1138752088 M * Aiken Bertl which scsi driver do you use with your alpha? 1138752489 J * marl ~matt@84.92.193.226 1138753668 M * Bertl Aiken: 0000:00:05.0 SCSI storage controller: LSI Logic / Symbios Logic 53c875 (rev 26) 1138753700 M * Bertl mugwump: http://vserver.13thfloor.at/Experimental/del-2.6.16-rc1-vs2.1.0.9/ (preliminary) 1138753720 M * Bertl mugwump: i.e. you get a better one later ... 1138753756 M * Bertl nap attack ... i.e. I'm off for now .. but maybe back in a few hours ... 1138753773 N * Bertl Bertl_zZ 1138754570 M * mugwump thanks Bertl_zZ 1138754771 M * jkl whats up bertl! 1138755924 A * mugwump wonders why applying these quilt patches always ends up lots of "Hunk #1 succeeded at XXX (offset N lines)" 1138756093 J * VoUgIoS-4- ~vougioss@ppp7-adsl02-228.the.forthnet.gr 1138756130 Q * VoUgIoS-4- Quit: 1138757407 A * mugwump watches with glee as stgit re-bases Bertl's patch against 2.6-head 1138759639 N * monrad_ monrad 1138761027 Q * michal_ Ping timeout: 480 seconds 1138761489 J * michal_ ~michal@www.rsbac.org 1138762551 J * Johnnie ~jdlewis@dynamic-acs-24-154-53-16.zoominternet.net 1138765649 J * anonc ~anonc@staffnet.internode.com.au 1138768071 N * ebiederm ebiederm_oO 1138769028 N * Bertl_zZ Bertl 1138769037 M * Bertl back now ... 1138769083 M * Bertl mugwump: the patches are _all_ against the base kernel :) 1138769084 M * Skram Yey! 1138769087 M * Skram heh 1138769107 M * Bertl mugwump: so if you stack them, there will be 'offsets' 1138769126 M * Bertl hey Skram! 1138769145 M * Skram hey 1138770091 J * Vudumen_ vudumen@perverz.hu 1138770121 Q * Vudumen Read error: Connection reset by peer 1138770864 N * _Medivh Medivh 1138770887 M * Bertl hey Medivh! 1138772054 J * Loki_muh loki@satanix.de 1138772054 Q * Loki|muh Read error: Connection reset by peer 1138776757 M * Hollow morning 1138776782 M * Bertl good morning Hollow! 1138776817 M * Hollow hey Bertl, still fit? ;) 1138776827 M * Bertl well, I had a longer nap :) 1138776855 M * Hollow ah, ic :) 1138776861 M * Hollow what's the new ioctl for? 1138776871 M * Bertl which one? 1138776885 M * Hollow http://vserver.13thfloor.at/Experimental/patch-holders.diff 1138776900 M * Bertl ah, that's just a test for the debian folks ... 1138776905 M * Hollow ok 1138776932 M * Bertl kilian/waldi complained that the e2fsck screwed up the filesystem (because it was mounted in a different namespace) 1138776962 M * Bertl which unfortunately is true, as the fsck does not open the device O_EXCL as we found out 1138776979 M * Bertl (but isntead relies to parsing /proc/mounts and similar voodoo) 1138776985 M * Bertl s/to/on 1138776990 M * Hollow bad fsck 1138777005 M * Bertl yeah, well, we first thought it is a kernel flaw 1138777008 M * Hollow hm, do we need this fix too in gentoo? 1138777024 M * Hollow i mean.. is it a upstream bug? 1138777025 M * Bertl the ioctl is not required, but the fsck should be 'improved' 1138777034 M * Hollow ok 1138777037 M * Hollow good to know 1138777059 M * Bertl got no feedback from kilian/waldi yet ... but I guess they will also push it upwards 1138777069 M * Bertl let me dig out the tests if you are interested 1138777126 M * Hollow well, the bug is quite clear (now you found it ;) 1138777151 M * Hollow so, if they push it upstream, all seems fine.. ;) 1138777193 M * Bertl okay, in case they don't and you need tests for gentoo (to push too), just let me know 1138777584 M * Hollow yup, thx 1138777595 M * anonc Hollow: when using bind mounted portage ina gentoo guest (ro with an rw distfiles) running under active-active drbd (one vserver on each node), it looks like it doesn't get unmounted properly when heartbeat stops the vserver (using the heartbeat vserver control script from the vserver mailing list). The problem doesn't happen when bind mounting only portage rw (without a seperate distfiles mount). Do you think this fix might help? 1138777636 M * Hollow well, it has to do with fsck, not with mount 1138777650 M * Hollow probably the order you try to umount those is wrong 1138777708 M * anonc the symptom is heartbeat fencing the node (rebooting it) to free up the drbd resource so that the second node can take over. What order do the vserver scripts unmount guest bind mounts? 