1208569010 J * cryptroni1 ~oli@p54A3BF3C.dip0.t-ipconnect.de 1208569353 Q * cryptronic Ping timeout: 480 seconds 1208569685 J * FireEgl FireEgl@adsl-226-25-164.bhm.bellsouth.net 1208571075 Q * Infinito Remote host closed the connection 1208572024 Q * hparker Quit: Read error: 104 (Peer reset by connection) 1208573024 J * hparker ~hparker@linux.homershut.net 1208575917 Q * hparker Quit: g'nite 1208578806 J * ktwilight ~ktwilight@87.66.200.207 1208579058 Q * ktwilight_ Ping timeout: 480 seconds 1208580839 Q * balbir Ping timeout: 480 seconds 1208582571 J * balbir ~balbir@122.167.181.56 1208583613 J * ktwilight_ ~ktwilight@87.66.197.44 1208583613 Q * ktwilight Read error: Connection reset by peer 1208583694 J * Satriani ~Nuhks@bl7-134-213.dsl.telepac.pt 1208584109 Q * Linus Ping timeout: 480 seconds 1208587182 J * dna ~dna@19.Red-81-35-110.dynamicIP.rima-tde.net 1208587727 Q * dna Quit: Verlassend 1208589464 J * gce2 ~gce2@75.144.228.105 1208589502 J * JonB ~NoSuchUse@77.75.164.169 1208590352 Q * _gh_ Read error: Connection reset by peer 1208590373 J * _gh_ ~gerrit@c-67-169-199-103.hsd1.or.comcast.net 1208590978 Q * nerdpunk Ping timeout: 480 seconds 1208591403 J * bonbons ~bonbons@2001:960:7ab:0:2c0:9fff:fe2d:39d 1208591568 Q * JonB Quit: This computer has gone to sleep 1208594182 J * ntrs ~ntrs@77.29.64.91 1208594211 M * ntrs why are dynamic contexts disabled in the latest patches? Can they be enabled back? 1208594305 M * Genghis- hey guys 1208594341 M * Genghis- i just tried to apply the new ipv6 patch ipv6-2.6.22.19-vs2.2.0.7-grsec2.1.11-20080415.diff 1208594360 M * Genghis- but it has stuff like this 1208594362 M * Genghis- |+++ linux-2.6.22.18-g-v-ipv6/Documentation/vserver/debug.txt 1208594375 Q * Satriani Quit: I'll by back 1208594377 M * Genghis- so it wont patch 1208594394 Q * balbir Ping timeout: 480 seconds 1208594401 J * Linus ~Nuhks@bl7-134-213.dsl.telepac.pt 1208594448 J * balbir ~balbir@122.167.181.56 1208595121 M * zbyniu ntrs: it's disabled because it's evil 1208595460 N * DoberMann[ZZZzzz] DoberMann 1208595580 N * ensc Guest714 1208595581 Q * Guest714 Remote host closed the connection 1208595590 J * ensc ~irc-ensc@77.235.182.26 1208596340 Q * ntrs Ping timeout: 480 seconds 1208597030 Q * balbir Read error: Operation timed out 1208598061 Q * duckx Remote host closed the connection 1208598223 J * duckx ~Duck@81.57.39.234 1208598308 J * JonB ~NoSuchUse@77.75.164.169 1208599219 Q * gce2 Ping timeout: 480 seconds 1208599816 N * Bertl_zZ Bertl 1208599820 M * Bertl morning folks! 1208599827 M * JonB hi 1208599916 J * larsivi ~larsivi@144.84-48-50.nextgentel.com 1208602915 J * Piet ~piet@tor.noreply.org 1208603717 Q * JonB Quit: This computer has gone to sleep 1208604168 J * Infinito ~argos@201-3-16-77.gnace701.dsl.brasiltelecom.net.br 1208604338 J * JonB ~NoSuchUse@77.75.164.169 1208604890 Q * JonB Ping timeout: 480 seconds 1208605346 J * JonB ~NoSuchUse@77.75.164.169 1208605548 Q * JonB 1208606359 J * docelic ~docelic@78.134.195.102 1208606747 M * Genghis- nice 1208606752 M * Genghis- problem 1 is not solved yet 1208606756 M * Genghis- and i allready have another one 1208606758 M * Genghis- elenet-Stats Up To [19/04/2008] 1208606758 M * Genghis- (13:51:15) (+Genghis) Used: [63159MB] of [66560MB] (95%) (••••••••••) Left: [3401MB] 1208606758 M * Genghis- (13:51:15) (+Arrigi) van de mijne weet ik het 1208606762 M * Genghis- damn 1208606766 M * Genghis- wrong paste ;)\ 1208606781 M * Genghis- Cannot change ownership to uid 201, gid 201: Disk quota exceeded 1208606791 M * Genghis- when trying to install an application 1208606797 M * Genghis- i have no quota set 1208606801 M * Genghis- i'm root 1208606807 M * Genghis- and no vserver dlimit set 1208606845 M * Genghis- /dev/hdv1 20G 1.1G 18G 6% / 1208606846 J * Alexander ~main@217.172.56.88 1208606930 Q * Aiken Remote host closed the connection 1208606987 M * Alexander hello. tell please, who tried to install DirectAdmin on vserver. Need help with quota 1208607576 M * Alexander http://paste.linux-vserver.org/12026 How to solve a problem ? Thanks 1208608202 N * DoberMann DoberMann[PullA] 1208608256 M * Genghis- you did this right? : http://linux-vserver.org/Frequently_Asked_Questions#Is_there_a_way_to_implement_.22user.2Fgroup_quota.22_per_VServer.3F 1208608468 M * Alexander yes 1208608468 M * Alexander run "vrsetup /dev/vroot1 /dev/md1" (/vservers partition) 1208608468 M * Alexander add "/dev/hdv1 / ufs rw,usrquota,grpquota 0 0" to /etc/vservers/vps/apps/init/mta 1208608468 M * Alexander add "quota_ctl" to /etc/vservers/vps/ccapabilities 1208608468 M * Alexander run "cp -af /dev/vroot1 /vservers/vps/dev/hdv1" 1208608578 M * Genghis- dunno then 1208608668 M * Bertl Alexander: what does /etc/mtab inside the guest contain? 1208608706 M * Alexander Bertl: srv:/# cat /etc/mtab 1208608706 M * Alexander none /proc proc defaults 0 0 1208608706 M * Alexander none /dev/pts devpts gid=5,mode=620 0 0 1208608725 M * Alexander srv:/# cat /etc/mtab 1208608725 M * Alexander "/dev/hdv1 / ufs rw,usrquota,grpquota 0 0 1208608725 M * Alexander none /proc proc defaults 0 0 1208608725 M * Alexander none /dev/pts devpts gid=5,mode=620 0 0" 1208608751 M * Bertl okay, and the filesystem used for that partition (/dev/md1) is? 1208608789 M * Alexander "/dev/md1 /vservers ext3 rw,tagxid 0 0" 1208608806 M * Bertl okay, you want to mount that with the quota option too 1208608846 M * Bertl and you are aware, that you cannot use a shared partition if you want user/group quota for the guest? 1208608965 M * Alexander yes, I set up, as is told in the instruction "http://linux-vserver.org/Quotas" 1208608996 M * Bertl okay, just because md1 is /vservers 1208609017 M * Bertl I'd expected that to be /vservers/guest1 or so 1208609132 M * Alexander I have correctly understood, what is required to create separate partition on a node and mount to /vservers/vps? 1208609176 M * Bertl is that a question? 1208609333 M * Alexander yes. /dev/md1 mount on /vservers (here are placed all VPS). I cannot use quota on "/vservers/guets" now? 1208609367 M * Bertl well, let's assume you have a bunch of guests, and you want to have quota on two of them 1208609400 M * Bertl then you could put all but those two on /dev/md1 (under /vservers/guest3,4,5,6 ...) 1208609448 M * Bertl but the two guests you want user/group quota on, should go to something like /dev/md2 and /dev/md3, mounted on /vservers/guestA and /vservers/guestB 1208609479 M * Bertl otherwise your user/group quota will get mixed with the other guests (which probably isn't what you want) 1208609502 M * Bertl nevertheless, you need to mount the partition for guests with u/g quota with the quota option if you are using ext3 1208609549 M * Alexander Bertl: many thanks and excuse for my English. I have understood now 1208609569 M * Bertl np, I'm no native speaker either ... 1208609849 M * heanol Bertl: are you a developer? 1208609859 M * heanol of vserver 1208609863 M * Bertl yes 1208609870 M * heanol cool, you're very helpful :) 1208609877 M * Bertl thanks 1208610529 Q * snooze Ping timeout: 480 seconds 1208611377 Q * pmjdebruijn Remote host closed the connection 1208611445 N * DoberMann[PullA] DoberMann 1208611614 J * pmjdebruijn ~pmjdebrui@jester.pcode.nl 1208613253 Q * Alexander Ping timeout: 480 seconds 1208613609 J * JonB ~NoSuchUse@77.75.164.169 1208615044 M * infowolfe Bertl, ping? 1208615095 A * infowolfe asks nicely about 2.6.25-vs 1208615366 J * MatBoy ~MatBoy@wiljewelwetenhe.xs4all.nl 1208615389 M * Bertl infowolfe: will take a little 1208615563 M * infowolfe Bertl, if you don't mind me asking, what's 'a little' 1208615569 Q * Infinito Quit: Leaving 1208615991 M * infowolfe Bertl, basically the issue is that if i put any load whatsoever on kernels between 2.6.20.