1397435083 M * Bert off to bed now ... have a good one everyone!
1397435092 N * Bert Bert_zZ
1397442446 J * SteeleNivenson ~SteeleNiv@pool-108-29-139-222.nycmny.fios.verizon.net
1397442458 J * SteeleNivenson_ ~SteeleNiv@pool-108-29-139-222.nycmny.fios.verizon.net
1397444675 Q * SteeleNivenson_ Quit: Leaving
1397453395 Q * jrayhawk Ping timeout: 480 seconds
1397454616 Q * SteeleNivenson Read error: Connection reset by peer
1397455057 J * Ghislain ~aqueos@adsl1.aqueos.com
1397455517 J * jrayhawk ~jrayhawk@nursie.omgwallhack.org
1397456345 J * urbee ~urbee@193.77.138.30
1397456414 M * urbee daniel_hozac, Bert_zZ, around?
1397456423 M * urbee regarding my route missconfiguration
1397456433 M * urbee <daniel_hozac> you most likely want source based routing
1397456668 M * urbee i think i have this set up..
1397456672 M * urbee rt_tables etc ..
1397458046 Q * snixor Quit: KVIrc 4.1.3 Equilibrium http://www.kvirc.net/
1397459913 Q * kiorky Ping timeout: 480 seconds
1397460116 J * kiorky ~kiorky@cryptelium.net
1397466089 N * Bert_zZ Bert
1397466094 M * Bert morning folks!
1397466289 J * BenG_ ~bengreen@85.255.235.91
1397466685 M * daniel_hozac Bert: have you dropped the l for a reason?
1397466701 N * Bert Bertl
1397466715 M * Bertl nope, thanks for letting me know :)
1397466797 M * Bertl otherwise I might have spent half of the day without 'l' :)
1397467028 M * daniel_hozac that would have been terrible
1397467061 M * Bertl yes, one of my reoccurring nightmares :)
1397467079 M * daniel_hozac hehe
1397467654 J * ensc|w ~ensc@www-old.sigma-chemnitz.de
1397467953 J * laurens ~laurens@static.70.47.40.188.clients.your-server.de
1397467957 M * laurens morning
1397467967 M * Bertl morning!
1397467973 M * laurens Hello Bertl :)
1397467989 M * laurens I'm trying to setup wheezy guests on wheezy, but I get errors :(
1397468000 M * laurens I tried the mailing list, but got no answer :(
1397468017 M * Bertl kernel/patch/util-vserver version and what errors?
1397468103 M * laurens ok, sec
1397468146 M * laurens kernel: 3.13.6-vs2.3.6.11-beng from psand repo
1397468158 M * laurens util-vserver: 0.30.216-pre3054; Nov 19 2013, 10:24:54
1397468257 M * laurens command & error: http://pastebin.com/Puw98DMv
1397468304 M * Bertl okay, rerun it with --debug please
1397468773 M * laurens ok, sec
1397469092 Q * BenG_ Quit: I Leave
1397469192 M * laurens http://pastebin.com/d68mUHK4
1397469249 M * Bertl E: Couldn't find these debs: libglib2.0dev
1397469271 M * Bertl I presume you typed it wrong
1397469369 M * laurens would that account for the rest of the errors?
1397469382 M * daniel_hozac yes.
1397469386 M * Bertl yes, very likely, try with a correct set of includes
1397469405 M * laurens let me try again
1397469613 M * laurens seems to work now, thank you very much! :)
1397469636 M * Bertl you're welcome!
1397470961 Q * urbee Ping timeout: 480 seconds
1397471380 M * fback Bertl: should multiple RT work with ipv6?
1397471557 M * Bertl I think so, at least it was kind-of included back in 2006
1397471757 M * fback http://paste.linux-vserver.org/58812
1397471858 M * Bertl looks like it works, no?
1397471913 M * fback nope, there's no 2a01:1d8:1:: interface in this guest
1397471921 M * fback just 2a01:1d8::
1397471939 M * Bertl well, I don't know your routing tables and rules
1397471961 M * fback the second get forces source, then I get expected result
1397472667 Q * Aiken Remote host closed the connection
1397472795 M * Bertl sorry, I still do not use ipv6 so I can't help much ... my suggestion is to try the setup without the Linux-VServer patch, and get it there working first
1397472835 M * Bertl once that works as expected (might need some kernel options) retest with the Linux-VServer patch and see if there are differences in the behaviour (which might hint a bug)
1397475598 M * laurens what exactly does "open(memory.stat): No such file or directory" when running vserver-stat?
1397475641 M * daniel_hozac that you don't have the memory cgroup mounted, but your kernel requires it to keep track of memory usage for guests.
