1345335002 J * PeteG ~vampirnat@77.118.247.191.wireless.dyn.drei.com 1345335076 Q * bonbons Quit: Leaving 1345335109 Q * PeteG_ Ping timeout: 480 seconds 1345336667 J * bergerx___ ~bergerx@178.233.12.223 1345337001 Q * bergerx_ Ping timeout: 480 seconds 1345342883 Q * nkukard Read error: Operation timed out 1345342935 J * nkukard ~nkukard@197.87.42.133 1345343297 M * Bertl off to bed now ... have a good one everyone! 1345343302 N * Bertl Bertl_zZ 1345349108 Q * fisted_ Quit: leaving 1345349117 J * fisted ~fisted@xdsl-87-78-83-193.netcologne.de 1345355835 J * yang_ yang@193.189.137.24 1345355835 Q * yang Remote host closed the connection 1345358875 J * theocrite ~Goddefroy@87.98.217.168 1345359006 P * theocrite 1345361156 Q * clopez Ping timeout: 480 seconds 1345364483 J * bonbons ~bonbons@2001:960:7ab:0:b154:a4ff:d420:b518 1345370800 J * fisted_ ~fisted@xdsl-87-78-11-216.netcologne.de 1345371219 Q * fisted Ping timeout: 480 seconds 1345375177 N * Bertl_zZ Bertl 1345375196 M * Bertl morning folks! 1345378187 J * ghislain ~AQUEOS@adsl2.aqueos.com 1345378744 Q * ghislain Quit: Leaving. 1345383146 Q * Aiken Remote host closed the connection 1345385871 Q * fisted_ Quit: leaving 1345388784 J * clopez ~clopez@65.27.165.83.dynamic.mundo-r.com 1345389164 M * Jb_boin is there a way to deactivate the vdu when doing a vserver start? 1345389220 M * Bertl I think yes, but please check with the 'flower page' or contact daniel_hozac 1345389458 Q * ensc|w Remote host closed the connection 1345389768 J * ensc|w ~ensc@www.sigma-chemnitz.de 1345391028 Q * ensc|w Remote host closed the connection 1345391126 J * ghislain ~AQUEOS@adsl2.aqueos.com 1345391336 J * ensc|w ~ensc@www.sigma-chemnitz.de 1345391462 M * Jb_boin because its pretty annoying when you reboot and have 10th of vservers that needs to be restarted quickly and they do a vdu on each one (knowing that we manually (re)calculate quotas each nights) 1345391801 J * fordfrog ~fordfrog@78.110.211.180 1345392007 M * fordfrog hi, i have issues with vserver and memory. recently i upgraded my server memory from 16gb to 32gb and since then i have problems with memory allocation in virtual server. though i have 16gb of free memory (and most of the used memory is used as file cache), i get errors on memory allocation. could anyone guide me so i could solve this issue? 1345392191 M * Bertl first, what system, kernel and util-vserver version? 1345392256 M * fordfrog Linux server 2.6.35-vs2.3.0.36.32-gentoo #3 SMP Sat Aug 18 22:02:40 CEST 2012 x86_64 Intel(R) Core(TM) i7-2600 CPU @ 3.40GHz GenuineIntel GNU/Linux, util-vserver-0.30.216_pre3025 1345392293 M * Bertl Jb_boin: certainly a quick and dirty hack would be to put a wrapper around vdu 1345392340 M * Bertl fordfrog: and the guest has any memory limits? 1345392347 M * fordfrog i have this non-default kernel memory configuration: http://pastebin.com/3Sbece0h 1345392356 M * fordfrog Bertl: no 1345392432 M * fordfrog Bertl: here is info on limits: http://pastebin.com/RhfLq2Gj 1345392434 M * Bertl well, never overcommit will lead to failing memory allocations eventually 1345392451 M * Bertl e.g. when you start java or similar for sure :) 1345392469 M * fordfrog even when having 16gb of free mem? 1345392491 M * Bertl IIRC, java allocate 2 or 4GB for every thread, by default 1345392505 M * Bertl so, you'll run out of memory pretty fast 1345392521 M * Bertl it doesn't actually use the memory for anything 1345392535 M * fordfrog here is current memory info: http://pastebin.com/7GUdtNKm 1345392559 M * fordfrog so now, when having 14gb of free mem, java cound not allocate even 2gb of mem? 1345392580 M * fordfrog (not counting 16gb used just for cache) 1345392583 M * Bertl actually you have 9GB 1345392594 M * Bertl CommitLimit:    33219008 kB 1345392599 M * Bertl Committed_AS:   24908236 kB 1345392629 M * Bertl but check _what_ fails, and see with strace -fF what it tries to allocate 1345392738 M * fordfrog Bertl: here is the output: http://pastebin.com/YNvXRTSc 1345392740 M * Bertl it's a shame that 'modern' applications do not do proper memory management but instead just assume that memory is unlimited 1345392758 M * Bertl but with overcommitment completely disabled you have to carefully select your software 1345392816 M * Bertl see the size it's allocating? 