1453431381 J * derjohn_mobi ~aj@x4db004df.dyn.telefonica.de 1453431822 Q * derjohn_mob Ping timeout: 480 seconds 1453435985 N * Bertl_zZ Bertl 1453435987 M * Bertl morning folks! 1453436166 M * gnarface 'morning Bertl 1453447091 J * Ghislain ~aqueos@adsl1.aqueos.com 1453447964 J * thierryp ~thierry@zeta.inria.fr 1453449186 Q * derjohn_mobi Ping timeout: 480 seconds 1453450897 J * derjohn_mobi ~aj@2001:6f8:1337:0:c1dd:c388:a6f3:c9d6 1453451307 J * nikolay ~Nikolay@199.91.137.248 1453451324 Q * nikolay Remote host closed the connection 1453452056 J * nikolay ~Nikolay@199.91.137.248 1453457483 Q * eyck_ Ping timeout: 480 seconds 1453461373 Q * transacid Remote host closed the connection 1453461388 J * transacid ~transacid@transacid.de 1453463456 Q * AndrewLe1 Ping timeout: 480 seconds 1453464578 J * Ghislain1 ~aqueos@adsl1.aqueos.com 1453464578 Q * Ghislain Read error: Connection reset by peer 1453464807 Q * arekm Remote host closed the connection 1453465028 J * arekm ~arekm@phobos.pld-linux.org 1453471654 J * AndrewLee ~andrew@210.240.39.201 1453472369 M * gnarface hey uh, vservers have swap access, don't they? 1453472371 M * gnarface or do they not? 1453472379 M * gnarface are they not allowed to actually use the host's swap partition? 1453472487 M * Guy- gnarface: I suppose you could set up resource limits that would prevent vserver processes from being swapped out, using the swap control group 1453472498 M * Guy- gnarface: but by default, there is no such limit 1453472558 M * gnarface Guy-: ok i was just having an issue (intermittent server silent i/o lock) that i wondered could be related to the server running out of ram despite there being a giant unused swap partition. 1453472669 M * gnarface Guy-: but if vservers can actually use it by default, that's ok, that's what i want actually 1453472712 M * Guy- gnarface: I think you may have the wrong idea, based on your syntax 1453472725 M * gnarface oh? 1453472731 M * Guy- it's not like "vservers can use swap" 1453472736 M * gnarface oh 1453472738 M * Guy- vservers are just glorified chroots 1453472747 M * gnarface so they *can't* use swap? 1453472751 M * Guy- so there is only one kernel, which schedules all processes, including the ones in vservers 1453472770 M * Guy- if the kernel decides to swap out a process, it will do so, whether the process is running inside a vserver or not 1453472784 M * Guy- it's the kernel that uses the swap, not the vserver 1453472798 M * Guy- the vserver has no mechanism to use swap, because it's just a bunch of userspace processes 1453472807 M * gnarface so uh... they CAN use swap? 1453472811 M * Ghislain1 you can use cgroups to limit memory and swap 1453472814 M * Ghislain1 yes they can 1453472824 M * Ghislain1 and they do unless you limit them 1453472828 M * Guy- gnarface: I think you'll be happy with a "yes", but it's not the technically correct answer 1453472838 M * Ghislain1 if you limit memory i advise you limit swap also 1453472847 M * gnarface Guy-: i get your meaning 1453472847 M * Guy- but in the sense you have in mind, yes they can 1453472881 M * Ghislain1 yeah the swap is not for the guest this is a shared ressource of the host htat can be limited per guest 1453472881 M * gnarface i just wanted to make sure vservers don't all require their own swap partitions to keep from being able to OOM the host 1453473033 M * Guy- gnarface: they can cause oom unless you limit their memory usage (including swap) 1453473050 M * gnarface hmm. Guy- would that cause an i/o lock? 1453473074 M * Guy- you mean a deadlock? 