1105574412 M * chrish01 so say i have ethernet coming over ATM, T1(aka E1 in europe), etc... each circuit is running ethernet (ethernet II) 1105574414 M * nayco so this circuit are 'things' that connect 'machines' 1105574448 M * chrish01 what i do is i map these circuits to vlans...then plug a vlan trunk into a single vserver host...then i can map those circuits into seperate vserver contexts to do routing 1105574472 M * nayco ....i'm computing.... 1105574481 M * chrish01 i can make a quick drawing 1105574493 M * nayco yep 1105574603 M * nayco Ok, I think I begin to understand.... 1105574618 M * prae it exist a backport of vserver 1.9.x on linux tree 2.4.x ? 1105574626 M * chrish01 prae, noe 1105574628 M * chrish01 err no 1105574680 M * nayco One or more vlans physically enters a vs host, each of these vservers processes a vlan to route communications between... 1105574682 M * nayco er... 1105574687 M * nayco between.... 1105574858 M * prae bummer :( 1105574869 M * prae thanks chrish01, nayco 1105574872 M * prae see'ya 1105574872 M * prae :) 1105574880 M * nayco bye ! 1105574887 M * chrish01 nayco 1105574888 M * chrish01 http://people.mosaix.net/chris/images/Untitled.png 1105574894 Q * prae Quit: . 1105574896 M * nayco ..... 1105574946 M * nayco the aggregation host... Is it what We call an IMUX ? 1105574973 M * nayco I'm a shame in networks, apart from ipv4 and vlans ;) 1105574976 M * chrish01 naw, its custom linux solution i wrote 1105574982 M * chrish01 using Linux ATM and iproute2 1105574992 M * nayco garglll.. Ok ;) 1105575022 M * chrish01 its much cheaper as well compared to commercial products 1105575059 M * chrish01 and better performing...the best thing about doing this stuff ourselves on linux, is we dont get stuck with our hardware manufacturers timeline for servinces like ipsec and such to be implemented 1105575072 M * nayco Ok, I think I begin to see the usefullness of vserver / virtual routeer... 1105575093 M * chrish01 totally 1105575108 M * nayco and stuck into proprietary extension/features ;) ? 1105575115 M * chrish01 i can only get about 10-20 umls doing this now...but with vserver and ngnet...that could be like 100! 1105575138 M * chrish01 yes! we only use opensource software here...any changes we make we fund publicly or push back to community 1105575156 M * chrish01 we have on proprietary product, and thats our NMS system for our customers to use to manage their network 1105575161 M * chrish01 s/on/one 1105575236 M * nayco Ok, i'll try to sum up: Your job is to route _many_ communication on different network/protocols/pipes between, say, the three "clouds" of you schema 1105575281 M * nayco And that would require _many_ routers. And maybe customers' communications to travel in the same pipes ? 1105575297 M * nayco So, you found a way to do this on a sigle host. 1105575302 M * chrish01 yes, normally we would have one central router for each customers set of circuits 1105575311 M * nayco ...With one ethernet interface ? 1105575314 M * chrish01 yes...we actually have more than one host...but you get the idea 1105575332 M * chrish01 yea, but have 2 gigabit per box so we can have optimum traffic 1105575346 M * chrish01 then we can Bond the ethernet lines 1105575372 M * nayco yep... Although i'm not enough strong on routing (ie. near zero) to get the full potential of your system :( 1105575383 M * nayco yes! 1105575400 M * chrish01 yea... (www.mosaix.net) if you want to pass the idea around to your friends :) 1105575475 M * nayco Ok, so now you may be able to explain how the vserver (Util|kernel) manages interfaces/aliases so the vserver gets an IP... I mean, 1105575536 M * chrish01 ok...so i get all those circuits into the vservers root context (0)...whats tricky is getting them into seperate contexts with seperate routing tables and firewalls 1105575546 M * nayco what bothers me is how come my vserver can get an ip when its network starting script does nothing or is even not started ? 1105575560 M * chrish01 this is where NGNet comes into play...it supports segmented routing tables and iptables 1105575583 M * chrish01 nayco, that is setup by vserver command from context 0, not the internal server 1105575600 M * nayco ... 1105575614 M * nayco well... 1105575617 M * chrish01 ngnet allows me to say..ok i have vlan 100 on eth0 (eth0.100)...i want to map that to context 1 as eth0 1105575636 M * chrish01 then i can pretend that context 1 is a *real* box and setup routing tables/firewall just as i would a normal box 1105575681 M * nayco ok for ngnet... I understand the need ;) 1105575696 M * chrish01 yup yup 1105575698 M * nayco It needs a lot of kernel hacking, so ? 1105575716 M * chrish01 Bertl has done quite a bit of it already...it needs some special ARP work and such though 1105575725 M * chrish01 and small bugfixes here and there 1105575730 M * nayco these guys are gurus :o ! 1105575741 M * chrish01 and ability to use it inside a vserver, not just contexts 1105575750 M * chrish01 nayco, they sure are 1105575806 M * nayco Which bring me to that: more globally, I find it strange that when "booting" my vserver, not all the scripts are run, or functionnal because of caps restriction 1105575847 M * chrish01 i havent actually used any of the stable stuff :( 1105575869 M * chrish01 i set one up once quick to see what it can do, but removed it right after cause it lacked routing table segmentation 1105575877 M * nayco but i works fine, and even gets an ip without doing it itself... I mean, my /vserver/.../etc/hostname is empty, .../sysconfig/network* too :o ! 1105575908 M * chrish01 thats because when you start with `vserver name start`...the vserver actually changes those via the kernel rather than calling startup scripts inside the vserver 1105575997 M * nayco Well, I tried first the unstable, but i lack features that force me to do by hand some thing I cannot/don't want to do. So I tried stable, but without vprocunhide i'm stuck too... So i'm testing different versions to find the one that fits best my needs.... 1105576038 M * chrish01 nayco, what exactly do you need...maybe some of my experience researching this stuff can be of help 1105576040 M * nayco I've got a test Apache vserver running fine, but made with a method I don't like ("copy" => 1 Gig :( ) 1105576122 A * chrish01 wonders if there is a CopyOnWrite (COW) system for vserver 1105576152 M * nayco Well, I need something simple :p ! Tools/methods allowing me to have a "main" vserver (The master template, nearly emty, I managed to get one by installing only "basesytem" with "urpmi --root /vserver/template" 1105576165 M * daniel_hozac i don't think COW is done yet. 1105576179 M * chrish01 daniel_hozac, k thanks 1105576190 M * nayco then backup this one, and clone it to "secondary" templates 1105576233 M * nayco (A samba one, Apache , Bind, MySQL...), then backup them too. 1105576282 M * nayco And when in need, clone a template, add additional stuff, create/mount/fstab partition for them (Thanks LVM and XFS) 1105576291 M * chrish01 so are you using this for link an internal server to do all your services on? 1105576306 M * nayco huh ? 1105576323 M * chrish01 s/link/like 1105576344 M * chrish01 so one vserver...with samba vserver, bind vserver...etc? 1105576347 M * nayco Not exactly Well, in first place, to build our new IT departement 1105576358 M * nayco server 1105576416 M * nayco used primarily for storing work documentation, backup purpose, management webapps 1105576489 M * nayco , wiki, etc... Each on a vserver, then, to fire on-demand vservers to test a new program/service, compile update for a particular distro version 1105576527 M * chrish01 ah cool...make sure your backup is on a seperate server though dude 1105576534 M * nayco without breaking a production machine, be it my own or the one where critical info is stored ;) 1105576537 M * chrish01 and preferably off site, or night tape backups 1105576553 M * chrish01 backup is essential (i speak from experience) 1105576651 M * nayco well... This is a point I prefer not to be verbose on: We haven't got ofsite backup, nor tape: Hard disk, in one of my servers room, and my backup software is rsync. Try to make them spit dosh when it come to IT, and you'll and up depressive. 1105576653 M * nayco :-- 1105576675 M * nayco ohh, many typos ;-) 1105576742 M * nayco Well, that's not the point: We are not a bank ;P ! By the way, I only begin to see the full potential of vservers, and I thinks it is just a matter of imagination 1105576766 M * nayco ....And needs to give answers to. 1105576810 M * chrish01 just know that if your vserver crashes, you loose all that data, people will be mad and they *will* blame you 1105576833 M * chrish01 by crash i mean hardware failure, fire, etc.. 1105576842 M * nayco Yep: But the backup isn't done on the vserver host itself ! 1105576881 M * chrish01 ah k 1105576958 M * nayco No, it is a different machine. When I talked of backup on the vserver, let's be more precise: We a 4 IT employes with personnal machines on our desks, and this is just in case we need to store things in 2 different places in case of crash of our own machines... ;-) 1105576986 M * chrish01 oh ok 1105576991 M * nayco Anyway, the vservers will be backed up (Well, on the poor backup server with IDe disks) 1105576992 M * chrish01 but what if your office burns down? 1105576998 M * nayco I get kille 1105576999 M * nayco d 1105577003 M * chrish01 hehe 1105577018 M * nayco Yes, I know, this comes to my mind often... 1105577039 M * chrish01 ;-) 1105577060 M * nayco Hope i'll have this discussed before I lose sleep... Or the office burns :-| 1105577070 M * chrish01 haha 1105577095 M * chrish01 the burdon of being a linux geek is that you see what can go wrong more often...which means you got to fix it as well :) 1105577140 M * nayco well, back in the topic... So, how do util-vservers boot my vservers so they get an IP not by themselves, but from "outside" (From the host) 1105577163 M * nayco This is still the magic in the vserver that I cannot understand. 1105577185 M * chrish01 that is done in your config...are you using develop util-vserver? 1105577188 M * chrish01 and what distro btw? 1105577209 M * nayco I use both stable and devel one two different test machines 1105577261 M * chrish01 im only familiar with vserver on debian...any chance you are running debian? 1105577311 M * chrish01 configuring the network stuff is different on devel and stable...in stable...see /etc/.conf and on devel there are config files in /etc//... 1105577350 Q * serving Ping timeout: 480 seconds 1105577405 M * nayco nope, only drake. I'd like to test debian, but i'm happy with drake, and it is not the time for me to test something at the same time that vservers ;-) 1105577506 M * chrish01 ah...i stopped using drake back around 7.1 days 1105577538 M * nayco Well, I saw that: But what impresses me it the distro in the vserver, although having its own network init scripts, has no conf file set inside its "chroot" (Because installed from rpms and not configured), but outputs nearly no errors upon booting, and still gets an IP _from outside_... 