1138777742 M * Hollow iirc they should do it in reverse order 1138777753 M * anonc thats what I would expect 1138777783 M * Hollow does it fail too if you restart with manually? 1138777798 M * Bertl anonc: which patch version? 1138777896 M * anonc Bertl: gentoo-0 2.6.15-vs2.1.0.4-gentoo which I think is the latest ebuild. I haven't tried with .9 yet (I will if you think it is worth it). 1138777917 M * Bertl no, that should be fine 1138777958 M * anonc Hollow: i can use the gentoo init scripts to stop the vserver on one node and bring it up on the other (after taking ownership of the relevant drbd partition). 1138778037 M * Hollow so, manually it works fine.. 1138778068 M * Hollow i neither know drbd nor heartbeat, so i can't help you with those 1138778131 M * Bertl anonc: chances are good that your drbd device is mounted (an thus claimed) in more than one guest 1138778189 M * Bertl anonc: the thing is, when you mount the drbd partition outside the guest's namespace, it will automagically be copied into every newly started guest 1138778201 M * anonc It happens when both nodes are running heartbeat, each vserver has a different home node but I have forced both vservers onto the same node (by rebooting the other). When it attempts failback of one of the vservers (which essentially involves heartbeat running 'vserver stop', it seems to think that the drbd device that the vserver was/is on is still in use. I was guessing it is related to the namespace issues with mounts inside vservers. 1138778235 M * anonc Hollow: incidentally thank you for making gentoo so easy to use with vservers 1138778244 M * Hollow :) 1138778251 M * Hollow you're welcome 1138778254 Q * matti Ping timeout: 480 seconds 1138778281 M * anonc Bertl: ah. so possibly the second active vserver could be locking it? 1138778295 M * Bertl very likely ... 1138778334 M * anonc hmm - then why does making portage a single rw bind mount inside the guest make the problem go away :) 1138778365 M * Bertl no idea :) 1138778367 M * anonc i suspect a timing issue with heartbeat. i might experiment with some timeouts 1138778419 M * anonc Hollow: is the reason for the -nptl glibc in the vserver stage3 just for maximum portability? 1138778609 M * Hollow no, it's default 1138778617 M * anonc ah 1138778648 M * Hollow with the profile move to default-linux/ the stages should be exactly the same, except baselayout 1138778861 M * anonc weird - recompiled gentoo vserver guest with +nptl +nptlonly yet running /lib/libc-2.3.5.so doesn't list Native POSIX Threads Library - only linuxthreads 1138778979 Q * Johnnie Quit: G'bye! 1138779007 J * Johnnie ~jdlewis@dynamic-acs-24-154-53-16.zoominternet.net 1138779078 M * phycho hmmm 1138779082 M * phycho should i change over to 'freevps'? 1138779116 M * anonc i wouldn't - but you may as well try both to see which suits you better 1138779127 M * phycho any particular reason why? 1138779139 M * phycho i want swap/virtual memory in servers. 1138779152 M * phycho i dont like the idea that it kills processes over the memory limit 1138779161 M * phycho i would rather it ran into virtual swap instead 1138779193 M * FaUl phycho: bertl rescently explained why you don't want this 1138779201 M * anonc what do you propose the job of virtual swap should be? 1138779217 M * anonc swapping means hitting the disk 1138779219 M * phycho faUl - i still want it 1138779227 M * anonc that kills things for every guest 1138779240 M * phycho anonc - because theres not enough ram in the system to just put everyones memory allowance _above_ what they should get. 1138779257 M * phycho aka if i limit them to 128mb, i want them to be able to go over it slightly 1138779261 M * phycho without it killing processes 1138779270 M * phycho but i dont want the overusage to go into real memory 1138779273 M * phycho i want it in swap 1138779310 M * phycho see what im trying to do now? 1138779376 M * FaUl phycho: yes, but you'll get the effect that the box will swap around a lot if 2 or 3 are above that swaplimit 1138779393 M * phycho i know.. but the swap wont be that large 1138779393 M * FaUl which actually will slow down the whole mashine significantly 1138779399 M * phycho probably 30megabytes at most. 1138779419 M * phycho is there a way to set softlimits in vserver? 