*-2.6.24.*, my boxes like to hardlock without actually outputting _why_ they're breaking 1208615995 M * infowolfe they just stop responding 1208616073 M * daniel_hozac and you think 2.6.25 has the magical fix? 1208616090 M * infowolfe s/think/know/ 1208616128 M * infowolfe been running 2.6.25 on identical hardware since 2.6.25-rc4-git7, under higher load without issue 1208616160 M * daniel_hozac so, run a git-bisect, find the fix, backport it. 1208616167 M * infowolfe lol 1208616217 M * infowolfe since it takes a minimum of 4 days for the machine to go AWOL, and i'd rather not have my main monitoring box (the one running a -vs kernel) go AWOL, methinks not 1208616364 M * infowolfe since from what i've seen, it could take up to 20 reboots in order for me to find the 'bad' patch via git-bisect, that's pretty close to 3 months minimum for me to find the 'bad' patch, for a kernel that no longer matters, because 2.6.25 doesn't exhibit the same issues 1208616466 M * daniel_hozac and it's always better getting other people to fix your issues... 1208616756 M * infowolfe actually, it's better just using a kernel that doesn't suck (as bad) 1208616774 Q * esa` Ping timeout: 480 seconds 1208616800 Q * JonB Quit: This computer has gone to sleep 1208616801 M * infowolfe when a kernel that's >1yr old (2.6.19.7) is the last kernel that is considered stable across all hardware run at a particular site, that means that someone's been slacking on the QA 1208616823 M * infowolfe but nobody wants to do QA, since it's not glamorous, so instead they just add more features 1208618065 M * Bertl you are testing 2.6.25 for several days now? (at least more than 4) 1208618370 M * infowolfe since rc4-git7 1208618394 M * infowolfe longest uptime was ~3wks prior to me upgrading/rebooting 1208618402 M * Bertl so how can you be sure that 2.6.25 works? 1208618430 M * infowolfe i've abused it rather thoroughly :) 1208618439 M * Bertl the last two days, yes? 1208618457 M * infowolfe if the last -rc prior to final is good, the final is usually good 1208618485 M * daniel_hozac wasn't the last rc9? 1208618525 M * infowolfe last one we touched was -rc8 1208618531 M * infowolfe (prior to final) 1208618579 M * infowolfe oh, and daniel_hozac, a big indicator of quality is if it kicks an OOPS in the first 4 days the box is up 1208618636 M * Bertl and it didn't (yet)? 1208618741 M * daniel_hozac you must be using some fancy drivers... 2.6.22 has been rock stable for me for months. 1208618763 M * Bertl yes, indeed, 2.6.22 ist very stable 1208618807 M * infowolfe daniel_hozac, i think you just don't have much load 1208618824 M * Bertl the changes between 2.6.23 and 24 as well as 24 and 25 are quite huge, and I'm expecting major issues there (just because of those changes) 1208618836 M * daniel_hozac what kind of load are you talking about? 1208618857 M * pmjdebruijn infowolfe: may I see your server 'lspci' ? 1208618861 M * infowolfe ~600CPS per box (dynamic webserving) 1208618871 M * daniel_hozac CPS? 1208618882 M * infowolfe connections per second, as measured by ipvs 1208618882 M * Bertl connections per second 1208618895 M * daniel_hozac that doesn't really say anything about the load... 1208618911 M * Bertl well, I have a machine running which does 12000 page impressions per second :) 1208618911 M * daniel_hozac network load, yes, but load usually refers to processes. 1208618916 M * pmjdebruijn we've had quadcore here tolerate load 35 without problems 1208618977 M * infowolfe pmjdebruijn, i'm not talking about loadavg, since loadavg of >1.0 * cores shows you're having contention issues 1208618981 M * infowolfe pmjdebruijn, http://pastebin.com/d1fb5594c 1208618997 M * pmjdebruijn contention? 1208619001 M * daniel_hozac not really. 1208619017 M * infowolfe or your disk subsystem is woefully overworked 1208619024 M * pmjdebruijn infowolfe: is that a real server? or just a converted desktop? 