1105577617 M * chrish01 you also need to configure your capabilities for that vserver if you want to do stuff like bind a service or something from inside 1105577755 M * nayco Well, I started with slack (At school), I found the packages management hard, and I quickly got to use Mdk because it is the first I found in a magazine... I simply haven't felt the need to change, (I tried Gentoo, but on an oldworld Mac, which is real suicide :P). Maybe Debian would be more suitable for production servers, but, to late, I'm "efficient" (I think i'll never be: This changes too fast, that's why I like this job) 1105577755 M * nayco with Drake ... 1105577835 M * nayco "...stuff like bind a service..." => Yep, I read the Faq, I'll try that... Lucky it is explained ;-) 1105577881 M * chrish01 debian is absolutely amazing with production servers because of apt-get 1105577969 M * nayco This one is hairy, but hell, let's feed it: Mdk has urpmi :p ! 1105577976 M * nayco Ok, it cannot complie 1105577990 M * nayco ^W build packages... 1105578013 M * nayco but is useful and efficient. 1105578480 J * anonc ~nwalsh@shaggy.internode.com.au 1105578575 Q * anonymous-coward Ping timeout: 480 seconds 1105578634 M * nayco chrish01: Which software did you use to draw the diagram you sent me ? 1105578898 M * chrish01 OmniGraffle for OS X 1105578958 M * nayco I look good. Is it free software ? 1105578967 M * nayco s/look/looks/ 1105578982 M * chrish01 no, it is quite expensive actually 1105578986 M * chrish01 www.omnigroup.com 1105579017 M * nayco ok. 1105579027 M * nayco Still better than Dia :-p ! 1105579061 M * nayco Well, time for me to sleep, thanks a lot, and read you soon.... bye, ! 1105579093 Q * nayco Quit: Bonne nuit ! 1105579687 Q * chrish01 Quit: Leaving 1105584063 J * monrad ~monrad@213083190130.sonofon.dk 1105584246 J * serving ~serving@213.186.172.180 1105586307 Q * pusling Read error: Connection reset by peer 1105586309 J * pusling_ ~pusling@195.215.29.124 1105587694 Q * pusling_ Read error: Connection reset by peer 1105587954 J * chrish01 ~chrish01@c-67-168-71-150.client.comcast.net 1105591488 J * nox- ~nox@c135103.adsl.hansenet.de 1105591835 Q * nox Ping timeout: 480 seconds 1105591848 N * nox- nox 1105592002 Q * TheSeer Read error: Operation timed out 1105592041 J * TheSeer ~theseer@212.12.45.62 1105592610 Q * tchan Ping timeout: 480 seconds 1105595513 Q * monrad Quit: Leaving 1105596331 Q * chrish01 Quit: Leaving 1105596840 J * monrad ~monrad@213083190130.sonofon.dk 1105597595 J * tchan ~tchan@c-24-13-81-164.client.comcast.net 1105600045 Q * tchan Ping timeout: 480 seconds 1105600555 J * chrish01 ~chrish01@69.90.131.10 1105600610 J * tchan ~tchan@c-24-13-81-164.client.comcast.net 1105604944 J * BWare ~bware@212.26.196.41 1105608573 Q * mcp Ping timeout: 480 seconds 1105609094 J * prae ~prae@sherpadown.net 1105609099 M * prae morning' 1105609198 M * chrish01 morning prae 1105610108 N * Bertl_oO Bertl 1105610145 M * Bertl morning folks! (finnaly back) 1105610330 M * Bertl chrish01: not yet, but we will address it soon ... (ngnet) 1105610331 J * Loki|muh loki@satanix.de 1105610342 M * Bertl morning prae! 1105610348 M * Bertl morning Loki|muh! 1105610362 M * chrish01 Bertl, thanks dude 1105610390 M * Bertl the priority atm is getting a stable release for 2.6.11 ... 1105610412 M * Loki|muh hi Bertl 1105610415 M * Loki|muh finally back 1105610460 M * Bertl hmm, me or you? 1105610580 M * Loki|muh me, lots of disconnects... another guy thought it would be funny taking my ip :/ 1105610587 M * chrish01 Bertl, ok sounds good 1105610708 M * Bertl Loki|muh: heh, just funny (we used the same words, see irc log ;) 1105610750 M * Loki|muh url? ;) 1105610760 M * Bertl (well, except for my spelling ;/) http://irc.13thfloor.at/LOG/LOG_2005-01-13.txt 1105611550 M * TheSeer Bertl: strange question ;) can a vserver crash? *g* 1105611563 M * TheSeer e.g. a vserver on a customer box stopped repsonding 1105611583 M * TheSeer no network reachability of any kind 1105611598 M * TheSeer the other vserver work fine and flawless 1105611610 M * Bertl strange ... no basically it just might happen that it shuts down ... 1105611630 M * TheSeer according to vserver-stat it's still up and running 1105611650 M * TheSeer hmm.-. i'll restart one of the two.. 1105611654 M * Bertl nope 1105611667 M * Bertl try to use tcpdump on the host, and look for traffic 1105611686 M * Bertl (best try to connect from outside to that vserver's services) 1105611713 M * Bertl if you see incoming packets, but no response ... 1105611751 M * Doener hmm... "secondary issue"? 1105611752 M * TheSeer k.. 1105611774 M * Bertl Doener: could be ... morning btw! 1105611779 M * Doener morning 1105611823 M * TheSeer can i bind tcpdump to a secondary interface? 1105611829 M * TheSeer like eth1:foo? 1105611844 M * TheSeer otherwise the traffic caught by tcpdump is killing me ;) 1105611847 M * Bertl nope, just tcpdump eth0 with 'host ' 1105611853 M * TheSeer oh;) 1105611857 M * TheSeer stupid me ;) 1105611864 M * TheSeer right.. k.. wait a sec ;) 1105611872 M * Bertl (e.g. tcpdump -vvnei eth0 host ) 1105611977 M * TheSeer i forgot about the host-param :/ 1105611987 M * TheSeer k.. i can see pings coming in.. 1105611997 M * TheSeer from my workstation here.. 1105612002 M * TheSeer but i don't get a reply 1105612012 M * TheSeer 11:26:44.697079 border.office.salesemotion.net > cust126.netera.de: icmp: echo request (DF) 1105612018 M * Bertl do you see echo replies leaving the host? 1105612035 M * Bertl (try with tcpdump -vvnei eth0 icmp ) 1105612064 M * TheSeer 11:27:33.690055 0:5:dc:2e:88:0 0:2:55:9c:70:95 ip 98: 212.12.45.62 > 213.238.32.126: icmp: echo request (DF) (ttl 58, id 4, len 84) 1105612066 M * TheSeer that's all 1105612071 M * TheSeer and thats incoming only 1105612115 M * Bertl k, does 'ip addr ls' list the ip of that vserver? 1105612183 M * TheSeer erm... 1105612193 M * TheSeer doesn't look like it... 1105612196 M * TheSeer wtf..? 1105612219 M * Bertl I guess Doener can explain what happened to you ;) 1105612233 M * TheSeer ouhm.. but restarting the vserver will fix that, right? 1105612247 M * TheSeer it's a 2.4 box btw ;) 1105612311 M * Bertl Doener: do you have time to explain? 1105612387 M * Bertl seems not so ... okay, you are using recent tools, right? 1105612585 M * Doener sorry, been on the phone... 1105612599 M * Bertl np 1105612627 M * Doener TheSeer: thing is, for each subnet, there is one primary address and any number of secondary addresses 1105612664 M * Doener if you ever remove the primary address, all addresses in that subnet are removed as well 1105612682 M * Bertl (most likely your 'other' server used the primary) 1105612762 M * Doener f.e.: if you add 10.0.0.1/8, 10.0.0.2/8 and 10.0.0.3/8 (in that order), 10.0.0.1 would be the primary address. If you now remove 10.0.0.2, nothing magical happens, 10.0.0.2 is gone. but if you then remove 10.0.0.1, 10.0.0.3 is gone as well. 1105612773 M * BWare Bertl: I don't know to much about he linux ip stack, but I noticed that almost always responses are sent from the primary interface 1105612807 M * Bertl yep, that's right ... 1105612820 M * Doener BWare: because the routing entries are generated when the primary address is setup, and the address goes into the source address field of the routing entry 1105612826 M * Doener IIRC 1105612834 M * BWare Exactly 1105612876 M * BWare So, if the default route is set to another interface, packets are sent from that interface 1105612980 M * BWare and what I noticed in a recent vserver setup, using iproute2 tables doesn't help a bit :( 1105613037 M * Bertl yes, it helps, but you ahve to 'use' them 1105613061 M * Bertl i.e. you have to get the outgoing packets to actually 'use' a different routing table 1105613099 M * Bertl then you will get different routing entries, and if they are done properly different ipsfor different subnets ... 1105613102 M * BWare yep, so you need to tag them ;) 1105613116 M * Bertl (doesn't work with different ips on the same subnet though) 1105613124 M * TheSeer i didn'T do anything on that box 1105613132 M * TheSeer let along remove the primary address 1105613145 M * TheSeer since that would successfully disconnect me from it ;) 1105613176 M * TheSeer i just got a call from the customer about 5 minutes before i started posting that his vserver doesn'T answer anymore 1105613194 M * TheSeer i didn't touch anything vserver related on that maschine in weeks 1105613207 M * TheSeer (the box has an uptime of 200+ days) 1105613227 M * TheSeer despite the fact it might be advised to do a kernel update *g* 1105613230 M * Bertl hmm .. okay, could you enter that vserver and do 1105613237 M * Bertl grep Cap /proc/self/status 1105613259 M * TheSeer i already restarted the vserver since i couldn'T keep the client waiting any longer 1105613264 M * matti Yuuuupiiii... 1105613269 M * matti I love RIPE ;] 1105613280 M * TheSeer but the ip was definetly gone from the list of active secondaries 1105613295 M * Bertl doesn't matter, do it anyways please ... 1105613300 M * TheSeer k 1105613305 M * Bertl hey matti! 1105613310 M * matti Hello Bertl :>> 1105613323 M * TheSeer [root@vserver:notes /]grep Cap /proc/self/status 1105613323 M * TheSeer CapInh: 0000000000000000 1105613323 M * TheSeer CapPrm: 00000000d40c04ff 1105613323 M * TheSeer CapEff: 00000000d40c04ff 1105613323 M * TheSeer CapBset: 00000000d40c04ff 1105613367 M * TheSeer (kernel 2.4.25-vs1.26 - util-vserver-0.29.2-0) 1105613417 M * Bertl hmm, looks fine (i.e. no idea what caused that issue) 1105613455 M * TheSeer interesting.. :) 1105613470 M * TheSeer could running low on physical ram cause that? 1105613507 M * TheSeer because the box only had ~92MB left this morning ( from 2GB of ram total) 1105613517 M * Bertl we get/got reports of such 'incidents' on 2.4/1.2 now and then (regarding a vserver losing his ip) but we had no chance to get a clue yet ... 1105613543 M * Bertl the secondary ip issue definitely is the explanation in most cases 1105613551 M * Bertl (or at least a good one ;) 1105613553 M * TheSeer hehe 1105613553 M * we2by is a amd opteron 1600mhz compared to an intel 1600mhz? 1105613562 M * TheSeer cheaper? 1105613565 A * TheSeer hides 1105613588 M * Doener intel 1600mhz what? p4, celeron, xeon? 1105613618 M * TheSeer and amd usually is a bit faster on the same mhz amount then intel ( thus their crappy rating labels) 1105613639 M * TheSeer assuming you compare equal cpu types 1105613729 J * jsambrook ~jsambrook@host-62-69-64-93.bsve.net 1105613738 M * Bertl welcome jsambrook! 1105613771 M * we2by a p4? 1105613792 M * TheSeer opterons should be faster then a p4 1105613801 M * TheSeer on same mhz level 1105613812 M * we2by but the opteron is only 1600mhz 1105613812 M * TheSeer besides that amd cpu's are cheaper 1105613825 M * TheSeer is a amd opteron 1600mhz compared to an intel 1600mhz? 1105613847 M * TheSeer amd has a p-rating (?) 1105613854 M * TheSeer e.g. they label a cpu 1600+ 1105613865 J * rs rs@ice.aspic.com 1105613872 M * rs hi 1105613874 M * Bertl welcome rs! 1105613874 M * TheSeer which means it's compareable to an Intel-CPU of the same type with this Mhz 1105613891 M * nox is there any doc about opnvpn in an alpha vserver ? 