1138779423 M * FaUl but why don't you allow everyone to use 30mb more? 1138779423 M * phycho i could set a soft limit of 128mb 1138779426 M * phycho but hard limit of 150? 1138779430 Q * Johnnie Quit: G'bye! 1138779434 M * Bertl phycho: yes, I explained that to you, remember? 1138779435 M * phycho faUl - because i dont want them all to use 30mb more of _real_ memory. 1138779441 A * phycho scrolls up 1138779449 M * FaUl hey Bertl 1138779456 M * Bertl hey FaUl! 1138779464 J * Johnnie ~jdlewis@dynamic-acs-24-154-53-16.zoominternet.net 1138779494 M * anonc can one dynamically adjust the token bucket? i'd probably use that if possible to starve them of cpu cycles if they go over a designated rss 1138779495 M * phycho k i see it 1138779511 M * Bertl anonc: yes, vsched does that 1138779518 M * phycho any way to set a vps to a certain amount of cpu? 1138779528 M * phycho i dont want one vps to hog cpu =) 1138779537 M * Bertl phycho: yes, with the token bucket scheduler 1138779541 M * phycho k 1138779543 A * phycho goes to google 1138779564 M * Bertl http://linux-vserver.org/Scheduler+Parameters 1138779573 M * Bertl phycho: FreeVPS does virtual swap? 1138779581 M * phycho yep 1138779587 M * phycho . If swap is available in VPS, is there any way to let all VPS share the system swap? 1138779587 M * phycho Starting from version 1.1, FreeVPS supports swap on virtual servers. 1138779594 M * phycho =) 1138779601 M * phycho http://www.freevps.com/docs/faq.html 1138779613 M * Bertl well, that doesn't say much, we support swap on guests too :) 1138779622 M * phycho heh 1138779628 M * phycho resident/virtual memory limit (RSS Limit); 1138779628 M * phycho - entire memory limit (Memory Limit) 1138779633 M * phycho you can set those though. 1138779650 M * phycho When a VPS exceeds RSS Limit, it starts to use the host's swap. 1138779656 M * phycho :p 1138779696 M * Bertl good, give it a try and let me know how it works for you ... 1138779717 M * phycho im going to =) 1138779721 M * phycho it will be under heavy load 1138779726 M * phycho because im offering free VPS's 1138779737 M * phycho so should be a lot of people trying it 1138779775 M * Bertl hmm, rh 2.6.9 kernel ... interesting 1138779838 M * phycho bah.. they only have redhat rpm's 1138779839 M * phycho useless. 1138779858 A * phycho hunts deeper 1138779873 M * phycho nope i see a 2.6.x kernel src =) woo 1138779890 M * phycho huh.. i see a freebsd jail patch.. wtf 1138780109 M * Bertl url? 1138780127 M * Bertl not for the freebsd, for the freevps patch/kernel 1138780144 M * phycho http://www.freevps.com/download/snapshots/ 1138780163 M * Bertl ah, yes, I'm looking at that, it's against the beforementioned rh kernel 1138780167 M * phycho yeah 1138780183 M * phycho i dont fancy using *puke* redhat 1138780189 M * phycho im going to try patch a debian src with it 1138780198 M * Bertl good luck :) 1138780206 M * phycho ill try it justnow and let you know how it goes 1138780218 M * Bertl make that :) 1138780259 M * phycho oh oh 1138780261 M * phycho power cut. 1138780262 M * phycho hahahha 1138780279 M * phycho hadnt hooked server up to ups yet :x 1138780280 M * phycho hope its ok 1138780293 M * phycho rest of my network is all on UPS =) 1138780296 M * phycho so i can stay online hehe 1138780384 Q * FireEgl Ping timeout: 480 seconds 1138780435 Q * cehteh Ping timeout: 480 seconds 1138780465 M * phycho thats strange 1138780470 M * phycho my neighbours lights are on.. 1138780486 M * phycho brb im away to check circuit breakers 1138780518 N * Smutje_ Smutje 1138780997 M * phycho bk 1138781004 M * phycho main one had just went 1138781344 M * Bertl well, could have been worse ... 