1397475654 M * daniel_hozac most likely you didn't run the util-vserver initscript.
1397475666 M * daniel_hozac or you have the Debian initscript.
1397475754 M * fback Bertl: I have no idea how to do it outside the guest
1397475780 M * Bertl multiple routing tables are not something which is Linux-VServer specific
1397475806 M * fback Bertl: this part works
1397475862 M * fback even with vserver patch
1397475921 M * daniel_hozac are you saying it doesn't work inside a guest?
1397475951 M * fback daniel_hozac: it takes wrong interface for ipv6
1397475975 M * fback in my case the one that's not assigned to the guest
1397475978 M * daniel_hozac does it actually? tcpdump says so?
1397475999 M * fback daniel_hozac: http://paste.linux-vserver.org/58812
1397476009 M * fback it's from the guest
1397476023 M * fback first ip rou gest should be the same as second
1397476047 M * daniel_hozac ip route get doesn't use the actual path.
1397476055 M * daniel_hozac do actual connections get the wrong one?
1397476060 M * fback yes
1397476070 M * daniel_hozac so tcpdump shows this?
1397476132 N * l0kit Guest6425
1397476137 J * l0kit ~1oxT@0001b54e.user.oftc.net
1397476366 M * fback daniel_hozac: http://paste.linux-vserver.org/58813
1397476391 M * fback this should go with other interface, vlan38
1397476393 M * daniel_hozac and what kernel is that on?
1397476406 M * fback 3.10.19-vs2.3.6.8-p4-1
1397476518 Q * Guest6425 Ping timeout: 480 seconds
1397476701 M * Bertl daniel_hozac: hmm, are we missing some virtualization in the kernel that ip route get doesn't report the correct data?
1397477780 M * laurens daniel_hozac: how do I mount cgroup?
1397477787 M * laurens mount -t cgroup vserver /dev/cgroup?
1397477814 M * Bertl it should be mounted correctly by the util-vserver startup scripts
1397477871 M * laurens hmmm
1397477883 M * laurens apparently it didn't do that...
1397477963 M * Bertl was it executed at host startup?
1397478116 M * laurens i'll doublecheck, sec
1397479014 M * laurens it is now :) sorry again
1397479278 M * Bertl no problem
1397480139 J * noob123 ~kvirc@cpe-31-15-133-178.cable.telemach.net
1397480147 N * noob123 snixor
1397480149 M * snixor hi
1397480154 M * Bertl hey
1397480171 M * snixor a quick question: is there a way to limit vserver guest size with cgroups?
1397480187 M * Bertl define 'guest size'
1397480200 M * snixor guest container (folder)
1397480209 M * snixor . /vservers/guest
1397480209 M * Bertl filesystem wise? no
1397480240 M * snixor what would be a best practice solution for limiting it's size?
1397480252 M * Bertl there are several options
1397480256 M * snixor seperate lvm?
1397480263 M * Bertl you can use a separate partition per guest, e.g. lvm
1397480277 M * snixor k
1397480280 M * snixor thank you Bertl!
1397480283 M * Bertl you can use a shared partition on a filesystem which supports group quota
1397480306 M * Bertl and you can use a shared partition with tagging on a filesystem which supports file tagging
1397480314 M * Bertl (and thus disk limits)
1397481085 J * benl ~benl@dockoffice.sonassihosting.com
1397481140 M * benl Hi troops
1397481144 M * benl Simple questsion
1397481146 M * benl *question
1397481207 M * benl When applying a cpuset cap to a guest
1397481217 M * benl How can you verify its working
1397481231 M * benl Inside the guest `/proc/cpuinfo` and top will still show all cores
1397481891 M * Bertl depends on your guest config, with VXF_VIRT_CPU it should only show allowed cpus
1397481907 M * benl Thanks Bertl
1397481915 M * benl Another question
1397481926 M * benl I'm investigating the effects of memory exhaustion in a guest
1397481968 M * benl And if I drive memory usage to 100% - I'll see "Memory cgroup out of memory"
1397481975 M * benl (expected)
1397481994 M * benl But load averages on the machine SOAR (ie. in excess of 300 for a 4 core host)
1397482028 M * benl This renders the host pretty much inacessible until the load subsides
1397482058 M * benl am I missing something - I would have thought the purpose of constraining a guest was to prevent it exhausting resources on the host?