1105613892 M * TheSeer the "real" clockrate is lower - like 1200 or something 1105613903 M * we2by mhh 1105613918 M * TheSeer the clockrate itself has not really anything to say 1105613924 M * we2by so a opetron 242 1600mhz is compared to a pentium x 1600mhz 1105613934 M * TheSeer no 1105613953 M * nox clockrate says as much as max speed on the tachometer *g 1105613954 M * TheSeer if it's 1600Mhz on both the AMD CPU is prolly faster 1105613967 M * we2by TheSeer, why? 1105613981 M * TheSeer because clockrate is a useless figure 1105613992 M * TheSeer you can only compare cpu's of the same type by clockrate 1105614007 M * TheSeer e.g. you can tell that an intel p4 1Ghz is slower than then intel p4 2Ghz 1105614021 M * TheSeer the clockrate itself ALONE is useless 1105614037 M * TheSeer e.g.a Risc CPU at 1 Ghz is *way* faster then any intel cpu ever will be ;) 1105614049 M * we2by mhh 1105614054 M * Bertl hmm? 1105614057 M * TheSeer let's compare it with a car 1105614077 M * TheSeer if your engine makes 6000rpm 1105614086 M * TheSeer it doesn't say anything about the speed you can go, or does it? 1105614096 M * we2by it does say a litle bit 1105614103 M * TheSeer bull shit 1105614109 M * chrish01 gears dude :) 1105614126 M * chrish01 and gears are all tuning...you can lengthen, shorten, etc 1105614127 M * TheSeer i can get my car to 6.000 rpms without even moving 1105614161 M * we2by a car without wheels? 1105614170 M * TheSeer no, just no gear set 1105614203 M * TheSeer the point is, it's all a question about gear.. 1105614217 M * we2by ok. I see 1105614220 M * TheSeer e.g. what's the rate of engine rpms <-> wheel rpms 1105614228 M * TheSeer same goes for a cpu 1105614229 M * we2by what would yo choose? a dual opteron or a p4 3.2ghz? 1105614236 M * we2by opteron model 242 1105614238 M * Zoiah we2by: opteron 1105614243 M * nox opteron 1105614246 M * TheSeer opteron 1105614257 M * Bertl TheSeer: now that you brought up that analogy, how to I change the gear on my CPU? 1105614266 M * we2by hehe 1105614271 M * Zoiah Bertl: the turbo button on your case. ;) 1105614277 M * TheSeer *lol* 1105614277 M * nox lol 1105614280 M * we2by :D 1105614293 M * matti ROTFL 1105614320 M * Bertl Zoiah: ah, and the sleep button must be 'neutral' then? *G* 1105614327 M * matti Bertl: :D 1105614329 M * Zoiah Bertl: you got it. ;) 1105614358 A * nox searching reverse *g 1105614425 M * Bertl ah, some computation you need to be undone? 1105614510 M * nox reverse engeneering *g 1105614523 M * Bertl okay, that one goes to you ;) 1105614553 M * nox hehe 1105614799 M * nox we2by ps opteron because of hyperlink and integrated memorycontroller and a nice 64bit idea 1105614821 M * we2by but not alot app use 64bit feature 1105614846 M * nox no therfore the slim amd64 is much better 1105614881 M * nox enough ram and you can address64bit without the big overhead of itanium 1105614883 M * Bertl we2by: you should also note that x86_64 is not as well supported as x86 (yet, which makes it a pita sometimes) 1105614910 M * Bertl (this will probably be a non-issue in a few month) 1105614947 M * nox we2by opteron runs 32 bit ... itanium emolates it afaik 1105614995 M * we2by mhh that x86_64 support changes my mind 1105615024 M * nox we2by in which way ? 1105615029 M * we2by ok, I better stay a few more months with my p4 1105615033 M * we2by app support 1105615041 M * we2by I will put windows on it 1105615049 M * nox we2by no that argument sux *g 1105615075 M * nox opteron is fastest x86 also 1105615091 M * we2by so? if my app does not run on it, make no sense 1105615105 M * nox opteron is fastest 32bit also 1105615120 M * Bertl we2by: well, win* will probably not use any x86_64 stuff, just the x86 parts ... which should make it quite compatible 1105615186 M * we2by beside such sytem with 2 opteron is too heavy for a home user like me 1105615204 M * we2by if it were a server, I would spend my money on it 1105615227 M * nox has anyone openvpn running on a alpha vserver ? 1105615259 M * chrish01 Bertl, message me if you get a change to do any ngnet work :) im dying to test out some stuff in a vserver 1105615261 M * Doener for home use you could go with an athlon64... cheaper, fast as well and AFAIK a lot cooler (as thus more quiet) than a p4... but if you already got a fast box, just keep it ;) 1105615264 M * chrish01 anway, night all 1105615280 M * Doener nig chrish01 1105615283 M * Bertl nox: I remember somebody talking about it, if that wasn't you, I'd grep the logs ... 1105615284 M * nox cu chrish01 1105615286 M * Bertl night chrish01! 1105615289 N * chrish01 chris|sleep 1105615290 A * Doener needs tab-completion for "night" 1105615327 M * Bertl better than tab completion for 'a' or 'I' ;) 1105615381 M * Doener heh 1105618973 M * rs re 1105619078 M * Bertl wb rs! 1105620430 M * Bertl Doener: do you remember, was the start/uptime issue we fixed an issue present with 2.6.9-vs1.9.3? or did that happen _after_ 2.6.9? 1105620571 M * Doener was changed in 2.6.9 1105620590 M * Bertl so, 1.9.3 should do the wrong thing there, right? 1105620786 M * Doener yes, 1.9.3 contains TheWrongFix (tm)... logs say you re-fixed it around 1.9.3.7 1105620797 M * Bertl okay, great, tahnks! 1105620808 M * Bertl btw, do you ahve an hour or so today? 1105620866 M * Doener guess so, maybe in an hour or so... i'll tell you then 1105620875 M * Bertl k, tx again! 1105621324 M * rs Bertl: I still have the sticky context problem sometimes, due to filehandles that remains open 1105621340 M * rs is there a way to force remove the xid ? 1105621353 M * Bertl nope, will be solved after 1.9.4 ... 1105621362 M * rs you found the bug ? 1105621384 M * Bertl there is no bug ... but we will work around the issue ... 1105621413 M * rs cool, could you explain how the work around will do? 1105621422 M * rs be 1105621433 M * Bertl yes, pls ask me later this evening ... 1105621441 M * rs ok 1105621462 M * rs I'm ready for some NGNET testing when you want 1105621477 M * rs by the way 1105621522 M * Bertl cool, get the patches and try to setup some test env ... make the vnet3_setup.sh work for you ... 1105621532 M * rs ok 1105621545 M * Bertl http://vserver.13thfloor.at/Experimental/NGNET/ 1105621558 M * Bertl http://vserver.13thfloor.at/Experimental/NGNET/diff-2.6.10-vs1.9.3.17-ng8.7.diff 1105621572 M * Bertl you need to patch iptables and compile the vnet tools too 1105621575 J * mcp ~hightower@81.17.110.148 1105621581 M * Bertl welcome mcp! 1105621621 M * rs yeah already do a patched iptables debian package IIRC 1105624848 Q * sannes Read error: Connection reset by peer 1105625569 M * pflanze we2by: I'd choose AMD any day. My latest experience comparing an Athlon 2088Mhz with an Intel Xeon 2392Mhz shows that the Athlon is 4 times faster. 1105625598 M * Bertl hmm, any benchmarks for that claim? 1105625614 M * pflanze I'd expect there's something wrong with that machine, but so far the guys owning the machine couldn't come up with a solution. 1105625626 M * TheSeer Bertl: wat is "stty: standard input: Inappropriate ioctl for device" trying to tell me? *g* 1105625658 M * pflanze If you run a single process eating cpu like perl -e 'for(1..10000000){}', they are both as fast (the athlon a little bit faster), 1105625658 M * Bertl hmm, that stdin isn't a tty? 1105625682 M * TheSeer that's from a the logfile of a domino server that used to run in the "crashed" vserver from ealier 1105625685 M * pflanze but as soon as you're doing real work with it like running tomcat and postgresql on the machines, both the java and db part are 4 times faster. 1105625698 M * TheSeer it refuses to start by now or actually keeps crashing on me.. 1105625713 M * TheSeer the domino server that is.. not the vserver 1105625731 M * Bertl pflanze: and the 4 times are the result of 'good observation' I guess? 1105625747 M * pflanze From measuring the possible hits/second rate. 1105625770 M * pflanze On both machines tomcat and postgresql are taking up about half of the cpu. 1105625780 M * Bertl i.c. and the system are otherwise identical (network, nic, disk system, memory) 1105625787 M * pflanze Thus I'm saying that it's not just java or postgresql which is slow, it's both., 1105625825 M * pflanze Both IDE disks, both 1G RAM, both 100mbit network (and network bandwidth is not the bottleneck, it's the cpu usage). 1105625856 M * pflanze There's one difference though: on the Athlon, vserver is running :)) 1105625888 M * pflanze And Debian, in contrast to Suse on the Intel. 1105625909 M * Bertl okay, so then let's agree on the following: amd over intel gives a 2 times performance improvement, and the same is true for linux-vserver over vanilla ;) 1105625924 M * pflanze :) 1105625941 M * Bertl (from close observation, of course) 1105626057 M * pflanze I'd say, there's a 1.26times improvement because of using the right dist, 1.26time because of using the right CPU, 1.26time because of using vserver, and 2 times because I'm the admin :) 1105626099 M * Bertl k, I can live with that ... 1105626135 M * pflanze Seriously, I'm suspecting it's the hyperthreading of the intel cpu. 1105626154 M * pflanze I guess the suse kernel is just broken in this respect (it's an older 2.4 SMP kernel). 1105626174 M * pflanze But since the guys owning the machine wouldn't turn off the ht in the bios, I don't know. 1105626231 M * pflanze "2.4.21-99-smp4G" 1105626247 M * pflanze built Wed Sep 24 14:13:20 UTC 2003 1105626461 M * Doener hm, sounds like a nice 'playground' ;) 1105626472 M * Doener Bertl: what is it that you want me for? 1105626507 M * Bertl well, I'v done a backport of 1.9.3.17 to 2.6.9 1105626532 M * Bertl to create a delta between 1.9.3 and 1.9.3.17 ... 1105626535 M * Doener why that? is 2.6.10 so bad for you? 1105626565 M * Bertl actually yes, there are many issues with 2.6.10 which seem linux-vserver unrelated 1105626593 M * Doener funny, i had freezes with 2.6.9 which are gone with 2.6.10 1105626607 M * Bertl but I wanted to get a precise delta to break down into smaller pieces ... 1105626644 M * Bertl (and the options were either backporting 1.9.3.17 or forward porting 1.9.3) 1105626683 M * Bertl anyway, I would like you to have a look at the broken out changes 1105626733 M * ndim Hmm. ifconfig displays IPv6 addresses in a vserver guest? 1105626740 M * Doener and another one falling for loadable caps support... *sigh* 1105626743 M * Doener (ml) 1105626791 M * Bertl http://vserver.13thfloor.at/Experimental/delta-split-vs1.9.3-vs1.9.3.17/ 1105626826 M * Bertl the first (huge) patch is an abstracted version of the changes done to restructure the includes ... 1105626857 M * Bertl (it should be clean of any functional changes as far as possible) 1105626900 M * Bertl the second patch is a nuber of changes in the included files, which should not have any side-effects 1105626935 M * Bertl the last patch XX_* is garbage, which I do not consider functional or useful 1105626981 M * Bertl 03_* are the interesting changes, where 06-09 are additions/features/fixes 1105627182 M * Bertl Doener: I'm not sure that the move to RCU is perfect yet, so I'd like to basically revert that change for 1.9.4 but keep it unchanged in 1.9.4.x (just redoing the refcounting, I have some ideas hwo to make this work) 1105627204 M * Bertl what is your opinion on that? 1105627307 M * Doener i agree, i still try to decide if rcu is the right choice for this 1105627475 M * Doener right now i'd say rcu is not made for this. most (all?) modifications we do to the contexts are lockless and rcu is just about removing those locks, that we don't have anyways. 1105627525 M * Bertl hmm, what about all the proc access (which was the reason for doing rcu in the first place)? 