1453473081 M * Guy- I don't think so 1453473085 M * gnarface maybe? so far at the point of the lockup though, and actually at basically all times munin graphs are reporting ZERO swap usage 1453473110 M * Guy- gnarface: I think you should try to obtain a kernel stacktrace when this happens 1453473115 M * gnarface hmm 1453473189 M * gnarface i was considering maybe just upgrading from 3.14.17-vs2.3.6.13 to 3.14.52-vs2.3.6.15 first in hopes that someone has already fixed it without me having to diagnose it 1453473238 M * Guy- that's certainly something you could try 1453473248 M * Guy- while you are at it, make sure to enable the various watchdog features of the kernel 1453473258 M * Guy- soft and hard lockup detection, mainly 1453473342 M * gnarface just trying to remember how i did it the last time 1453473354 M * gnarface and trying to figure out if i can start from the same kernel config safely 1453473359 M * Ghislain1 longterm:3.14.58 is the last kernel 1453473380 M * Ghislain1 i mean in 3.14 1453473384 M * gnarface oh? i was just going off what was linked on the website, Ghislain1 1453473412 M * gnarface here: http://linux-vserver.org/Welcome_to_Linux-VServer.org 1453473436 M * Ghislain1 the patch is for 52 but the latest linux kernel is 58 dont know if the patch apply without error to 58 1453473457 M * gnarface ah i see. yes i remember someone recently saying it worked for them at least 1453473465 M * gnarface ok good to remember 1453473486 M * gnarface it could be unrelated but i'm also getting apache logs missing in one of the vservers 1453473498 M * gnarface the logs get rotated, and the new log gets created and apache gets restarted presumably 1453473509 M * gnarface but the old log just disappears 1453473544 M * gnarface i'm assuming its an unrelated configuration error on my part, but its also possible that they're disappearing because its during the rotation that its crashing every time 1453473595 M * gnarface (i guess technically during the "compression" phase, since the existing already-compressed files are still getting their numbers iterated, so i have now the first log, and 5-53, but i'm missing 2,3, and 4 1453473603 M * gnarface ) 1453474459 J * eyck ~eyck@u28n61.nowanet.pl 1453475097 M * gnarface hmm. actually maybe they're not going missing, maybe the main file is just never being rotated off... that's weird 1453475155 M * Ghislain1 in lenny i have issue in logrotate that try to /etc/init.d/rsyslog rotate but launch that with the apache user or the dovecot user 1453475161 M * Ghislain1 that fails completly 1453475171 M * Ghislain1 sorry i meant jessie 1453475192 M * gnarface well this was all working, until i tried to add in awstats and make it run as its own user, which required changing the permission of the apache logs 1453475199 M * gnarface apparently i've misunderstood something 1453475212 M * gnarface (or found a bug, but probably misunderstood something) 1453475258 M * Ghislain1 permission are tricky, just chmod -R 777 / that solves everything ! :p 1453475282 M * gnarface heh, well i'm not gonna be doing that 1453475282 M * Ghislain1 this is a joke dfo not ever do that 1453475291 M * Ghislain1 :p 1453475291 M * gnarface i actually understand unix permissions pretty well 1453475305 M * gnarface so the misunderstanding here is more about wtf apache or logrotate is mad about 1453475317 M * Bertl yeah, also use 03777 instead 1453475344 M * gnarface so the logs are awstats:adm 0640 1453475347 M * Ghislain1 lol 1453475368 M * gnarface and the log directory (/var/log/apache2) is awstats:adm 0750 1453475441 M * gnarface but files like mydomain.com-access.log.4.gz get renamed to mydomain.com-access.log.5.gz without the replacement #4 being put in its place 1453475473 M * gnarface i thought it was disappearing, but the main "mydomain.com-access.log" file has entries old enough to suggest its just never getting rotated 1453475553 M * gnarface apache is running as www-data as per debian default, but it should create logs as root, i thought 1453475573 M * gnarface so i don't understand how it could be continuing to log and rotation continues to happen for the old files but not new ones 1453475631 M * gnarface unless i've run into some rule i was unaware about like logrotate refuses to create new files if the /var/log/apache2 directory isn't owned by root or something weird? 