1397482068 M * Bertl sounds like a mainline cgroup problem to me
1397482097 M * benl old kernel thing?
1397482103 M * Bertl could be
1397482112 M * benl 3.2.50-vs2.3.2.16-beng
1397482121 M * Bertl note that sometimes the interpretation is misleading, because some applications increase the load under memory pressure
1397482133 M * benl Yeah, CPU load was nil
1397482145 M * benl But load averages were extraordinary
1397482156 M * benl But it was having a dramatic effect on the host
1397482159 M * benl SSH was non-functional
1397482162 M * Bertl yeah, task x gets killed from OOM, task y restarts task x
1397482164 M * benl serial console "hung" etc.
1397482180 M * Bertl that repeats over and over again
1397482207 M * Bertl but recent kernels (i.e. 3.13) should mitigate at least the I/O overhead involved, maybe even more
1397482225 Q * undefined Quit: Closing object
1397482305 J * undefined ~undefined@66-190-97-211.dhcp.unas.tx.charter.com
1397482320 M * benl how does 3.10 fair?
1397482333 M * benl I've been unable to use 3.13 as it doesn't seem to work with DRBD 8.4.4
1397482343 M * Bertl no idea, you've probably have to try it
1397482354 M * benl no problem
1397482367 M * benl I assume there aren't any "flags" in vserver that could mitigate/lessen this
1397482524 M * Bertl well, you can limit kernel resource, tasks, I/O, etc as well (with cgroups or Linux-VServer mechanisms)
1397482537 M * benl any good examples?
1397482551 M * Bertl it should be possible to keep the impact low, but it won't eliminate the issue completely
1397482567 M * benl I'm just trying to keep the host recoverable from the situation
1397482587 M * benl nothing worse than using a vserver guest to try and cap resources - and protect the host
1397482610 M * benl then find when it does run out of resources, it takes the host with it anyway
1397482693 M * Bertl well, if you limit the cpu and I/O resources to the guest, and reserve enough for the host, it shouldn't bring the host down
1397482718 M * benl I've capped CPU and memory
1397482724 M * Bertl if it does, then the kernel has a bug in the cgroup subsystem and you need to upgrade and/or report it upstream
1397482724 M * benl What else can/should I cap
1397483012 M * Bertl depends on what brings your host down
1397483038 M * Bertl check with magic sysrq what the cpus are doing (over serial console) to get an idea what happens
1397483049 M * Bertl but in general, I'd first try to update the kernel
1397483221 M * daniel_hozac note that if you have a memory cgroup limit including swap, you'll be going to swap when you hit the limit.