1208619043 M * pmjdebruijn infowolfe: indeed, our SAN is doing overtime 1208619057 M * infowolfe pmjdebruijn, tyan s2927g2nr mobo w/ 2x opteron 2352's (was 2210's until yesterday) 1208619083 M * Bertl nVidia chipser, I'm not surprised :) 1208619084 M * pmjdebruijn ah 1208619088 M * pmjdebruijn indeed 1208619088 M * Bertl *chipset 1208619104 M * infowolfe Bertl, can you suggest a better chipset for amd opteron? 1208619113 M * pmjdebruijn I've only seems them in 1337 sp33dy g4m3r setups :) 1208619131 M * infowolfe pmjdebruijn, you haven't bought any amd gear lately then? 1208619160 M * pmjdebruijn (/my apolagizes for the lame remark) 1208619160 M * pmjdebruijn infowolfe: we use HP ProLiant DL3{60|80} G5's 1208619160 M * pmjdebruijn which are Xeons 1208619178 M * infowolfe pmjdebruijn, i feel sorry for your power infrastructure :-p 1208619183 M * pmjdebruijn uhr 1208619187 M * pmjdebruijn infowolfe: why? 1208619204 M * infowolfe we only have one xeon box... draws freaking 5A just sitting in the bios :( 1208619213 M * infowolfe 5A @ 110VAC 1208619218 M * pmjdebruijn infowolfe: a Pentium 4 based Xeon perhaps? 1208619220 M * Bertl their chipsets are as good as the nVidia graphics cards and drivers 1208619228 M * infowolfe pmjdebruijn, try E5462 1208619255 M * Bertl you know that the nvidia drivers (for windows) have a watchdog which reboots the graphics card on a regular basis? 1208619264 M * infowolfe Bertl, 9.7m pageviews yesterday 1208619271 M * pmjdebruijn infowolfe: our Xeons are just 80Watts 1208619272 M * infowolfe Bertl, xp drivers, vista drivers or all drivers? 1208619278 M * pmjdebruijn E5405 1208619311 M * Bertl infowolfe: all of them, as I heard ... 1208619311 M * pmjdebruijn the entire server does less than 220 Watts. so that less than 1Amp @ 220/230 1208619320 M * infowolfe pmjdebruijn, how much ram? 1208619325 M * infowolfe and ddr2 or fb-dimm? 1208619335 M * pmjdebruijn infowolfe: typifaclly 4GB 1208619338 M * pmjdebruijn typically* 1208619341 M * pmjdebruijn DDR2 1208619361 M * infowolfe 16*2gb fb-dimm is what i blame for most of the 5A draw 1208619365 M * pmjdebruijn those are diskless systems though 1208619369 M * infowolfe lol 1208619396 M * infowolfe mine's not... 16x 74g seagate 15k.5 SAS 1208619404 M * pmjdebruijn uhr 1208619407 M * pmjdebruijn well there your power issue 1208619422 M * daniel_hozac indeed.... 1208619443 M * pmjdebruijn anyway, these days Intel is the way to go for servers 1208619452 M * pmjdebruijn early Opterons just beat the Pentium 4's to death 1208619454 M * Bertl infowolfe: btw, a financial contribution will definitely speed up a 2.6.25 release (hint) 1208619489 M * infowolfe pmjdebruijn, actually, intel's not the way to go if you have to pay for power (i do) 1208619547 M * pmjdebruijn infowolfe: we do as well 1208619563 M * pmjdebruijn we have our own datacenter 1208619582 M * Bertl well, Intel has improved a lot on the power side recently 1208619598 M * infowolfe i can fit 3 of my dual-socket 2352's in the same power footprint as my master db 1208619611 M * Bertl not as good as AMD, that's correct, but OTOH, the stability issues with AMD are a good point against it 1208619627 M * Bertl not the CPUs but the chipsets (usually) 1208619662 M * daniel_hozac infowolfe: and they have the same amount and type of RAM and disks? 1208619696 M * infowolfe ... 1208619709 M * infowolfe daniel_hozac, you realize that there isn't an amd board on earth built for fb-dimms? 1208619712 M * pmjdebruijn I wonder how much power a Quad 2ghz (12mb cache) opteron draws? 