1453476361 Q * AndrewLee Ping timeout: 480 seconds 1453478102 J * AndrewLee ~andrew@210.240.39.201 1453479290 M * gnarface ah, i may have discovered the issue 1453479302 M * gnarface not the lockup, but the logrotation issue 1453479329 M * gnarface real dumb, prerotate script was returning non-zero status because awstats user's home directory was /nonexistant 1453479676 Q * thierryp Remote host closed the connection 1453481273 Q * derjohn_mobi Ping timeout: 480 seconds 1453482155 Q * nikolay Remote host closed the connection 1453483223 M * Bertl off for a nap ... bbl 1453483243 N * Bertl Bertl_zZ 1453486551 M * gnarface you guys still recommend NOT compiling KVM in to the linux-vservers kernel right? 1453486639 M * gnarface so i should just uncheck this entire thing, right? [*] Virtualization ---> 1453487268 M * daniel_hozac_ no, that's fine. 1453487275 M * daniel_hozac_ i run both kvm and vserver on the same host. 1453487516 M * gnarface daniel_hozac_: ok, so its safe, but definitely not required? 1453487720 M * daniel_hozac_ sure 1453487747 M * gnarface k, thanks 1453490576 M * gnarface alright, building 3.14.58-vs2.3.6.15 now 1453490616 M * gnarface the patch for 3.14.52 only needed one edit to apply to 3.14.58 1453491504 M * arekm daniel_hozac_: hi, probably something is still problematic, that's 4.1.15+vs+delta patch - http://sprunge.us/ZJdA ... pid_revalidate thing inside 1453493156 M * arekm daniel_hozac_: reproduced by sshing into guest, strace -f -F -s 200 -p XYZ (some running process; in my case it was "less"); then ssh second session and sudo killall -9 strace 1453493191 M * arekm (strace 4.11 in my case, not sure if that matters) 1453494589 M * daniel_hozac_ arekm: can you addr2line that? 1453494598 M * daniel_hozac_ killall and strace anyway 1453494750 J * derjohn_mob ~aj@x4db004df.dyn.telefonica.de 1453494929 M * daniel_hozac_ the strace doesn't work, right? 1453494961 M * arekm seems so, doesn't print anything 1453494979 M * arekm and can't quit straced 'less' 1453494995 M * daniel_hozac_ it looks like it's in yama 1453494998 M * arekm so killall not needed 1453495057 M * arekm (addr2line pending, getting vmlinux first) 1453495229 M * daniel_hozac_ which is probably why we haven't seen it before. 1453495252 M * daniel_hozac_ (honestly i hadn't even heard of yama... but i'm kind of out of the loop) 1453495429 M * arekm yama are bunch of some security protections gathered from other projects (or other ideas) 1453495437 M * arekm seems like one feature actually, https://www.kernel.org/doc/Documentation/security/Yama.txt 1453495449 M * daniel_hozac_ yeah, i read up on it now 1453495743 M * arekm linux-4.1/security/yama/yama_lsm.c:302 1453495774 M * arekm http://sprunge.us/iMdT 1453496071 M * daniel_hozac_ is that RIP? 1453496075 M * daniel_hozac_ or the first in the trace? 1453496130 M * arekm [] yama_ptrace_access_check+0x148/0x1c0 1453496146 M * daniel_hozac_ [53400.638059] [] task_is_descendant.part.1+0x2a/0x70 would be good too 1453496182 M * arekm ffffffff812eae0a is yama/yama_lsm.c:226 which is http://sprunge.us/DRPb 1453496290 M * daniel_hozac_ could you also get objdump -D ./vmlinux | grep -A 1000 ' :'? 1453496415 M * arekm yes. rebooting, so a sec... 1453496466 M * daniel_hozac_ i'm still building a new kitchen sink kernel with yama enabled so i can test some myself... it could use some config cleaning. 1453499807 M * arekm objdump -D ./vmlinux | grep -A 1000 ' :' -> http://sprunge.us/CjWW (grep -A 1000 ' :' found nothing) 1453500290 Q * Defaultti Quit: Quitting. 1453500376 Q * Ghislain1 Quit: Leaving. 1453500440 J * Defaultti defaultti@lakka.kapsi.fi 1453504136 M * daniel_hozac_ arekm: are you using a fake init? 1453504613 M * arekm daniel_hozac_: style plain so real init 1453505033 M * daniel_hozac_ right, well, that's fake init ;) 1453505067 M * daniel_hozac_ okay. this is probably due to the same issue Guy- observed with vps faux not listing guest processes, because the guest init is its own parent 1453505101 M * daniel_hozac_ i haven't had time to look at it yet, planning on doing that this weekend though. 1453506293 M * arekm ok 1453506294 M * arekm sleep time