1138781352 M * phycho hehe yeah 1138781359 A * Bertl was 18 hours without power last month 1138781360 M * phycho all my stuff is on ups anyway 1138781364 M * phycho lol ouch 1138781369 M * phycho ive never had a proper outage 1138781374 M * phycho its pretty reliable in the uk 1138781385 M * Bertl well, we thought so in austria too :) 1138781399 M * phycho hehe 1138781421 M * phycho i _really_ should upgrade the kernel on my gateway 1138781425 M * phycho its running an insecure pile of crap 1138781425 M * phycho lol 1138781435 M * Bertl IPC? 1138781447 M * phycho its 2.6.9 standard 1138781453 M * phycho im sure some vulns came out after that.. 1138781459 M * phycho olddd 1138781556 M * phycho http://www.freevps.com/download/snapshots/freevps26-1.5.5-11.diff.gz 1138781559 M * phycho is that 2.6.9 patchset 11 1138781559 M * phycho or 1138781562 M * phycho 2.6.11 patchset 9? 1138781573 M * Bertl it's against 2.6.9 1138781578 M * phycho k 1138781590 M * phycho thats what i figured but wasnt sure 1138781611 A * phycho probably shouldnt be doing this on a non-devlopment server 1138781613 M * phycho its live :p 1138781986 M * phycho oh oh 1138781986 M * phycho Hunk #16 succeeded at 1390 (offset -14 lines). 1138781986 M * phycho 3 out of 16 hunks FAILED -- saving rejects to file net/socket.c.rej 1138782063 M * phycho any ideas where i can get 2.6.9.11 1138782080 M * phycho its not on kernel.org anymore 1138782214 M * phycho bah cant find it 1138782217 A * phycho cant be bothered 1138782248 M * Bertl I don't think there was a vanilla 2.6.9.11 release 1138782278 M * phycho there was 1138782279 M * phycho i had it 1138782281 M * phycho ages ago 1138782285 M * Bertl really? 1138782287 M * phycho yep 1138782291 M * phycho i remember having 2.6.9.11 1138782296 M * phycho whether it was a patch or not 1138782297 M * phycho i cant remember 1138782303 M * phycho i think it might have been 2.6.9 with a patch on top 1138782328 M * phycho patch-2.6.11.4 1138782336 M * phycho nm 1138782393 M * phycho cant find it anywhere ;/ 1138782642 M * phycho openvps based on vserver as well? 1138782649 M * Bertl yep 1138782660 M * phycho heh.. how many different ones are there 1138782673 M * Bertl openvps is something different :) 1138782677 M * phycho ah 1138782679 M * phycho is it any good? 1138782694 M * Bertl well, it is supposed to be(come) a frontend 1138782700 M * phycho ah 1138782701 M * phycho sweet 1138782742 M * phycho any screenshots anywhere? 1138782764 M * Bertl well, probably on their site :) 1138782770 M * phycho having a look 1138782783 M * phycho it will be handy for what im trying to do 1138782784 M * phycho =) 1138782790 M * phycho aka http://www.darktech.org.uk 1138782852 M * Bertl well, good for you :) 1138782852 M * phycho http://www.openhosting.com/utility 1138782855 M * phycho thats damned sweet 1138782855 M * phycho =) 1138783111 Q * shedi Quit: Leaving 1138783221 J * Smutje_ ~Smutje@xdsl-87-78-18-236.netcologne.de 1138783329 Q * Smutje Ping timeout: 480 seconds 1138783329 N * Smutje_ Smutje 1138783446 J * matti matti@linux.gentoo.pl 1138783607 M * phycho http://openvz.org/documentation/tech/features 1138783611 M * phycho that loosk nice too 1138783639 M * Bertl yup, it's basically advertizement for their commercial release 1138783644 M * phycho i know 1138783654 M * phycho i wont touch the commercial crap with a barge pole 1138783658 M * phycho might give the free one a go t hough 1138783678 M * Bertl make that, and again, please let me know how it goest 1138783681 M * Bertl *goes 1138783683 M * phycho hehe k 1138783693 M * phycho hopefully its easy =) 1138783861 M * phycho gonna do it on centos though 1138783994 M * phycho hopefully i can migrate my web/mail vserver over to it with no trouble 1138784445 M * phycho will have centos in 39mins :-) 1138784597 J * startx ~startx@p508F3F43.dip0.t-ipconnect.de 1138784628 M * Bertl welcome startx! 1138784636 M * startx hiya 1138785649 M * phycho bertl - openvz has debian templates now too 1138785650 M * phycho =) 1138785668 M * Bertl phycho: I know :) 1138785707 M * phycho its based on 2.6.8 though =( 1138785718 M * phycho i wanted a newer kernel 1138785718 M * Bertl phycho: I know that too :) 1138785764 M * phycho would it still be secure running such an old kernel? 1138785776 M * phycho i dont fancy my box being rooted. 1138785778 M * Bertl well, they argue that older kernels are more secure ... 