1105627618 M * Bertl (i.e. __lookup_vx_info() would require a lock then) 1105627977 M * Doener true... 1105628040 M * Bertl well, probably not that relevant, because a) it should only be for a short time, and b) the possibility for concurrency is relatively small ... 1105628071 M * Bertl (i.e. if we walk the hash chain for more than 5 elements, something _is_ broken) 1105628112 M * Bertl and of course, we could use a per chain lock (if that buys us anything) 1105628244 M * Doener not much i guess. 1105628595 M * Bertl okay, any other comments on the changes (especially 03_changes)? 1105628648 M * Bertl (take your time) 1105628667 M * Doener hm, why does it contain the reverse for a hunk in XX_junk? 1105628744 M * Bertl ah, which one? 1105628756 M * Doener the first hunk in both 1105628819 M * Bertl that's a whitespace cleanup ... 1105628864 M * Bertl (in this case replacing spaces by tabs at the beginning of a line) 1105628953 Q * pflanze Quit: [x]chat 1105629211 M * Seraph Bertl: when running x86_64 as kernel and 32bit (i386) userland.. i get this when running the vprocunhide script: 1105629214 M * Seraph /proc/sys/fs/aio-max-nr: Invalid argument 1105629216 M * Seraph /proc/sys/fs/mqueue/msgsize_max: Invalid argument 1105629219 M * Seraph /proc/sys/fs/quota/cache_hits: Invalid argument 1105629219 M * Seraph actually many more.. 1105629221 M * Seraph /proc/sys/dev/raid/speed_limit_max: Invalid argument 1105629224 M * Seraph /proc/sys/dev/cdrom/check_media: Invalid argument 1105629226 M * Seraph . 1105629236 M * Seraph is that due to the 32bit userland or is the kernel not properly vserver-enabled? 1105629274 M * Bertl it's more that most likely your distro has mounted sysfs on /proc/sys ... 1105629337 M * Seraph there's nothin in /etc/fstab. and mount only tells about /sys for sysfs 1105629351 M * Seraph but i can see /proc/sys actually 1105629367 M * Doener hm, those are not sysfs files 1105629368 M * Seraph so that's nothin my distro would have caused ;) 1105629397 M * Bertl Doener: right, that is /proc contents ... 1105629435 M * Bertl Seraph: do you have a complete log of what entries are reported? if so, could you upload it somewhere (e.g. pastebin.com)? 1105629443 M * Doener uhm, i guess i mixed up some stuff *crouches* 1105629451 M * Seraph it's 2.6.10 vanilla+1.9.3.16 compiled under i386 as x86_64 1105629497 M * Seraph Bertl: uploading isn't the prob.. but getting it ;) 1105629512 M * Seraph it's /etc/init.d/vprocunhide start 2>&1 >~/vprocunhide.log still echoing to console.. 1105629515 M * Seraph *hrmpf* 1105629524 M * Bertl what about vprocunhide >/tmp/log 2>&1 ? 1105629564 M * Bertl (the other way will not do what you expect, which is expected ;) 1105629632 M * Seraph http://vserver.lauft.normalerweise.net/mit/vprocunhide.log 1105629639 M * Bertl tx 1105629640 J * pflanze ~chris@unk-110.ethz.ch 1105629641 M * Seraph np 1105629748 M * Bertl do those entries exist on your system? 1105629789 M * Seraph # ls /proc/sys/net/ipv6/mld_max_msf -l 1105629789 M * Seraph -rw-r--r-- 1 root root 0 2005-01-13 16:23 /proc/sys/net/ipv6/mld_max_msf 1105629796 M * Seraph picked one at random.. 1105629799 M * Seraph looks fine 1105629827 M * Bertl okay, let's try to strace -fF the script ... 1105629895 M * Seraph http://vserver.lauft.normalerweise.net/wegen/vprocunhide.strace 1105630303 M * Seraph hmm, is my IRC broken or is there no comment on this yet? *g* 1105630363 M * Bertl well, it's a lot of log to read, don't you think? anyway, could you try 'setattr --admin /proc/net/snmp6' ? 1105630395 M * Seraph root@nyx:~# setattr --admin /proc/net/snmp6 1105630396 M * Seraph root@nyx:~# 1105630400 M * Seraph is fine 1105630424 M * Seraph Bertl: yes sure it's a lot ;) but usually a "very interesting" or something alike would help that the log is at least valid ;) 1105630437 M * Bertl hmm, very interesting ;) 1105630442 M * Seraph *g* 1105630712 M * Bertl no idea what would cause this ... 1105630726 M * Bertl do you have vserver debugging enabled? 1105630794 M * Seraph nope 1105630803 M * Bertl hmm, not really relevant, as we have no debug messages there (as I just found out ;) 1105630810 M * Seraph hehe ;) 1105630814 M * Seraph so how do i fix that? 1105630856 M * Bertl we do not even know why it happens, so we can't fix it right now .. if you want to track it down, you have to do some changes to the kernel ... 1105630872 M * Seraph tell me 1105630891 M * Bertl kernel/vserver/inode.c, line 104, let's add 1105630964 M * Bertl printk("%p [%u,%x,%x] is_proc=%d\n", de, xid, flags, mask, is_proc) 1105630973 M * Bertl +; 1105630978 M * Bertl right after: 1105630983 M * Bertl is_proc = (in->i_sb->s_magic == PROC_SUPER_MAGIC); 1105630993 M * Bertl and 1105631057 M * Bertl printk("%p [%p,%d,%d] has_xid=%d\n", de, in, in->i_sb->s_flags, in->i_sb->s_magic, has_xid); 1105631061 M * Bertl right after: 1105631068 M * Bertl has_xid = (in->i_sb->s_flags & MS_TAGXID) || 1105631068 M * Bertl (in->i_sb->s_magic == DEVPTS_SUPER_MAGIC); 1105631087 M * Bertl and at line 179 1105631091 M * Bertl before 1105631098 M * Bertl ret = user_path_walk_link(vc_data.name, &nd); 1105631140 M * Bertl printk("%p [%u,%s]\n", id, vc_data.name\n"); 1105631163 M * Bertl oops ... 1105631176 M * Bertl printk("%p [%u,%s]\n", id, vc_data.name); of course ... 1105631205 M * Seraph alright, that's it? 1105631227 M * Bertl yep, if the compiler complains, please let me know (we then have to fix something) 1105631248 M * Seraph alright, let's rock'n'roll ;) 1105631868 M * Seraph ok, new kernel is up 1105631959 M * Seraph http://vserver.lauft.normalerweise.net/wegen/vprocunhide/ 1105631963 M * Seraph i've added the dmesg output 1105631995 J * sannes ~ace@home.skarby.no 1105632094 Q * prae Quit: Lost terminal 1105632128 M * Seraph Bertl: hope that's ok.. ask for more if you need anything else =) 1105632785 M * Bertl tx 1105632788 M * Seraph np 1105632937 M * Bertl hmm, the last printk should spew out a warning (now that I see the output) 1105632960 M * Seraph so? 1105632970 M * Bertl because of the missing argument 1105633073 M * Bertl well, let's fix that up then ... 1105633081 M * Bertl instead of printk("%p [%u,%s]\n", id, vc_data.name); 1105633087 M * Bertl we actually want: 1105633203 M * Bertl ret = user_path_walk_link(vc_data.name, &nd); 1105633205 M * Bertl { 1105633221 M * Bertl char *tmp = getname(vc_data.name); 1105633223 M * Bertl 1105633253 M * Bertl printk("%p [%u,%s] err=%d\n", nd.dentry, id, tmp, ret); 1105633264 M * Bertl putname(tmp); 1105633269 M * Bertl } 1105633275 M * Bertl if (!ret) { 1105633277 M * Bertl ... 1105633288 M * Seraph ok, hold on a sec 1105633328 M * Seraph talking about warnings: 1105633329 M * Seraph CC kernel/vserver/inode.o 1105633330 M * Seraph kernel/vserver/inode.c: In function `__vc_set_iattr': 1105633330 M * Seraph kernel/vserver/inode.c:104: Warnung: unsigned int Format, Zeiger Argument (Argument 3) 1105633333 M * Seraph kernel/vserver/inode.c:104: Warnung: unsigned int Format, Zeiger Argument (Argument 4) 1105633336 M * Seraph kernel/vserver/inode.c:104: Warnung: unsigned int Format, Zeiger Argument (Argument 5) 1105633339 M * Seraph kernel/vserver/inode.c:111: Warnung: int Format, anderer Typ Argument (Argument 4) 1105633342 M * Seraph kernel/vserver/inode.c:111: Warnung: int Format, anderer Typ Argument (Argument 5) 1105633347 M * Seraph ... 1105633350 M * Seraph is that alright? 1105633354 M * Seraph or shall we fix that first? 1105633429 M * Bertl you really use a 'german' compiler? 1105633438 M * Seraph a german locale.. 1105633445 M * Bertl well, so be it ... 1105633449 M * Seraph *g* 1105633455 M * Seraph yep, so be it 1105633460 M * Bertl okay, let's see the lines in question ... 1105633462 M * Seraph but that doesn't answer the question 1105633468 M * Seraph ;) 1105633553 M * Bertl ah, okay, yes, we should change that ... 1105633597 M * Bertl printk("%p [%u,%x,%x] is_proc=%d\n", de, xid, flags, mask, is_proc); 1105633602 M * Bertl should actually be: 1105633617 M * Bertl printk("%p [%u,%x,%x] is_proc=%d\n", de, *xid, *flags, *mask, is_proc); 1105633682 M * Bertl and we can change the 1105633682 M * Bertl printk("%p [%p,%d,%d] has_xid=%d\n", de, in, in->i_sb->s_flags, 1105633683 M * Bertl in->i_sb->s_magic, has_xid); 1105633686 M * Bertl to 1105633706 M * Bertl printk("%p [%p,%d,%d] has_xid=%d\n", de, in, (int)in->i_sb->s_flags, 1105633707 M * Bertl (int)in->i_sb->s_magic, has_xid); 1105633747 M * Seraph jup, no more complaints from gcc ;) 1105633780 M * Bertl good, please let's redo the vprocunhide then (with the new kernel) 1105633800 M * Seraph alright ;) 1105633804 M * Seraph just a sec.. 1105633985 M * Seraph ok, all files updated 1105633992 M * Bertl tx 1105633998 M * Seraph np 1105634062 M * Seraph Bertl: mmh, how long will you be in tonight? 1105634075 M * Bertl probably some time ... 1105634075 M * Seraph Bertl: i should somewhat get going.. and would be back later tonight.. 1105634076 M * Seraph ok 1105634090 M * Seraph if you have any wishes, thoughts.. just msg me in the meantime 1105634098 M * Bertl just one last question 1105634101 M * Seraph sure 1105634129 M * Bertl hmm, no, I can answer that myself, thanks! 1105634138 M * Seraph *g* 1105634139 M * Seraph ok 1105634350 M * brc BERTL 1105634353 M * brc what was the /proc trick ? 1105634364 M * Bertl hmm, trick? 1105634378 M * brc vcontext: vc_create_context(): File exists 1105634379 M * brc to fix this 1105634398 M * Bertl mount -o remount,rw /proc (did help sometimes) 1105634398 Q * sannes Read error: Connection reset by peer 1105634413 M * brc didnt help 1105634422 M * brc is there any other way of "killing" that context ? 1105634428 M * brc the vserver is allready stoped. 1105634436 M * Bertl nope, it has still some reference ... 1105634448 M * Bertl (check /proc/virtual//info ) 1105634451 Q * BWare Quit: using sirc version 2.211+KSIRC/1.3.10 1105634462 M * brc [root@localhost root]# cat /proc/virtual/1111/info 1105634462 M * brc stupidawy: 1111 1105634462 M * brc Info: daa5e000 1105634462 M * brc Init: 0 1105634472 M * Bertl ( and probably /proc/virtual//limit) 1105634485 M * brc [root@localhost root]# cat /proc/virtual/1111/limit 1105634485 M * brc PROC: 0 88 -1 0 1105634485 M * brc VM: 5356 661080 -1 0 1105634485 M * brc VML: 0 4 -1 0 1105634485 M * brc RSS: 271 39990 -1 0 1105634486 M * brc FILES: 60 669 -1 0 1105634486 M * brc SOCK: 0 0 -1 0 1105634488 M * brc [root@localhost root]# 1105634522 M * Bertl hmm, that one is new ... sec 1105634533 M * rs I have exactly the same issue with some customers 1105634544 M * Bertl okay, that one is known ;) 1105634592 M * Bertl brc: what do the other entries in /proc/virtual/ contain? could you upload that somewhere? 1105634601 M * brc yes. hold 1105634653 M * Bertl rs: you see the same with PROC=0 ? 1105634682 M * Bertl (i.e. VM and RSS used, but no processes?) 1105634695 M * brc http://brc.ath.cx/~bruce/proclog 1105634705 M * rs Bertl: yeah 1105634706 M * Bertl and do you have VSERVER_DEBUG enabled on those systems? 1105634712 M * rs bvdsn001:~# cat /proc/virtual/787/limit 1105634712 M * rs PROC: 0 25 1000 0 1105634712 M * rs VM: 370 22804 2560000 0 1105634712 M * rs VML: 0 0 -1 0 1105634712 M * rs RSS: 122 11050 131072 0 1105634714 M * rs FILES: 4 213 5000 0 1105634716 M * rs SOCK: 0 8 -1 0 1105634728 M * brc hmm. i haven't touched vserver_debug.. if it is enbaled by default , yes.. if not, i dont have ! 