1345392827 M * fordfrog no, what line? 1345392834 M * Bertl 846 1345392838 M * Bertl 8GB 1345392868 M * Bertl and that's not the first allocation 1345392905 M * fordfrog this line? [pid 13833] rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0 1345392915 M * fordfrog ah, sry :-) 1345392915 M * Bertl [pid 13833] mmap(0x605000000, 8506048512, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS|MAP_NORESERVE, -1, 0) = -1 ENOMEM (Cannot allocate memory) 1345392928 M * Bertl size = 8506048512 1345392931 M * fordfrog y 1345392974 M * Bertl plus 50331648, plus 2555904 1345393030 M * Bertl and if you scan the trace, you'll see that quite a number of times such allocations already happened 1345393055 M * Bertl in total, it will be just above the 9GB you have left 1345393078 M * fordfrog how can i find amount of memory committed to each running process? to find out where the memory is committed but not used if i get it right 1345393116 M * Bertl the RSS will give the used part 1345393147 M * Jb_boin you can specify at commandline the memory allocation for java : -Xms512M -Xmx3000M for example 1345393205 M * fordfrog i know, i just wonder where all the memory went when just few gigs are reported to be used 1345393245 M * fordfrog if i get it right, the rest is committed but not used so just reserved and wasted 1345393254 M * Bertl correct 1345393272 M * fordfrog i will check my other java processes first 1345393402 M * fordfrog hm, setting -Xmx for two tomcat instances reduced the as quite a lot: Committed_AS: 9948596 kB 1345393505 M * fordfrog honestly i did not even know java by default allocates 8gb of memory in my case 1345393522 M * fordfrog + the extras 1345393531 M * Bertl yeah, java is really good for your memory :) 1345393644 M * fordfrog y, seems so, i should put all my java processes in cages :-) thanks a lot for help, i will have to tune my java processes now 1345393814 J * fisted ~fisted@xdsl-84-44-223-138.netcologne.de 1345393844 M * fordfrog hm, this is interesting, setting -Xms on java processes did not affect committed as memory at all 1345394436 M * Bertl you're welcome! 1345394499 M * Jb_boin what is strange is that you didnt hit the issue with less ram 1345394518 M * fordfrog i guess with more ram java decided it can allocate more and did so 1345394652 M * fordfrog before, one of my tomcat instances ended often with oom, so i had to increase -Xmx to 2g 1345394865 M * fordfrog just one more question, is committed memory used by kernel for caching files until it is really needed or it is never used for anything else once committed? 1345394918 M * Bertl never used for anything AFAIK 1345394927 J * ensc ~irc-ensc@p54ADED3B.dip.t-dialin.net 1345394969 M * fordfrog hm, thanks 1345395357 M * daniel_hozac Jb_boin: vdu only runs the first time you start the guest 1345395363 M * daniel_hozac subsequent restarts will use a cache 1345395718 M * Jb_boin well, it might be because it rebooted on a newer kernel then 1345395774 M * Jb_boin from a 2.6.32 from ubuntu semi official repo to a 3.2 from beng it did a vdu on every vservers 1345395905 M * daniel_hozac or you didn't shut them down nicely. 1345395985 M * Jb_boin yep, server crashed but i should be able to bypass vdu or at least that it doesnt stops vserver start if i kill vdu process 1345396072 M * Jb_boin knowing that the /vserver partition wasnt even tag mounted at boot :) 1345396820 M * daniel_hozac well, if it crashes, there's not really much you can do. 1345396973 Q * ntrs Ping timeout: 480 seconds 1345397318 J * ntrs ~ntrs@vault08.rosehosting.com 1345399010 Q * ensc|w Remote host closed the connection 1345399317 J * ensc|w ~ensc@www.sigma-chemnitz.de 1345402931 Q * nenolod Quit: brb 1345405039 Q * m_ueberall Read error: Connection reset by peer 1345408387 Q * bonbons Quit: Leaving 1345408428 J * Aiken ~Aiken@2001:44b8:2168:1000:21f:d0ff:fed6:d63f 1345411169 Q * ntrs Ping timeout: 480 seconds 1345411504 J * ntrs ~ntrs@vault08.rosehosting.com 1345412268 Q * ghislain Quit: Leaving. 1345413911 Q * Aiken Remote host closed the connection 1345414165 J * Aiken ~Aiken@2001:44b8:2168:1000:21f:d0ff:fed6:d63f