1397483226 M * daniel_hozac usually killing any and all IO.
1397483297 J * AndrewLee ~andrew@210.240.39.201
1397483659 J * urbee ~urbee@193.77.138.30
1397483715 M * benl how do you define a cgroup limit without swap
1397485894 M * Bertl by specifying the swap space as zero?
1397487836 Q * padde Remote host closed the connection
1397487841 J * padde ~padde@patrick-nagel.net
1397488129 Q * Guy- Remote host closed the connection
1397488137 J * Guy- ~korn@elan.rulez.org
1397488618 Q * karasz Ping timeout: 480 seconds
1397488754 J * karasz ~karasz@00015555.user.oftc.net
1397490554 M * benl Where do you define it per vserver?
1397490572 M * Bertl with the cgroup configuration
1397490586 M * Bertl you specify a limit for memory and for memory + swap IIRC
1397490611 M * benl http://linux-vserver.org/util-vserver:Cgroups
1397490645 M * benl So set both memory.memsw.limit_in_bytes and memory.limit_in_bytes to the same value?
1397490652 M * Bertl yes
1397490828 M * benl Great, thanks.
1397490831 M * benl I'll give this a shot now
1397490834 M * Bertl np
1397490942 Q * urbee Ping timeout: 480 seconds
1397491293 M * benl What is VXF_VIRT_CPU supposed to alter?
1397491310 M * benl It doesn't seem like it affects top or /proc/cpuinfo
1397491358 M * Bertl it affects the stat accounting
1397491403 M * Bertl so it should show up in top, but not in /proc/cpuinfo (which is just status information)
1397491446 M * benl I think I want  VIRT_CPU
1397491481 M * benl Nah, that's no use either
1397491517 M * benl I'm only really looking to verify if the cpuset is working as it should
1397491605 M * Bertl also check with: vattribute --xid <guest-xid> --get
1397491611 M * Bertl that the flag is actually active
1397491830 M * benl it is active
1397491840 M * benl I'm seeing the expected behaviour from `virt_mem`
1397492218 M * benl I'm able to replicate the load issue in the guest
1397492232 M * benl I'll swap over to a 3.10 kernel (from 3.2)
1397492296 M * benl There is no I/O wait during the load sprial
1397492362 M * benl CPU load is 100% idle
1397492380 M * Bertl on the host or on the guest?
1397492383 M * benl host
1397492390 M * benl But SSH is unresponsive
1397492394 M * benl serial is unresponsive
1397492410 M * benl Its just blind luck I left `vtop` running on serial console
1397492418 M * benl Load is at 104 now :|
1397492425 M * Bertl magic sysrq should work fine
1397492435 M * Bertl dump the cpus and memory allocations
1397492440 M * benl Barely works through ipmi
1397492460 M * Bertl get a proper console then :)
1397492466 M * benl Not possible!
1397492471 M * benl ipmi sol is as good as it gets
1397492575 M * benl looks like the guest isn't swapping at this point
1397492583 M * benl curious right?
1397492599 M * benl I'm going to switch kernels and see if I can still make it happen
1397492895 J * bonbons ~bonbons@2001:a18:209:4501:d817:bff4:9068:31f2
1397493269 M * benl Well
1397493280 M * benl Did a test on 3.10.33-vs2.3.6.8-beng
1397493309 M * benl Got to the same point, stalled, dumped a tonne of `Memory cgroup out of memory` to the serial console
1397493329 M * benl But recovered itself almost immediately - and load stayed normal
1397493343 M * benl Looks like it was a  kernel issue
1397493506 M * Bertl not too unexpected ... the early 3.x kernels were quite bad
1397493516 M * benl :(
1397493521 M * benl That's not good to hear
1397493663 M * laurens I've configured all my vservers with public ip addresses, but they can't access the internet it seems... Any idea whether I have to configure anything on my host (iptables is ok)
1397493689 M * benl No requirement for IPtables on the host (well, no NAT at least) if they are using public IPs
1397493700 M * Bertl please elaborate on "can't access the internet"
1397493806 M * laurens Bertl: I'm still checking but I can't access outside hosts...
1397493819 M * Bertl how do you check?
1397493882 M * laurens apt-get isntall :)
1397493885 M * laurens ping
1397493900 M * Bertl what do you ping? and what do you get when you ping?
1397493943 M * laurens for instance the google public dns servers
1397493958 M * Bertl what's the command you use?
1397493963 M * laurens ping 8.8.8.8
1397493968 M * laurens and I get no output
1397493976 M * laurens when I ctrl-c, I get all packet loss
1397493986 M * laurens I'll check further, maybe I missed something...
1397493989 M * Bertl okay, run 'tcpdump -vvnei ethX icmp' on the host
1397494006 M * laurens ok, ty
1397494012 M * Bertl then start the ping and see what tcpdump records, ethX is the outgoing interface
1397494664 Q * benl Quit:  HydraIRC -> http://www.hydrairc.com <- *I* use it, so it must be good!
1397497298 J * SteeleNivenson ~SteeleNiv@207.97.167.52
1397498056 J * zerick ~eocrospom@190.187.21.53
1397498188 J * fisted ~fisted@xdsl-81-173-189-133.netcologne.de
1397498607 J * fisted_ ~fisted@xdsl-87-78-181-48.netcologne.de
1397499062 Q * fisted Ping timeout: 480 seconds
1397499062 N * fisted_ fisted
1397502277 Q * SteeleNivenson Ping timeout: 480 seconds
1397502808 Q * click_ Read error: Connection reset by peer
1397502817 J * click click@ice.vcon.no
1397504138 J * SteeleNivenson ~SteeleNiv@pool-108-29-139-222.nycmny.fios.verizon.net
1397504445 J * Aiken ~Aiken@2001:44b8:2168:1000:21f:d0ff:fed6:d63f
1397505350 Q * macmaN Read error: Connection reset by peer
1397505388 J * macmaN ~chezburge@145.183.35.213.dyn.estpak.ee
1397506338 Q * bonbons Quit: Leaving
1397507953 Q * SteeleNivenson Quit: Leaving
1397507956 J * SteeleNivenson ~SteeleNiv@pool-108-29-139-222.nycmny.fios.verizon.net
1397516156 Q * Ghislain Quit: Leaving.