1105634738 M * rs Bertl: lets check this 1105634770 M * rs Bertl: yes 1105634783 Q * Snow-Man Ping timeout: 480 seconds 1105634802 M * Bertl okay, let me think about that for a moment ... 1105634821 M * brc There are other xids with proc=0.. The old ones which are still locked. 1105634843 M * rs Bertl: this context is blocked for a day now 1105634879 M * Bertl do we have more than one context hanging? 1105634902 M * Bertl if so, what are the ratios of VM/FILES for them? 1105634942 M * Bertl and what does 'grep Cnt /proc/virtual//status' say? 1105635003 M * rs there is two hanging context on this machine for instance 1105635028 M * brc i though you were talking with me 1105635029 M * brc hhehe 1105635043 M * Bertl I'm talking to both of you, actually ;) 1105635063 M * ntrs By the way I just want to say that we have the same issue, vservers stopped and then cannot be started because the context is still alive. A workaround is to change the ctx number in the vserver config and then restart. 1105635064 M * rs /proc/virtual/738/limit:VM: 11898 284396 2560000 0 1105635067 M * rs /proc/virtual/738/limit:FILES: 34 497 5000 0 1105635070 M * rs /proc/virtual/787/limit:VM: 370 22804 2560000 0 1105635073 M * rs /proc/virtual/787/limit:FILES: 4 213 5000 0 1105635110 M * Bertl okay, ntrs, rs, brc, let's try to find a common nominator/pattern ... 1105635123 M * ntrs But the old context will still be alive even when you start the new one with a different ct 1105635125 M * ntrs ctx 1105635131 M * Bertl my assumptions so far: 1105635138 M * brc bertl 1105635139 M * brc http://brc.ath.cx/~bruce/procvirtual.tar.gz 1105635146 M * brc i targziped /proc/virtual 1105635148 M * ntrs If it helps any, this has been going for a while now 1105635150 M * rs this command would help to find info if interested: grep -lE 'PROC:[[:space:]]*0' /proc/virtual/*/limit|xargs grep -E 'VM:|FILES:' 1105635166 M * Bertl (assumption A) VM, RSS and FILES are != 0 1105635185 M * brc and here you have the active contexts http://brc.ath.cx/~bruce/active-contexts 1105635190 M * Bertl (assumption B) PROC = 0 1105635190 J * Snow-Man ~sfrost@snowman.net 1105635204 M * rs Bertl: I have another machine wiht a lot more hanging contexts 1105635218 M * rs like 21 contexts hanged 1105635223 M * Bertl (assumption C) UseCnt > RefCnt > 0 1105635273 M * Bertl (assumption D) nr_threads = nr_running = nr_onhold = 0 1105635294 M * ntrs Assumption A is not necessarily true 1105635297 M * Bertl is there any 'hanging' context which violates any of those assumptions so far? 1105635299 M * ntrs Here's a hanging ctx 1105635300 M * ntrs # cat /proc/virtual/102/limit 1105635300 M * ntrs PROC: 0 200 200 11 1105635300 M * ntrs VM: 453 782630 -1 0 1105635300 M * ntrs VML: 0 0 -1 0 1105635300 M * ntrs RSS: 0 255312 262144 0 1105635302 M * ntrs FILES: 5 2203 -1 0 1105635305 M * ntrs SOCK: 0 140 -1 0 1105635311 M * ntrs RSS is 0 1105635312 M * rs here it's true 1105635337 M * rs yeah me too but for all the files counter is != 0 (as well as VM) 1105635342 M * Bertl okay, so we have to reduce A to VM, and FILES != 0 1105635399 M * Bertl now are there some vservers which 'prefer' to hang, and others which shutdown cleanly? 1105635406 M * ntrs B is correct 1105635409 M * ntrs C is correct 1105635420 M * brc Bertl: thats it 1105635421 M * Bertl (or is it more a random, sometimes it hangs, the other time it's fine) 1105635434 M * brc everythingis correct 1105635439 M * brc sometimes it hangs 1105635441 M * brc after a long time up 1105635443 M * brc they hang 1105635451 M * brc i just changed the context for this vserver and i can stop/start it 1105635454 M * ntrs Both examples I have here have hit the max process limit at least once 1105635460 M * brc after smoe time up they start locking up 1105635505 M * ntrs Yes, usually after some time up they do that. If you just start it with a new ctx number you can probably start and stop them numerous times without a problem 1105635518 M * rs none of them here (about the reached limit) 1105635519 M * ntrs maybe it has something to do with the max process limit 1105635527 M * ntrs oh, ok, forget that then 1105635532 M * Bertl okay, so there is no 'test' vserver which definitely 'hangs' every time you start it, right? 1105635543 M * rs Bertl: no 1105635560 M * ntrs not that I can remember 1105635568 M * rs here the work around is to reboot the vserver on another node, and the vserver can work well for days or month then 1105635610 M * ntrs But I am glad that finally I can prove myself that I am not going crazy. I have brought this up several times in the past and the conslusion was that it was probably just me and my servers 1105635611 M * ntrs :) 1105635625 M * rs :) 1105635635 M * ntrs Right on the same server you have to change the ctx number. 1105635654 M * ntrs You also cannot bring that vserver to that same node anymore at least until you reboot that node. 1105635685 M * ntrs It's good that this is getting some mainstream attention. :) 1105635709 M * Bertl hmm, okay, another (probably hard to verify) question ... 1105635712 M * rs btw, assumption C is verified here too 1105635724 M * rs for all cases 1105635739 M * brc Are those lcoked contexts consuming memory ? 1105635757 M * brc if so, it would be interesting to have a host reboot somtiems 1105635758 M * brc hehe 1105635810 M * Bertl ntrs: well, that's a known problem with people ;) 1105635810 M * Bertl just remember the astronauts seing flashes of light, and not reporting them ... because they thought it is something which should not happen (at least not to them ;) 1105635810 M * Bertl okay, the question: 1105635810 M * Bertl is there a vserver which has not transmitted any networking packets yet but is hanging? 1105635812 M * ntrs I just saw one other case here that did not hit the max processes limit and has hung too 1105635834 M * rs and D too 1105635860 M * ntrs No. This only happens after the vserver has been up for quite some time and there was definitely some network traffic 1105635867 J * Wicked187 ~ddortch@68-248-179-131.ded.ameritech.net 1105635880 M * Wicked187 hey al 1105635885 M * Wicked187 I am in a pickle 1105635896 M * Bertl welcome Wicked187! 1105635899 M * Wicked187 I run Slackware 10.0 as my base system 1105635922 M * Wicked187 I have a nice tape library, and I am looking for some nice backup software 1105635932 M * Wicked187 I have tried NovaNet by Novastore 1105635936 M * rs Bertl: seems to not happen here too 1105635936 M * Wicked187 and it is horrible 1105635948 M * Bertl Wicked187: hmm, and how is that linux-vserver related? 1105635953 M * Wicked187 So, we are looking at Brightstor from Computer Associates... 1105635954 M * rs all hanging contexts transfered something 1105635964 M * Wicked187 they only support RedHat and SuSe 1105635992 M * Wicked187 would it be possible to setup a child environment with Fedora that could properly access the tape library hardware? 1105636020 M * Bertl guess so, just add the appropriate devs 1105636026 M * Wicked187 okie 1105636042 M * Wicked187 it is probably going to work out kinda of wierd... 1105636057 M * TheSeer Wicked187: how do they only support a specific distro? 1105636060 M * Wicked187 I would need to access the root server to backup everything 1105636067 M * Bertl rs, ntrs, brc: do we have a context with a very low FILES number? 1105636071 M * Wicked187 TheSeer: that is what they say 1105636081 M * TheSeer Wicked187: well there is support and there is support 1105636083 M * rs Bertl: all 1105636083 M * Wicked187 and I have had difficulties getting to to work 1105636103 M * TheSeer Wicked187: so they deliver a binary package with a kernel module or something? 1105636105 M * Wicked187 they will not provide support, and they cannot guarantee it will work 1105636108 M * Bertl rs: one with FILES=1 maybe? 1105636117 M * Wicked187 TheSeer, they provide RPMs 1105636123 M * TheSeer src rpms? 1105636129 M * Wicked187 and I ran rpm2tgz and installed them 1105636132 M * Wicked187 just rpms 1105636138 M * TheSeer okay.. what's in those? 1105636142 M * Wicked187 this is computer associates 1105636143 M * rs lowest I have is 3 1105636145 M * Wicked187 no source 1105636147 M * TheSeer kernel modules? 1105636161 M * Wicked187 no, it doesn't appear to be kernel modules 1105636166 M * Bertl brc, ntrs: can you do better? ;) 1105636171 M * TheSeer then the distro shouldn't really matter 1105636173 M * rs lots of 4 and some others with 7-22 1105636183 M * Wicked187 that is how I feel about it 1105636188 M * Wicked187 but I cannot get it to work 1105636197 M * TheSeer what's its problem? 1105636203 M * Wicked187 well 1105636206 M * TheSeer uhm.. query ;) this is getting off topic 1105636208 M * Wicked187 there are so many components 1105636211 M * Wicked187 sure 1105636235 M * ntrs Bertl, I am checking 1105636235 M * rs Bertl: it's really often 3-5 1105636251 M * Bertl rs: okay, let's see the complete /proc for all <= 3 if possible 1105636270 M * Bertl (i.e. the /proc(virtual/ that is) 1105636271 M * ntrs I have files 5 or more 1105636284 M * rs ok 1105636357 M * Bertl hmm, brc, I see VML: 0 4 -1 0 1105636361 M * ntrs lowest I have is 5 do you want the complete proc for those too? 1105636427 M * Bertl hmm, guess it would be useful to dump the /proc/virtual/ for all of those 'hanging' contexts just to make sure (and upload it somewhere) 1105636497 M * brc lemme check 1105636499 M * Bertl maybe with the one liner brc did 1105636508 M * brc Bertl: i've uploaded mine 1105636524 M * rs http://rs.rhapsodyk.net/vserver/hanging-file-le-3.txt 1105636539 M * rs there is only two context in that case btw (on this node) 1105636547 M * rs you want the full list too ? 1105636601 M * Bertl yes, if possible (but in a way that I can tell the xid and the entry) 1105636613 M * rs k 1105636672 M * Bertl XID=100; cd /proc/virtual/$XID; echo "========= $XID =========" ; for OPA in *; do echo "-------------- $OPA --------------"; cat $OPA; done 1105636677 M * Bertl (for example) 1105636830 M * brc Do this 1105636832 M * brc cd /proc/virtual 1105636834 M * brc for OPA in *; do echo ============== CONTEXT $OPA ===================; for OPA2 in $OPA/*; do echo -------------------$OPA:$OPA2; cat $OPA2; done; done 1105636847 M * brc dont forget to > /logfile 1105636847 M * brc :) 1105636876 M * brc hmm 1105636879 M * brc only the locked ones? 1105636879 M * brc ok 1105636917 M * Bertl yep, just the locked ones 1105636974 M * Bertl another assumption: 1105637003 M * Bertl (assumption E) UseCnt - RefCnt = 2 1105637021 M * Bertl (please prove me wrong on this one ;) 1105637061 M * rs http://rs.rhapsodyk.net/vserver/hanging-ctx-dump.txt 1105637120 M * rs Bertl: I can't :) 1105637138 M * rs E is true 1105637197 M * Bertl brc: you are on an UP machine, right? (the hanging context I mean) 1105637293 M * brc done 1105637303 M * brc Bertl: what is a UP machine?? 1105637338 M * brc http://brc.ath.cx/~bruce/proc-locked-xids 1105637339 M * ntrs Bertl, all our machines are SMP 1105637343 M * brc Ahh 1105637346 M * brc Up machine right!! 1105637411 M * ntrs Bertl, assumption E is correct here 1105638150 J * timothy ~timothy@212.123.1.4 1105638166 M * timothy good evening 1105638170 M * Bertl welcome timothy! 