1208619734 M * daniel_hozac infowolfe: so... your comparison is moot. 1208619739 M * infowolfe hardly 1208619749 M * infowolfe the only thing i care about is how many amps it draws 1208619762 M * pmjdebruijn uhr Opteraon 2350 2000mhz, 2048kb does 75watts 1208619775 M * infowolfe pmjdebruijn, that's thermal envelope 1208619797 M * Bertl the 3GHz Quad Core sets at 65W 1208619802 M * infowolfe but ddr2 ram's going to draw less than fb-dimms because it doesn't have to work around crappy architectural decisions 1208619814 M * Bertl (intel that is) 1208619852 M * pmjdebruijn doesn't Broadcom have AMD chipset? 1208619866 M * infowolfe pmjdebruijn, they're crappier than the nvidia chipsets for performance 1208619878 M * pmjdebruijn hmm ok 1208619887 M * infowolfe and if you want, i can measure the amps drawn on my desktop a bit later 1208619894 M * Bertl via should be an option for AMD 1208619900 M * infowolfe i have some b2 stepping 2352's in my workstation 1208619901 M * pmjdebruijn Bertl: for a server? 1208619906 M * infowolfe Bertl, via isn't an option for anything 1208619912 M * pmjdebruijn infowolfe: right on! 1208619927 M * Bertl pmjdebruijn, infowolfe: actually the chipsets are better than the nVidia ones 1208619940 M * pmjdebruijn for a while I had an EPIA in a 1U casing... I replaced with with an old HP Proliant 1208619952 M * pmjdebruijn Bertl: well, that doesn't take much, does it :p 1208619952 M * infowolfe the biggest complaint i have about the nvidia chipsets are the onboard nics, which are worked around via e1000's 1208619969 M * pmjdebruijn are those forcedeth? 1208619977 Q * FireEgl Quit: Leaving... 1208620011 M * pmjdebruijn infowolfe: just curious, but why do you need 4 NICs? 1208620025 M * infowolfe oob/ext/bonded-int 1208620034 M * pmjdebruijn ah 1208620054 M * Bertl the intel dual/quad GB nics are nice for such things 1208620076 M * Bertl and PCIe handles that nicely with 2 lanes 1208620091 M * pmjdebruijn true 1208620095 M * infowolfe Bertl, i know ;) 1208620106 M * infowolfe 03:00.0 Ethernet controller: Intel Corporation 82571EB Gigabit Ethernet Controller (rev 06) 1208620106 M * infowolfe 03:00.1 Ethernet controller: Intel Corporation 82571EB Gigabit Ethernet Controller (rev 06) 1208620106 M * infowolfe 04:00.0 Ethernet controller: Intel Corporation 82571EB Gigabit Ethernet Controller (rev 06) 1208620106 M * infowolfe 04:00.1 Ethernet controller: Intel Corporation 82571EB Gigabit Ethernet Controller (rev 06) 1208620115 M * infowolfe EXPI9404PT 1208620121 M * pmjdebruijn to be honest that's the only crappy thing about those HP DL360/380 G5's... they have Broadcom NICs (bnx2). which are not bad... 1208620131 M * pmjdebruijn But I'd trade them for e1000's in a heartbeat 1208620203 M * pmjdebruijn infowolfe: does Tyan actually "support" Linux for that motherboard? 1208620227 M * infowolfe pmjdebruijn, they ship linux drivers for it, so i'd say yes 1208620232 M * pmjdebruijn ah 1208620282 M * infowolfe the reason i picked it is for its size... dual socket in ATX formfactor 1208620287 M * infowolfe http://tyan.com/product_board_detail.aspx?pid=595 1208620336 M * Bertl nice page :) 1208620422 M * Bertl now imagine using two quad cores there :) 1208620456 M * pmjdebruijn Bertl: our proxy is :) 1208620466 M * pmjdebruijn 16Ghz of power! 1208620482 M * pmjdebruijn image doing some Chess against that :) 1208621217 J * dowdle ~dowdle@71-32-20-58.blng.qwest.net 1208621237 M * Bertl wb dowdle! 1208621469 J * dna ~dna@19.Red-81-35-110.dynamicIP.rima-tde.net 1208621597 M * Bertl wb dna! 1208621605 M * mc wb bertl! 1208621624 M * Bertl mc: thanks :) 1208621634 M * mc :P 1208621772 Q * dna Quit: Verlassend 1208621841 M * Bertl infowolfe: so please drop me a note when you decided to speed up Linux-VServer development ... 1208621854 M * infowolfe Bertl, as soon as my boss wakes up, i'll poke him ;-) 1208621902 M * Bertl excellent! 1208622027 M * daniel_hozac Bertl: would you mind walking through the scheduler with me a bit? 1208622046 M * daniel_hozac (i.e. the Linux-VServer modifications) 1208622073 M * Bertl sure, no problem 1208622091 M * Bertl are you going to adapt the TBS to the RB tree stuff? 