozac: you mean, removing that vnamespace calling in /usr/sbin/vserver and rewrite vserver.stop to only use the context there? 1302200524 M * daniel_hozac no 1302200538 M * daniel_hozac remove vnamespace, use it where it's needed. 1302200607 M * AlexanderS yes, thats what I wanted to describe... okay I will try it 1302200626 M * daniel_hozac note that it's needed in a lot of places. 1302200629 M * daniel_hozac that's why it was moved. 1302200668 M * AlexanderS okay 1302201852 Q * imcsk8 Quit: This computer has gone to sleep 1302202148 J * hparker ~hparker@2001:470:1f0f:32c:beae:c5ff:fe01:b647 1302202599 N * ensc Guest1032 1302202609 J * ensc ~irc-ensc@p5DF2BE71.dip.t-dialin.net 1302202831 J * zbyniu ~zbyniu@ip-62.181.188.13.static.crowley.pl 1302203019 Q * Guest1032 Ping timeout: 480 seconds 1302203143 J * imcsk8 ~ichavero@148.229.1.11 1302205392 J * derjohn_mob ~aj@c135079.adsl.hansenet.de 1302207033 Q * cuba33ci Read error: Connection reset by peer 1302207113 J * cuba33ci ~cuba33ci@111-240-166-37.dynamic.hinet.net 1302207668 Q * krushik Remote host closed the connection 1302207937 J * thal ~thalunil@walledcity.de 1302210299 Q * bonbons Quit: Leaving 1302210776 Q * dowdle Remote host closed the connection 1302212094 Q * derjohn_mob Ping timeout: 480 seconds 1302214686 Q * ghislain Quit: Leaving. 1302215862 Q * imcsk8 Quit: This computer has gone to sleep 1302219125 Q * hparker Quit: Quit 1302219303 J * hparker ~hparker@2001:470:1f0f:32c:beae:c5ff:fe01:b647