1105638178 M * timothy heyhey :) 1105638227 M * timothy got a question, can't seem to find the answer. Inside a vserver is it actually possible to see the ip with ifconfig? 1105638249 M * Bertl yes, if you use aliases for the vserver ips ... 1105638278 M * Bertl (otherwise you have to use ip from iproute2) 1105638424 M * timothy I was under the impression those changes can't be made inside a vserver 1105638430 M * timothy with ip that is 1105638443 M * Bertl which changes? 1105638458 M * timothy changes to ip settings and such 1105638485 M * Bertl yes, they are not allowed (yet), but I'm missing the conenction here?! 1105638527 M * timothy I really need vserver to know what ip it's got 1105638547 M * Bertl every vserver 'knows' what ip he has ... 1105638574 M * timothy the OS that's running in a given vserver 1105638617 M * Bertl as I said, you have two options here ... 1105638631 M * Bertl a) use the 'ip' tool from iproute2 (inside the vserver) 1105638647 M * Bertl b) configure the vserver to use aliases 1105638717 M * timothy ok will try the aliases first 1105638718 M * timothy cheers 1105638763 M * Bertl you're welcome! 1105638820 P * timothy 1105638979 M * Bertl okay, if somebody else is interested in looking at the hanging contexts, the collected information is available here: http://vserver.13thfloor.at/Stuff/HANG/ 1105639216 M * Bertl ntrs, rs, brc: another question: what did you do when you discover that a vserver hangs? 1105639229 M * Bertl a) stop a bunch of vservers then restart them 1105639238 M * Bertl b) stop a single vserver then restart it 1105639265 M * Bertl c) just stop a vserver 1105639298 M * rs I don't understand your question :) 1105639311 M * Bertl okay, let me rephrase ... 1105639345 M * Bertl when you discover that context X hangs ... and you think about what you did when you stopped that vserver ... 1105639367 M * Bertl is it usually true that you a) did stop several vservers at once 1105639383 M * Bertl b) you did just stop and start context X 1105639405 M * Bertl c) you somehow discover that the context hangs after stopping it (and only that context) 1105639428 M * rs hmm it wasn't stopped/started by me but by the user 1105639462 M * rs I often discover it when the user send a message on our support board to tell "hey! my vserver can't restart" 1105639486 M * Bertl okay, so we don't know exactly in your case ... 1105639495 M * rs and that situation doesn't happen each time a context hang, because most of the time, when you restart a vserver on our arch, the vserver is started on another node 1105639539 M * rs it only happen when the node on where you was is the less loaded node 1105639563 M * Bertl ah, that is an interesting aspect, are there contexts which hang, although they are restarted on a different one? 1105639591 M * Bertl (i.e. is stopping it enough to hang it, or is it a restart/reboot which causes it) 1105639638 M * rs we never restart vservers 1105639644 M * rs they are always stopped 1105639663 M * rs then started on another node or the same in some cases (explained earlier) 1105639667 M * Bertl hmm, what about reboot via vshelper? 1105639692 M * rs yeah this case is the only case when we have restarts yes :) 1105639731 M * Bertl do we know that there are cases where this didn't happen? 1105639764 M * Bertl (i.e. do we know of a vserver which did not execute a vshelper reboot, but does hang?) 1105639789 M * rs no sorry 1105639791 M * Bertl ntrs, brc: are you using the vserver side reboot at all? 1105640188 M * rs gotta leave, Bertl: if you have further question, please drop me a mail 1105640200 M * Bertl okay, thanks! 1105640211 M * rs cya folks 1105640216 Q * rs Quit: leaving 1105640849 J * prae ~prae@sherpadown.net 1105640860 M * prae 'noon all :) 1105640997 M * Bertl welcome prae! 1105641009 M * prae hi Bertl 1105641569 J * sannes ~ace@home.skarby.no 1105641578 M * Bertl welcome sannes! 1105641622 M * ntrs Bertl, I usually try to stop a vserver to make a config change for example and then try to restart it. So b) applies to me. 1105641645 M * Bertl okay, thanks! 1105641763 M * ntrs And we do not use the vserver side reboot at all lately, because we thought this issue could lead to problems. Plus no client needs to reboot a vserver, since they cannot do hardware changes and they cannot upgrade the kernel, so why reboot? 1105641874 Q * sannes Read error: Connection reset by peer 1105642055 J * nayco ~nayco@lns-vlq-47-nan-82-252-224-206.adsl.proxad.net 1105642060 Q * serving Ping timeout: 480 seconds 1105642060 M * nayco 'llo ! 1105642116 M * Bertl welcome nayco! 1105642171 M * Bertl ntrs: the idea behind the reboot is to check if the service/application setup does work as intended (when the host will be restarted) without restarting the host ... 1105642245 M * ntrs I know 1105642264 M * ntrs That's a legitimate use. But we can assist with any such requirement. 1105642280 M * Bertl sure, no problem with that ... 1105642289 M * ntrs Anyway, just wanted to let you know what the situation here was 1105642302 M * ntrs I will be back in about 30 minutes. 1105642310 N * ntrs ntrs_away 1105642607 M * Bertl okay, thanks! cya! 1105642785 M * nayco Bertl: Still got a problem with aliases (util-vserver devel).. I build and configure a vserver, and when starting it, it cannot find the network alias for eth0... Because it's not created automatically (As said here previously). So, I create "ifcfg-eth0:0" manually, then bring it up (ifup eth0:0), but the "vserver start" still doesn't work... The "/etc/vservers/..../interface/..." are correctly configured by "vserver...build" . 1105642815 Q * prae Quit: Pwet 1105642843 M * nayco Note: I deleted all the symlinks in /vservers/.../etc/rc.d/rc*.d/ 1105642884 M * Bertl nayco: you've got mail ;) 1105643151 P * jsambrook 1105643176 N * ntrs_away ntrs 1105643225 M * nayco yep, but I left my imap mailer on at work, so I cant get my mail here (home)... :( 1105643233 M * nayco Thanks anyway ;-) 1105643240 M * Bertl np, I can repeat it ... 1105643267 M * Bertl very likely the issue is that you have (somehow) misconfigured your vserver ... 1105643287 M * Bertl so what I would like to see is the (complete) build command you used 1105643301 M * nayco ok: 1105643348 M * Bertl http://archives.linux-vserver.org/200501/0063.html 1105643384 M * Bertl sladen: thanks for the real-time archive! 1105643402 M * nayco thank, i gonna read it now. 1105643415 M * nayco I cant remember the exact command 1105643463 M * nayco but it looked like the one found on the pdf presentation (In german ;-)) found somewhere on the utils-vservers site. 1105643466 M * nayco : 1105643612 M * Bertl let take the example from http://linux-vserver.org/alpha+util-vserver 1105643625 M * Bertl and adjust it so that it is likely to fail for you ;) 1105643639 M * Bertl * vserver foo build -m apt-rpm --hostname vs.foo.org --netdev eth0 --interface 192.168.3.1/21 --context 42 -- -d fc1 1105643664 M * Bertl if you change the eth0 to eth0:wossname it will probably fail ... 1105643720 M * Bertl what you initially wanted to do is more like this: 1105643750 M * nayco I cant find the pdf... So, I give it a try: vserver modele build -m skeleton --(network stuff) 1105643775 M * Bertl --interface modele=eth0:192.168.3.1/21 1105643790 M * nayco mkdir -p /vservers/modele/var/lib/rpm 1105643799 M * nayco urpmi --root /vservers/modele basesytem 1105643821 M * nayco that's all.... Then vserver modele start 1105643825 M * Bertl yep, but the --(network stuff) is the important part you got wrong, IMHO 1105643845 M * nayco I know :((( I'm googling around.... 1105643867 M * Bertl better read (and understand) what I pasted here ;) 1105643963 M * nayco Ok : # vserver test2 build -m skeleton --hostname test2.nowhe.re --interface 10.0.1.2 --netdev eth0 --netprefix 23 1105643977 M * nayco this is what i read. So, no, my command line: 1105644030 M * nayco # vserver modele build -m skeleton --hostname modele.f.q.d.n --interface 172.x.x.x --netdev eth0:modele --netprefix 22 1105644038 M * sladen Bertl: np 1105644047 M * nayco ...er.... Is "22" the netmask ? 1105644047 M * Bertl nayco: which is wrong! 1105644055 M * nayco ahhhh :) 1105644066 M * Bertl there is no device 'eth0:modele' 1105644073 M * nayco uhh ? 1105644080 M * Bertl so specifying a device 'eth0:modele' will fail 1105644093 M * nayco I got to create it _before_ ? 1105644100 M * Bertl eth0:modele is an _alias_ 1105644107 M * nayco argghhhh... 1105644120 M * Bertl you do not need to create it beforehand 1105644126 M * nayco so, eth0 is enough ? 1105644135 M * Bertl you have 3 options here ... (listen closely) 1105644137 J * CosmicRay ~jgoerzen@2002:4463:7269:1:20e:a6ff:fe66:c5a3 1105644146 M * Bertl welcome CosmicRay! 1105644153 M * nayco yep ? 1105644158 M * CosmicRay Nice to see you again bertl 1105644170 M * nayco 'llo, CosmicRay ! 1105644174 M * CosmicRay you can tell whenever I set up a new box with vserver -- I delurk for 2 days :-) 1105644186 M * Bertl nayco: a) you create it beforehand (as you do now) and use --interface 172.x.x.x/22 without any --netdev 1105644198 M * CosmicRay which is a testament to vserver. Our main machine has been running without a hitch for a year. 1105644206 M * CosmicRay hi nayco 1105644216 M * nayco Well, This channel is logged on my machine, good ;-) 1105644234 M * CosmicRay I do have somewhat of a theory question. 1105644235 M * Bertl nayco: b) you use --netdev eth0 --interface 172.x.x.x/22 (which will work fine but not create any alias) 1105644264 M * CosmicRay I am wanting to run vserver for a personal project. I have an Alpha and 1 static IP from my ISP. I have some things listening in the master environment. 1105644276 M * CosmicRay I am wanting to set up a vserver or two on the machine 1105644281 M * Bertl nayco: c) you use --interface modele=eth0:172.x.x.x/22 (which will create the alias for you, ons tartup) 1105644289 M * CosmicRay to isolate things from everything else (this server is also my firewall) 1105644306 M * nayco Ah: C) is seems the best option, thanx :))) 1105644306 M * Bertl sounds fine ... 1105644309 M * CosmicRay I thought I could maybe set up dummy0 on an unused netblock, bind the vserver to it, and use SNAT to redirect to it... 1105644316 M * CosmicRay does that make sense? 1105644318 M * nayco Bertl: Thanks ! 1105644325 M * Bertl nayco: you're welcome! 1105644330 M * CosmicRay or would I be better of just locking it to the IP of my external interface? 1105644333 M * CosmicRay s/of/off/ 1105644348 M * CosmicRay (can I lock several vservers to the same IP?) 1105644360 M * CosmicRay assuming they don't bind to the same ports, of course... 1105644365 M * Bertl CosmicRay: both should be fine ... but the SNAT is propably more flexible ... 1105644388 M * CosmicRay is it even possible to lock a vserver to one of the IPs in use by the "master" server? 1105644398 M * Bertl yes 1105644401 M * nayco Bertl: So basically " --netprefix 22" is not useful ? 1105644405 M * CosmicRay sweet. 1105644421 M * CosmicRay Bertl: btw deleting that one line is letting my kernel compile. thanks. 1105644428 M * Bertl nayco: not required if you specify it with --interface (as done in the examples) 1105644432 M * Bertl CosmicRay: thought so ... 1105644461 M * Bertl CosmicRay: let me know if you encounter any issues on Alpha ... (and check with the testme.sh script) 1105644511 J * _are_ ~are@dsl-084-056-158-001.arcor-ip.net 1105644524 M * Bertl welcome _are_! 