1208622123 M * Bertl (or just curious in regard to the PL folks) 1208622136 M * daniel_hozac i'd like to give it a shot, but it's also in regard to PL. 1208622210 M * daniel_hozac Sapan idenitified an infinite loop in the idle time skipping, which i think is due to Andy's patch... but i don't fully understand how it affects matters. 1208622211 M * Bertl i.c. no problem ... 1208622404 M * daniel_hozac there are a couple of variables i don't think i fully understand. vx_tokens_recalc's delta_min, what's that used for? 1208622623 J * yarihm ~yarihm@84-75-103-252.dclient.hispeed.ch 1208622623 Q * yarihm 1208622626 M * Bertl we do integral updates to the TB, i.e. we add a certain amount (rate) every interval, right? 1208622641 M * daniel_hozac right. 1208622675 M * Bertl now for a TB below the require minimum, we want to know the time (ticks) required to fill up right above that amount 1208622681 M * Bertl *required 1208622753 M * daniel_hozac okay... 1208622774 M * Bertl the formula for that is: delta = (Tn - Tn-1) 1208622917 M * Bertl tokens = int(delta/I)*R; 1208622994 M * Bertl integral = int(delta/I)*I 1208623013 M * Bertl which leaves a 'delta' to the next interval 1208623014 Q * Guy- Server closed connection 1208623019 J * Guy- ~korn@elan.rulez.org 1208623034 M * Bertl this is delta - integral 1208623043 M * daniel_hozac right 1208623057 M * Bertl (i.e after that time (ticks) the next R will be added) 1208623090 M * Bertl I - (delta - integral) that is 1208623122 M * daniel_hozac okay. 1208623127 M * Bertl the (delta - integral) is stored in delta_min[0] 1208623161 M * Bertl similar is done for the idle time 1208623200 M * Bertl at the end we use that for the HARD scheduler to schedule the next action 1208623214 M * daniel_hozac so delta_min is whatever's leftover after this context? 1208623260 M * Bertl no, delta_min is the location (timewise) in the current interval 1208623295 J * docelic_ ~docelic@78.134.199.186 1208623339 M * Bertl look at it like this, we have an amount of time which elapsed since last check 1208623365 M * Bertl we can break that down by divinding by the Interval(s) 1208623395 M * Bertl we then get an integral part (tokens/R) 1208623431 M * Bertl and a remainder, which is delta_min 1208623448 Q * docelic Ping timeout: 480 seconds 1208623483 M * daniel_hozac right 1208623498 M * Bertl note that we do not add the delta_min(s) to the time stored in the context 1208623512 M * Bertl (only the integral part is added there) 1208623601 M * Bertl we also calculate the ticks required to get a change in the number of tokens for this context (at the end) and store them in delta_min 1208623688 M * daniel_hozac at the end, what do the ifs mean? 1208623711 M * Bertl below the 'next interval' comment? 1208623716 M * daniel_hozac right. 1208623728 M * Bertl well, we have three cases here: 1208623746 M * Bertl 1) no rate is set (R=0) (that was added recently 1208623819 M * Bertl 2) we need less than R (rate) tokens to get the context going 1208623828 M * Bertl (this is the else case) 1208623857 M * Bertl 3) we need a bunch of I (intervals) to get enough tokens 1208623863 M * daniel_hozac oh, tokens is the number of tokens required to run again. i see. 1208623868 M * Bertl (this is the elseif case) 1208623886 M * Bertl yep, the code reuses the variables very heavily 1208623899 M * Bertl which might need one or the other comment :) 1208623956 M * daniel_hozac another thing i was wondering about... shouldn't sched_pc->norm/idle_time be set to *norm_time/*idle_time? won't they get out of sync, such that the check for adding new tokens always suceeds if a context is idle for a while? 