1412294401 Q * fisted Remote host closed the connection
1412294418 J * fisted ~fisted@xdsl-84-44-237-221.netcologne.de
1412294434 N * l0kit Guest311
1412294440 J * l0kit ~1oxT@0001b54e.user.oftc.net
1412294553 Q * Guest311 Ping timeout: 480 seconds
1412298441 Q * zerick Ping timeout: 480 seconds
1412312972 M * Bertl off to bed now ... have a good one everyone!
1412312980 N * Bertl Bertl_zZ
1412315604 J * zerick ~zerick@190.118.28.252
1412316177 J * Ghislain ~aqueos@adsl1.aqueos.com
1412323960 J * BenG ~BenG@cpc29-aztw22-2-0-cust128.18-1.cable.virginm.net
1412334575 M * eyck Is it possible that network namespaces are slower then even bridging? 
1412335676 M * Ghislain i do not use them but i guess this depnd on the kernel version, testing is the best way to be sure when in doubt
1412335713 M * Ghislain kernel 3.5 has perf regression on network namespace
1412335796 M * Ghislain http://www.opencloudblog.com/?p=96
1412335899 Q * Aiken Remote host closed the connection
1412337601 Q * fisted Remote host closed the connection
1412337618 J * fisted ~fisted@xdsl-81-173-188-155.netcologne.de
1412338060 J * benG__ ~bengreen@5751ac80.skybroadband.com
1412338127 Q * benG__ 
1412338856 N * Bertl_zZ Bertl
1412338866 M * Bertl morning folks!
1412344184 M * eyck Ghislan: does this regression persists until 3.9.4 or 3.10?
1412344186 M * eyck morning
1412344229 M * Bertl @slower, I'd suspect so, after all it basically includes bridging
1412344262 M * Bertl but the nic is local, so that should be a little faster than a real physical nic
1412344339 M * eyck Bertl: devs did some benchmarks, and in their benchmarks vserver with network namespaces turned out slower then openvz with bridges.
1412344363 M * eyck IIRC vserver had the fastest networking around, and bridges add their own delay too
1412344391 M * eyck that's why I'm suspecting net namespaces 
1412344400 M * Bertl yeah, they should have tested Linux-VServer with IP isolation
1412344412 M * Bertl which would have been the fastest of them all :)
1412344515 M * Bertl But I'm positively surprised that they didn't test KVM performance on a Linux-VServer kernel ...
1412344636 M * eyck yeah... you might be joking allright, but I do keep kvm on vserver kernel around ;) 
1412344854 M * Bertl yes, I do so too, but it has no relevance regarding Linux-VServer performance
1412344869 M * Bertl (and the same is obviously true regarding network namespaces)
1412344918 M * Bertl if network namespaces are particularily slow on a Linux-VServer kernel compared with a mainline kernel, then we have a problem which needs our attention
1412352460 J * _zerick_ ~eocrospom@190.118.28.252
1412352553 Q * _zerick_ 
1412353212 M * yoshi314_ maybe a stupid question. but are there any potential issues with hugepage support in guests? i might be doing something wrong on my end
1412353247 M * yoshi314_ when using transparent ones, it seems to work fine. i wanted to try old manual setup ones and i got problems with  java. 
1412353267 M * yoshi314_ although that could be because i forgot to disable transparent ones in kernel variables
1412353306 M * yoshi314_ not sure what happens in such situation
1412353696 M * Bertl there should be no problems with recent Linux-VServer kernels, given that your guest has all the required permissions
1412353742 M * Bertl but note that hugepages are probably not well tested, so if you get a test case which works on an unpatched kernel and fails on a Linux-VServer kernel, it is very likely a bug
1412353756 M * Bertl (which we would like to know about so that we can fix it :)
1412354963 Q * BenG Quit: I Leave
1412355606 M * yoshi314_ well for now i suspect java, still reading up on it
1412356221 M * Bertl please keep us updated what you figure out
1412356316 M * yoshi314_ btw is that normal that kernel.shmmax changes in the guest even though i did not set any sysctl parameters for instance ? 
1412356327 M * yoshi314_ unless i redefine it same way the host has
1412356370 M * Bertl changes when you do what?
1412356449 M * Bertl or does it suddenly change for no apparent reason?
1412356490 M * yoshi314_ it's different in the guest.
1412356519 M * yoshi314_ i've set it to ~30gb on host, it seemed to be set at 30 or 300mb in the guest (when looking in /proc)
1412356519 M * Bertl the guest gets a new space, so it might revert to the initial kernel settings
1412356553 M * yoshi314_ hm, so i would need to provide all custom sysctl variables in guest config
1412356582 M * Bertl or set it as .default in the util-vserver config
1412356599 M * yoshi314_ i just need it on one guest
1412356613 M * yoshi314_ not sure what would happen if each guest would define it
1412356621 M * Bertl okay :)
1412356625 M * yoshi314_ wrt to ram usage
1412356794 M * yoshi314_ also, the guests sysctl params are independent from hosts?
1412356846 M * Bertl yes, most of them should be, at least all which are covered by namespaces
1412362528 Q * ensc|w Remote host closed the connection
1412363002 J * ensc|w ~ensc@www-old.sigma-chemnitz.de
1412364333 J * Aiken ~Aiken@d63f.h.jbmb.net
1412364531 J * bonbons ~bonbons@2001:a18:20d:d701:dd2b:ce49:ce6a:fcd8
1412372854 M * Bertl off for a nap ... bbl
1412372855 N * Bertl Bertl_zZ
1412374743 Q * Defaultti Quit: Quitting.
1412374851 J * Defaultti defaultti@lakka.kapsi.fi
1412376083 Q * Ghislain Quit: Leaving.
1412377578 J * PowerKe_ ~tom@d515270C2.access.telenet.be
1412377988 Q * PowerKe Ping timeout: 480 seconds
1412378528 Q * bonbons Quit: Leaving