1105644544 M * CosmicRay Bertl: will do. btw, are you supporting amd64? 1105644562 M * _are_ hi 1105644585 M * Bertl CosmicRay: yes, the list of 'supported' archs is here ... http://linux-vserver.org/Syscall+Switch+Info 1105644608 M * _are_ samba is unable to use oplocks within a vserver, neither kernel, nor normal ones, it seems. anyone with a hint here? 1105644620 M * Bertl CosmicRay: (well, actually there are more arch supported than listed there) 1105644623 M * CosmicRay Bertl: wow, that is a much larger list than it was last a looked. 1105644632 M * CosmicRay Bertl: I was just going to say alpha is missing :-) 1105644646 M * Bertl http://linux-vserver.org/Linux-VServer-Paper-10 1105644655 M * Bertl (this one is better ;) 1105644667 M * CosmicRay excellent 1105644690 M * Bertl _are_ where does it fail? 1105644709 M * CosmicRay _are_: I just disable oplocks on my samba installation... dunno. 1105644766 M * _are_ Jan 13 02:47:02 samba smbd[28952]: [2005/01/13 02:47:02, 0] smbd/oplock.c:receive_local_message(187) 1105644766 M * _are_ Jan 13 02:47:02 samba smbd[28952]: receive_local_message: invalid 'from' address (was 10a010a should be 127.0.0.1) 1105644782 M * _are_ and after some tries: 1105644791 M * _are_ Jan 13 02:47:34 samba smbd[27245]: [2005/01/13 02:47:34, 0] smbd/oplock.c:request_oplock_break(1055) 1105644791 M * _are_ Jan 13 02:47:34 samba smbd[27245]: request_oplock_break: no response received to oplock break request to pid 28952 on port 53950 for dev = fd0c, inode = 382955, file_id = 222 1105644834 M * Bertl hmm, seems like samba can't handle the missing localhost properly ... 1105644899 M * Bertl maybe there is a way to tell the smbd that should be treated identical to localhost (127.0.0.1) 1105644911 M * _are_ well, I already gave the machine 127.0.0.1 1105644920 M * _are_ won't help 1105644933 M * Bertl yes, because the networking code will rewrite it ... 1105644963 M * _are_ alternativly I tried to use the interface binding, but this part still won't work, helped with other issues 1105644968 M * nayco someone wrote on the the missing LO on his site... 1105645046 M * nayco http://mirabellug.org/wikini/wakka.php?wiki=VServers 1105645060 M * nayco But it's in french... Is this guy right ? 1105645175 M * nayco 1105645175 M * nayco "reduce the netmask to 16 : 127.0.0.1/8 => 127.0.0.1/16 1105645175 M * nayco then use dummy in /etc/modprobe.conf : 1105645175 M * nayco options dummy numdummies=5 1105645175 M * nayco 1105645194 M * nayco and in the vserver conf: 1105645197 M * nayco IPROOT="dummy1:127.1.0.1 eth0:10.0.1.3" 1105645197 M * nayco IPROOTDEV= 1105645207 M * nayco finally in /etc/hosts: 1105645212 M * nayco 127.1.0.1 localhost 1105645214 M * nayco 1105645477 M * nayco That's what I read, but I don't understand it all... So I cannot say this work for samba. 1105645478 M * nayco works 1105645478 N * nayco nayco_smokes 1105645500 J * prae ~prae@sherpadown.net 1105645537 M * _are_ hmm 1105645684 M * Bertl as far as I understand it, it's a funny way, but will not help here ... 1105645711 M * _are_ well, the good way would probably be patching samba to honour it's own settings. 1105645750 M * _are_ after all it is only allowed to bind to 10.1.10.1 accoding to its config, so why does it insist on 127.0.0.1? to me it looks like a bug. 1105646863 N * chris|sleep chrish01 1105647098 M * Bertl morning chrish01! 1105647143 M * chrish01 hey Bertl, how goes? 1105647160 M * Bertl much work to do, but everything fine so far ... 1105647176 M * chrish01 good 1105647215 M * Bertl and how are you? 1105647294 M * chrish01 pretty good...i too am swamped with work 1105647340 M * chrish01 i do a lot with the mono project...so between virtual server/router stuff and that, free time is sparse :) 1105647378 M * Bertl how is mono? 1105647410 M * chrish01 its great...lots of new stuff going to be in 1.2 release (april or so) 1105648335 M * brc bertl still there? 1105648339 M * brc i was away now i am back 1105648346 M * Bertl yep, still there 1105648347 M * brc your question about how we discover the context locked 1105648364 M * brc i myself stopped the vserver, then when i start it fails with "context allready in use" 1105648380 M * brc all the cases were like that 1105648385 M * brc and i was not stop/starting other vservers, just that one 1105648405 M * Bertl so it's more a single server stop/start instead of a multiple server stop, or a vserver reboot, right? 1105648442 M * brc right! 1105648454 M * brc i had a case 1105648464 M * brc where i had to stop all vservers, and all of them had been running for a long time 1105648469 M * brc almost every context was locked up 1105648473 M * brc so i had to change all tehir contexts. 1105648490 M * brc the ones which didn't hang were the ones which were not used 1105648497 M * brc The owner didn't even connect to it 1105648508 M * Bertl i.c. 1105648534 M * brc it's weird cause there isn't a "logic" 1105648541 M * brc i am checking here the vservers 1105648548 M * brc i was using contexts from 100 to 116 1105648561 M * brc when they locked up, i putted a "1" in front of the context number 1105648570 M * brc so 100 becamse 1100 .... 101 becamse 1101 1105648584 M * brc and i still have 7 on the "---" 1105648630 N * nayco_smokes nayco 1105648682 M * Bertl ntrs, brc: another question: you all said vservers running for some time hang ... how long was the shortest time such a vserver had to run until it did hang? (rough estimations) 1105648816 M * ntrs I have no idea but if I would have to guess I would say days at least, maybe a couple of weeks. 1105648854 J * serving ~serving@213.186.182.244 1105648882 J * sannes ~ace@home.skarby.no 1105649071 M * brc Bertl: maybe a week or two 1105649077 M * brc i think it doesn't have to do with time 1105649085 M * brc but with the "things" done inside a vserver 1105649092 M * Bertl hmm, like? 1105649110 M * brc the most "used" ones 1105649178 M * Bertl brc, ntrs: another one: do you do some polling on /proc/virtual/* to provide some kind of records or stats (maybe graphs)? 1105649534 M * brc No 1105649541 M * brc i didn't even know about the existence of it :) 1105649565 M * Bertl okay, that is a valuable information ;) 1105649617 M * brc hehehe 1105649969 J * Tbery ~tb@rt-pha-1.karneval.cz 1105649973 M * Tbery HI 1105649984 M * Bertl welcome Tbery! 1105650001 M * Tbery I had trobles... 1105650021 M * Tbery but today I finish.. 1105650028 M * Tbery it.. 1105650042 M * Bertl k 1105650051 M * Tbery but when Im writing vservers-stats 1105650057 M * Tbery but when Im writing vserver-stat 1105650080 M * Tbery after is show only 2 servers but running 5 1105650085 M * Tbery why?? 1105650133 M * Bertl because just 2 are running the other three were started but stopped automatically after the last process died ... 1105650161 M * Tbery no they are running.. 1105650173 M * Tbery ssh...dhcp..apache.. 1105650217 M * Bertl if you enter the context, do you get them listed by ps? 1105650294 M * Tbery how?? 1105650297 M * Tbery to try?? 1105650308 M * Bertl do they have names? 1105650315 M * Tbery yes 1105650330 M * Tbery quido netboot aptcacher named2 1105650334 M * Bertl then try 'vserver exec ps' 1105650354 M * Tbery ain:~# vserver named2 exec ps 1105650354 M * Tbery ipv4root is now 192.168.0.100 1105650354 M * Tbery New security context is 49152 1105650354 M * Tbery PID TTY TIME CMD 1105650354 M * Tbery 4598 pts/0 00:00:00 ps 1105650355 M * Tbery smain:~# 1105650370 M * Bertl so no service is running there ... 1105650386 M * Bertl try again with: 1105650401 M * Bertl 'vserver exec ps aux' 1105650415 M * Tbery runing.. 1105650432 M * Tbery all are runing.. 1105650443 M * Bertl does this list some processes? 1105650450 M * Tbery but but whenvserver-stat 1105650450 M * Tbery CTX PROC VSZ RSS userTIME sysTIME UPTIME NAME DESCRIPTION 1105650450 M * Tbery 0 49 78MB 9kB 1m18.77 m17.63 1h30m58 root server 1105650450 M * Tbery 49152 60 803MB 161kB m03.91 m00.80 1h06m17 quido 1105650450 M * Tbery smain:~# 1105650511 M * Tbery it was working.. 1105650520 M * Tbery but no isn|t.. 1105650533 M * Tbery can be troble somewhere is proc?? 1105650568 M * Bertl could be, but more likely you have no services configured inside your vservers, so none will be started on startup 1105650588 J * lithium ~lithium@CPE005004cf14b9-CM024350002823.cpe.net.cable.rogers.com 1105650591 M * Bertl (this leaves no process/service running inside, and terminates the context right after startup) 1105650598 M * Bertl welcome lithium! 1105650611 M * lithium Thanks Bertl...it's been a while. 1105650699 M * lithium Has anyone ran LTSP inside a vserver? 1105650726 A * Bertl doesn't even know what that acronym means ... 1105650765 M * lithium Linux Termainal Server Project 1105650768 M * lithium As in http://www.ltsp.org/ 1105650818 M * Bertl hmm, isn't that what X does (basically)? 1105650856 M * lithium Mostly yes...but LTSP sets up a nice server for network booting terminals. 1105650894 M * Bertl hmm, isn't that what pxe/netboot does? 1105650895 M * sannes and maybe they do some things that are nicer for slow links.. like vnc .. 1105650943 M * lithium Yeah, I was hoping to set up an LTSP server inside a vserver for a PXE client. 1105650944 M * ndim Well, the server would just take network packets, interpret them and send other packets out. Oh, and do some HD I/O. All of that works inside a vserver guest. 1105650952 M * Bertl okay, so they basically rearrange the different technologies to form something like 'knoppix' or so?! 1105650978 M * Bertl (just trying to figure out where the issues could be) 1105651051 M * lithium I'm wondering about hosting NFS, BOOTP and XDCMP (may have screwed up the acronym) inside a vserver on a host that say...already serves NFS. 1105651080 M * Bertl you could do that, but just with userspace nfs servers (which are rare) 1105651100 M * Bertl bootp requires raw network access to work ... 1105651117 M * Bertl xdcmp should work fine ... 1105651156 M * lithium Well, maybe I won't bother with BOOTP then...I'll just use my existing DHCP server. 1105651181 M * Bertl is probably easier ... 1105651194 M * lithium I guess I'm just wondering about limitations in network virtualization...in terms the daemons I can run. 1105651212 M * Bertl nothing is really virtualized (yet) 1105651213 M * lithium My current vservers don't do anything amazing...just http and VNC. 1105651295 Q * serving Ping timeout: 480 seconds 1105651310 M * lithium I noticed with 1.29 that I can't see the vserver's network interfaces anymore...not sure when that changed. 1105651317 Q * chrish01 Quit: Leaving 1105651323 M * lithium ...from that host I mean. 1105651337 M * Bertl hmm? 1105651370 M * lithium I used to be able to do an ifconfig and see all my vserver's interfaces on the host... 1105651396 M * Bertl you are using newer tools I guess? 1105651433 M * lithium Yep. 1105651447 M * Bertl so you have to use 'ip addr ls' to see them 1105651464 M * Bertl (unless you specify to use aliases in the config) 1105651482 M * lithium I see... 1105651501 M * lithium So it's a change in the userland tools and not the vserver kernel patch then? 1105651524 M * Bertl yep, and it's not really new stuff, as ifconfig is pretty old ;) 1105651612 M * lithium Interesting. I remember being able to see everything up to at least when support for Slackware's init scripts was added to the userland utils...which was a patch I wrote actually. 