1208623977 M * daniel_hozac i mean, obviously they don't, but i don't see don't understand why :) 1208623987 M * daniel_hozac -don't see 1208624009 J * snooze ~o@1-1-4-40a.gkp.gbg.bostream.se 1208624019 M * Bertl nah, setting them to the time would lose the deltas 1208624047 M * Bertl i.e. we add the integral part, and thus we save the fractions (deltas) for next time 1208624060 M * Bertl /* advance idle time */ 1208624060 M * Bertl sched_pc->idle_time += integral; 1208624073 M * Bertl same for the normal time, of course 1208624120 M * daniel_hozac right, but assuming my interval is something that doesn't evenly divide the HZ, won't that eventually ensure that there's always enough of a delta that i get new tokens? 1208624168 M * Bertl no, you are getting this wrong here, let me try to describe it from a different side 1208624189 M * daniel_hozac sure. 1208624220 M * Bertl let's see the sched_pc->norm/idle_time as clocks which can only assume discrete values like 1R, 2R, 3R ... 1208624270 M * Bertl when we get into *recalc() we get a time which might be 1007.3R 1208624288 M * Bertl the sched_pc->norm has 1005R 1208624306 M * Bertl that now makes 2R integral, and 0.3R delta 1208624329 M * Bertl the clock in sched_pc only gets the 2R, which makes 1007R 1208624346 M * Bertl the delta is not added there (because the clock cannot assume the fractional part) 1208624415 M * Bertl btw, R is a bad choice here, make that I :) 1208624433 M * daniel_hozac but what happens if the context is then idle for a while, so *recalc gets 1200.5I, but sched_pc->norm is still 1007I? 1208624449 M * Bertl same thing 1208624462 M * Bertl 1200-1007 = integral part 1208624472 M * Bertl 0.5 remains as delta 1208624475 M * daniel_hozac aahhhh... i see, we add the entire integral. 1208624488 M * daniel_hozac regardless of potential spill due to bucket size 1208624502 M * Bertl now, when we want to know, how many ticks we will need to get a context going, we do 1208624521 M * Bertl tokens_needed = min - current 1208624542 M * Bertl intervals = tokens_needed/rate 1208624599 M * Bertl delta = time to next interval 1208624624 M * Bertl ticks = intervals * I + delta 1208624692 J * hparker ~hparker@linux.homershut.net 1208624696 M * daniel_hozac right 1208624728 M * Bertl and that's what we return in the min_delta(s) 1208624736 Q * sannes Server closed connection 1208624737 J * sannes ace@har.sagt.no 1208624743 M * Bertl so that the cpu can sleep for that time 1208624880 J * dna ~dna@19.Red-81-35-110.dynamicIP.rima-tde.net 1208624958 M * daniel_hozac thanks, i think i'm starting to get the hang of it. 1208624995 M * Bertl you're welcome! feel free to ask more if something is unclear 1208625002 M * daniel_hozac will do. 1208625243 J * Infinito ~argos@201-3-16-77.gnace701.dsl.brasiltelecom.net.br 1208626300 Q * Infinito Read error: Connection reset by peer 1208626325 Q * fatgoose Quit: fatgoose 1208626634 Q * hparker Quit: Read error: 104 (Peer reset by connection) 1208626641 J * hparker ~hparker@linux.homershut.net 1208627362 Q * dna Quit: Verlassend 1208627559 J * bfremon ~ben@ANantes-252-1-87-136.w81-48.abo.wanadoo.fr 1208627829 J * esa bip@ip-87-238-2-45.static.adsl.cheapnet.it 1208629104 Q * hparker Read error: Connection reset by peer 1208629186 M * Bertl nap attack .. bbl 1208629206 N * Bertl Bertl_zZ 1208630224 J * hparker ~hparker@linux.homershut.net 1208630321 J * virtuoso ~s0t0na@ppp91-122-186-241.pppoe.avangarddsl.ru 1208630440 J * JonB ~NoSuchUse@77.75.164.169 1208630728 Q * virtuoso_ Ping timeout: 480 seconds 1208631961 Q * bfremon Ping timeout: 480 seconds 1208632708 J * bfremon ~ben@ANantes-252-1-87-136.w81-48.abo.wanadoo.fr 1208633905 J * fatgoose ~samuel@76-10-149-199.dsl.teksavvy.com 1208634199 Q * [PUPPETS]Gonzo Server closed connection 1208634214 J * [PUPPETS]Gonzo gonzo@fellatio.deswahnsinns.de 1208634705 J * onox ~onox@kalfjeslab.demon.nl 1208634755 Q * docelic_ Read error: Connection reset by peer 1208637887 Q * bfremon Remote host closed the connection 1208639308 J * ntrs ~ntrs@77.29.66.99 1208639366 J * esa` bip@ip-87-238-2-45.static.adsl.cheapnet.it 1208639389 Q * esa Ping timeout: 480 seconds 1208639855 J * ntrs_ ~ntrs@77.29.66.146 1208640249 Q * ntrs Ping timeout: 480 seconds 1208640416 Q * MatBoy Remote host closed the connection 1208640486 Q * larsivi Ping timeout: 480 seconds 1208640613 M * ntrs_ Does anyone know why bind mount partitions are not visible with 'df' in a CentOS 5 guest? They are fine in any other guest, ubuntu, even CentOS 4. 