1138785786 M * phycho lol 1138785805 A * phycho hopes the packages on openvmz.org are all patched up :x 1138785892 M * phycho ill give you root on one to mess about with it 1138785897 M * phycho so you can check it out 1138785979 M * Bertl good, tx 1138785986 M * phycho np 1138786823 J * dothebart ~willi@xdsl-213-196-242-5.netcologne.de 1138787009 M * phycho bertl - do you know how the openvz filesystem works/ 1138787013 M * phycho is it a file based mount 1138787018 M * phycho or is it like vserver just a folder? 1138787053 M * Bertl if I got that right, it's some kind of symlink copy of an original 1138787068 M * phycho what do you mean 1138787074 M * phycho copy of an origional 1138787084 M * phycho dd file mount? 1138787088 M * phycho or what 1138787092 M * Bertl for each file, there is a symlink pointing to the 'real' file 1138787098 M * phycho ah 1138787102 M * phycho so it symlinks to the host os files? 1138787106 M * phycho that cant be right 1138787112 M * phycho since you can install different distros 1138787114 M * Bertl no, to the templates 1138787118 M * phycho ah 1138787127 M * phycho so its just folders/symlinks then? 1138787139 M * phycho no virtual mounts with flat files? 1138787141 M * Bertl should be 'special' symlinks 1138787148 M * phycho k 1138787158 A * phycho doesnt want to have to create virtual devices/drives with dd. 1138787166 M * phycho so im hoping it doesnt use that =) 1138787168 M * Bertl but best would be to ask the developers on #openvz (freenode) 1138787174 M * phycho k 1138787217 M * phycho was already on that network =) 1138787260 M * phycho im hoping its resource uses are going to be as good as vserver is 1138787263 M * phycho its great 1138787264 Q * tudenbart Ping timeout: 480 seconds 1138787269 M * phycho just not good enough in terms of resource management 1138787278 M * phycho otherwise id be staying with it. 1138787307 M * Bertl what is not good enough? 1138787322 M * phycho cpu limiting, memory, open files 1138787329 M * Bertl in openvz? 1138787334 M * phycho openvz has a far better way of doing it 1138787336 M * phycho no, in vserver. 1138787350 M * Bertl ah, interesting ... well, how is the openvz way? 1138787391 M * phycho sec 1138787414 M * phycho http://openvz.org/documentation/tech/features 1138787474 M * Bertl sec 1138787487 M * Bertl http://linux-vserver.org/Release+FAQ 1138787488 A * phycho is lookin for the actual info on how to set it 1138787496 M * Bertl (scroll down to the feature matrix) 1138787523 M * Bertl http://linux-vserver.org/Resource+Limits 1138787530 M * Bertl http://linux-vserver.org/Scheduler+Parameters 1138787575 M * phycho bah cant find it.. 1138787610 M * Bertl well, so we 'know' that they are 'better' but not in what regard or how ... LOL 1138787629 M * phycho i had found it erlier 1138787632 M * phycho and thats why i said they are better 1138787636 M * phycho but cant find it again 1138787638 M * phycho haha 1138787795 M * phycho user_beancounters /proc settings 1138787809 M * phycho just cant find the info on setting it up.. 1138787828 M * Bertl well, did you look at the links I pasted? 1138787830 M * phycho found it 1138787831 M * phycho http://forum.openvz.org/index.php?t=rview&goto=743&th=148#msg_743 1138787839 M * phycho im about to 1138787858 M * phycho the post in that forum shows what can be limited 1138787879 M * Bertl I hope they can do more :) 1138787897 M * phycho heh its pretty much the same 1138787905 M * phycho between vserver + openvz 1138787913 M * Bertl nah, they have 3 dummies there 1138787923 M * phycho only difference as far as i can see is openvz does have swap 1138787929 M * Bertl 4 even 1138787944 M * phycho whats the dummies for? 1138787957 M * Bertl don't ask me, ask the openvz folks :) 1138787963 M * phycho hehe 1138787972 M * phycho i guess 'dummy' is where they are going to be adding new features 1138787973 M * phycho or something 1138787976 M * phycho aka, just a placeholder 1138787981 M * phycho or at least, thats just a guess 1138787987 M * Bertl just to make the list longer :) 1138787990 M * phycho lol 1138788004 M * phycho its quite easy to set options tho 1138788006 M * phycho look further down 1138788012 J * shedi ~siggi@tolvudeild-196.lhi.is 1138788019 M * phycho vzctl set id --option limit 1138788064 M * Bertl which of course is much easier than: 1138788091 M * Bertl vlimit --xid -H --