1208640640 M * daniel_hozac how are you mounting them? 1208640656 M * daniel_hozac what initstyle? 1208640661 M * ntrs_ they are specified in the guest's config fstab file 1208640684 M * ntrs_ I am not sure about the initstyle 1208640702 M * daniel_hozac cat /etc/vservers//apps/init/style 1208640764 M * ntrs_ there is no such file 1208640774 M * daniel_hozac so you're using the default, sysv. 1208640779 M * ntrs_ ok 1208640796 M * daniel_hozac in that case, nothing should be scrubbing /etc/mtab in the guest apart from the vserver utils. 1208640825 M * ntrs_ on the same host other guests are showing the df information correctly except in centos 5. 1208640845 M * ntrs_ what do you mean? You think something is deleting /etc/mtab in the guest? 1208640858 M * daniel_hozac it's never deleted, just truncated. 1208640912 M * ntrs_ emptied? 1208640961 M * daniel_hozac i can't reproduce your issue though. my CentOS 5 guests show bind mounts. 1208640972 M * ntrs_ hah, look at this: 1208640975 M * daniel_hozac what filesystem have you put in your fstab? 1208640976 M * ntrs_ # Clear mtab 1208640977 M * ntrs_ (> /etc/mtab) &> /dev/null 1208640985 M * ntrs_ this is from the rc.sysinit 1208640990 M * daniel_hozac yes. 1208640999 M * ntrs_ I think that is the problem here. 1208641012 M * daniel_hozac you don't run that script if you're using the sysv initstyle. 1208641041 M * daniel_hozac does /etc/mtab in the guest contain any information pertaining to that filesystem? 1208641055 M * ntrs_ hmm, let me check 1208641100 M * ntrs_ /dev/hdv1 / ufs defaults 0 0 1208641101 M * ntrs_ none /proc proc 0 0 1208641101 M * ntrs_ none /dev/pts devpts gid=5,mode=620 0 0 1208641108 M * ntrs_ '/vservers/guest/tmp /tmp none 0 0 1208641111 M * ntrs_ '/vservers/guest/var/tmp /var/tmp none 0 0 1208641125 M * ntrs_ please ignore the ticks at the front of the last two lines 1208641143 M * daniel_hozac what utils are you using? 1208641180 M * ntrs_ .212 1208641207 M * daniel_hozac you probably want to upgrade those. 1208641211 M * ntrs_ ok 1208641218 M * daniel_hozac but you also want to change the filesystem type in your guest's fstab. 1208641227 M * ntrs_ but they are working fine with all the other guests on the host. 1208641235 M * daniel_hozac set it to ext3 or something. 1208641292 M * ntrs_ none /proc proc defaults 0 0 1208641293 M * ntrs_ none /dev/pts devpts gid=5,mode=620 0 0 1208641295 M * ntrs_ '/vservers/guest/tmp /tmp none bind,noexec 0 0 1208641297 M * ntrs_ '/vservers/guest/var/tmp /var/tmp none bind,noexec 0 0 1208641307 M * ntrs_ this is what I have currently in fstab 1208641332 M * ntrs_ so, I should replace none with ext3? 1208641340 M * daniel_hozac yes. 1208641352 M * ntrs_ ok, let's try that 1208641404 M * ntrs_ ok, it's working now. 1208641417 J * Aiken ~james@ppp121-45-192-61.lns1.bne1.internode.on.net 1208641424 M * ntrs_ Now, what I don't understand is why it does work with all the other guests. 1208641434 M * daniel_hozac different versions of df. 1208641460 M * ntrs_ Ok, I guess that's it then. 1208641464 M * ntrs_ Thanks daniel_hozac 1208641526 M * daniel_hozac you're welcome 1208642469 Q * JonB Quit: This computer has gone to sleep 1208646462 Q * hparker Quit: Read error: 104 (Peer reset by connection) 1208647041 J * larsivi ~larsivi@144.84-48-50.nextgentel.com 1208649291 J * Infinito ~argos@200-101-44-160.gnace701.dsl.brasiltelecom.net.br 1208649377 Q * Infinito 1208649410 J * Infinito ~argos@200-101-44-160.gnace701.dsl.brasiltelecom.net.br