1105651617 M * CosmicRay this is weird.... 1105651626 M * CosmicRay # vserver serv1 start 1105651626 M * CosmicRay Starting the virtual server serv1 1105651626 M * CosmicRay Error: /proc must be mounted 1105651626 M * CosmicRay To mount /proc at boot you need an /etc/fstab line like: 1105651626 M * CosmicRay /proc /proc proc defaults 1105651626 M * CosmicRay In the meantime, mount /proc /proc -t proc 1105651628 M * CosmicRay Server serv1 is not running 1105651630 M * CosmicRay SIOCSIFBRDADDR: Cannot assign requested address 1105651632 M * CosmicRay SIOCSIFFLAGS: Cannot assign requested address 1105651634 M * CosmicRay ipv4root is now... 1105651636 M * CosmicRay then it seems to proceed normally 1105651653 M * CosmicRay the fstab for that vserver does have proc 1105651664 M * CosmicRay and it is mounted. 1105651677 M * Bertl did you see our FAQ yet? 1105651688 M * CosmicRay damn, that's a FAQ? 1105651688 M * Bertl http://linux-vserver.org/Linux-Vserver+FAQ 1105651689 M * CosmicRay sigh. 1105651695 M * CosmicRay sorry. 1105651704 A * CosmicRay still has solucorp.ca stuff bookmarked :-) 1105651721 M * CosmicRay newvserver failed to complete normally for this reason too, apparently. 1105651788 M * CosmicRay that seems to provide a hint for /proc but it doesn't address the SIO* stuff 1105651805 M * lithium Hey bertl, is it still necessary to chbind daemons? 1105651824 M * Bertl CosmicRay: the SIO issue is probably a misconfiguration on your side 1105651855 M * Bertl I'd say you configured to create an ip entry which already exists on the host 1105651873 M * CosmicRay indeed I did... I thought you said that would work :-) 1105651880 M * CosmicRay so is this perhaps just a harmless message? 1105651899 M * Bertl yes and no, what you actually want is to change your config ;) 1105651929 M * Bertl do you use legacy (i.e. one config file) config? 1105651943 M * CosmicRay one config file, yes -- the one debian's newvserver generated for me 1105651951 M * CosmicRay it has IPROOT and IPROOTDEV specified in there 1105651953 M * Bertl lithium: you mean on the host? if so, then yes if you want to avoid clashes ... 1105651969 M * CosmicRay incidentally, loopback is also not working 1105651974 M * Bertl CosmicRay: could you upload it somewhere (e.g. pastebin.com)? 1105651978 M * CosmicRay I noticed that it didn't define IPROOTMASK or IPROOTBCAST... 1105651987 M * CosmicRay bertl: it's short enough I could msg it to you if that's ok 1105651989 M * CosmicRay about 10 lines 1105651997 M * Bertl yes, should be fine ... 1105652019 M * CosmicRay ok, you should have it. 1105652030 M * CosmicRay I probably don't really need CAP_NET_RAW, I'll take it out once things are up. 1105652033 M * Bertl typical debian config *sigh* 1105652059 M * Bertl ULIMIT="-H -n 1024" -> ULIMIT="-HS -n 1024" 1105652071 A * CosmicRay fixes 1105652078 M * Bertl S_CAPS="CAP_NET_RAW" -> S_CAPS="" 1105652093 M * lithium Ack...chbinding is an annoying task in Gentoo. Sometimes I regret switching my host over from Slackware...though I can still run Slackware vservers, which is QUITE handy. 1105652098 M * Bertl S_FLAGS="lock nproc" -> S_FLAGS="nproc" 1105652115 M * CosmicRay why do I not want lock? 1105652117 M * Bertl CosmicRay: and what you actually want: 1105652122 M * Bertl IPROOTDEV="eth0" -> IPROOTDEV="" 1105652149 M * Bertl lithium: why is that so? 1105652196 M * Bertl (I mean the 'annoying task' part) 1105652254 M * lithium Well, Slackware has very clean and simple rc.d scripts...so it's easy enough to add 'chbind' in there, but Gentoo's daemon scripts are not meant to be edited by hand...so they're a bit more work to figure out and modify. 1105652347 M * lithium Just a question from something CosmicRay mentioning...in my vserver conf files I specify IPROOTDEV to be eth0...should I be leaving that empty instead? 1105652376 M * Bertl well, if you 'reuse' ip addresses from the host, then definitely yes! ;) 1105652414 M * lithium My vservers all get their own IP addresses... 1105652416 M * Bertl http://vserver.13thfloor.at/Stuff/VServer-IP-Setup-0.1.txt <-- for details 1105652531 M * CosmicRay Bertl: why did you suggest I remove lock from S_FLAGS? from what I understand, that seems pretty essential 1105652560 M * lithium Is it possible to set an IPROOT and then not set an IPROOTDEV? 1105652803 M * Seraph re 1105652809 M * Seraph Bertl: alright, any new thoughts? 1105652876 M * Bertl nope, not really ... 1105652886 M * Seraph ok, so how do we proceed from here? 1105652887 M * Seraph ;) 1105652945 Q * nayco Quit: Bonne nuit ! 1105653100 M * Bertl CosmicRay: what do you expect it to do? 1105653112 M * CosmicRay # lock: Prevent the vserver from setting new security context 1105653118 M * Bertl Seraph: we have to do some more debugging ... 1105653133 M * Bertl CosmicRay: can you do that without that flag? 1105653142 J * chrish01 ~chrish01@69.90.131.10 1105653152 M * Bertl wb chrish01! 1105653159 M * chrish01 thanks :) 1105653170 M * CosmicRay Bertl: I don't know; I guess I've always used it; it's been a default and that comment made me believe it was important. does util-vserver do that automatically these days? 1105653185 M * Seraph Bertl: alright, talk me through what you need =) 1105653240 M * Bertl CosmicRay: yes, it won't hurt, but it's not really used atm ... 1105653265 M * Bertl Seraph: we can do one more debug statement in the kernel, then we have to resolve to userspace debugging ... 1105653300 M * lithium Bah...apparently I can see my vserver's eth0 alias afterall. 1105653323 M * Seraph Bertl: sure, let's go for it.. 1105653425 Q * chrish01 Quit: 1105653810 M * Bertl Seraph: sec, looking for the best spot ... 1105653827 M * lithium Bertl: Have the recommended vserver conf settings changed much over the past 1.5 or years? 1105653837 M * lithium +so 1105653847 M * Bertl you mean for legacy configs? ;) 1105653914 M * lithium Well, I've always been using the util-vserver userland utils and the stable vserver patches...but I'm explicity stating things like S_FLAGS="lock fakeinit" 1105653929 J * chrish01 ~chrish01@69.90.131.10 1105653968 M * lithium explicitly* 1105654027 M * Bertl if it makes you happy (the lock flag ;) 1105654091 M * lithium Well, like CosmicRay I'm not sure what's implied or what needs to be set nowadays. 1105654145 M * Bertl the thing is, util-vserver has reasonable settings (the mainline version), older versions and/or vendor specific adaptations have 'other' defaults ;) 1105654242 M * lithium So, assuming that I'm running a vserver without any specific needs (i.e. it doesn't do traceroute, run BIND, etc.) then you think I can leave the S_FLAGS and S_CAPS empty? 1105654305 M * lithium The util-vserver package from Portage should be unmodified...at least to my knowledge. 1105654319 M * Bertl yes, that's the idea (although fakeinit is special here) 1105654330 M * lithium How so? 1105654349 M * Bertl it basically is a different way to start your vserver 1105654383 M * Bertl (in the lack of per context caps and/or startup options, this flag was introduced) 1105654436 M * lithium What are the init options now? I remember fakeinit being recommended to keep things sane. 1105654471 M * lithium For whatever reason I like the idea of a fake init process in each context...maybe it's just irrational now though. 1105654498 M * Bertl with 2.6/1.9.x you can start a new init process inside each vserver 1105654523 M * Bertl but the idea of 'faking' the pid 1 is still there ... 1105654548 M * lithium How about 2.4/1.29? 1105654574 M * Bertl for 2.4/1.2x the fakeinit is fine (if you want to see a pid=1) 1105654814 M * Tbery Bertl, my trobles growing.. 1105654814 M * Tbery some virtual make wrong.. 1105654814 M * Seraph Bertl: any idea where to put that new debugging? 1105654814 M * Tbery are you using 2.6.8 secure proc?? 1105654814 M * Tbery I use debian.. 1105654814 M * lithium Well, it's probably time for a review of my seemingly outdated conf files. I already found a type that made my S_CAPS irrelevant. 1105654814 M * Tbery with that.. 1105654814 M * Bertl Seraph: working on it ... 1105654814 M * lithium typo* 1105654814 M * Bertl Seraph: thing is, my network connection is more than flakey atm ... so it's hardly usable ... 1105654814 J * JonB ~NoSuchUse@kg67.kollegiegaarden.dk 1105654814 M * Bertl evening JonB! 1105654814 M * JonB hey Bertl 1105654821 M * Bertl Tbery: you mean proc security? 1105654929 M * JonB Bertl: how stable is the stable vserver branch 1105654951 M * JonB Bertl: would i dare to run it at my works most importent server? 1105655045 M * Tbery irtual Root device support (CONFIG_BLK_DEV_VROOT) [N/y/m/?] (NEW) 1105655059 M * Tbery Virtual Root device support (CONFIG_BLK_DEV_VROOT) [N/y/m/?] (NEW) 1105655061 M * Tbery Y 1105655063 M * Tbery ?? 1105655142 M * Bertl Seraph: let's add a printk to vc_get_iattr (kernel/vserver/inode.c line 78) 1105655196 M * Seraph ok, tell me 1105655219 M * Bertl something like printk("vc_get_iattr(%d)\n", id); 1105655547 T * * http://linux-vserver.org/ | latest stable 1.29, devel 1.3.9, 1.9.3, ng8.7 1105655547 T * Doener - 1105655702 M * Bertl okay, I don't know what's up here but my connection is unusable ... I'll be back later ... 1105655752 M * JonB Bertl: okay, bye 1105655753 M * Seraph Bertl: alright, all logfiles on the webserver updated 1105655791 M * JonB how stable is the vserver patch, would i dare to use it at my works most busy and importent server, or would i have to hard reset it rarely? 1105655806 M * CosmicRay JonB: I've found it quite stable in 2.4 1105655817 M * CosmicRay JonB: I haven't used 2.6 muich because at the time of deployment it was incompatible with cups 1105655819 N * Bertl Bertl_8| 1105655824 M * JonB CosmicRay: i dont like 2.6 1105655828 M * JonB CosmicRay: it is not stable 1105655834 M * JonB CosmicRay: so 2.4 sounds good 1105655835 M * CosmicRay JonB: we have been using 2.4 for a main file/print/firewall server here, where we have approximately 100 users hitting it daily 1105655851 M * JonB CosmicRay: what load does it have? 1105655862 M * CosmicRay JonB: no trouble to speak of once it's up. flags may need slight tweaking for things liks samba or dhcp, but those are mostly documented various places 1105655876 M * JonB for samba? 1105655876 M * CosmicRay JonB: you want to know my load average? 1105655893 M * CosmicRay oh we serve up about 40GB of files to about 100 accounts with samba 1105655897 M * CosmicRay the machine is our pdc 1105655903 M * JonB CosmicRay: yes, that gives a hint of how battle tested your server is 1105655913 M * JonB CosmicRay: what was the trouble with samba? 1105655919 M * CosmicRay oplocks, iirc 1105655923 M * JonB CosmicRay: i dont run a dhcp server inside it 1105655927 M * JonB what is oplocks 1105655961 M * CosmicRay see the oplocks option in smb.conf(5) 1105656027 M * CosmicRay the machine itself is a dual Xeon 3GHz with 200GB of storage across a RAID-5 spanning 3 (for now) drives 1105656032 M * CosmicRay an HP somethingorother server :-) 1105656041 M * JonB okay 1105656047 M * JonB we mostly use it for CVS 1105656077 M * CosmicRay we have 14 vservers running on this thing 1105656083 M * JonB nice 1105656086 M * CosmicRay yup 1105656089 M * JonB okay, i'll dare to deploy it 1105656091 M * JonB thanks 1105656098 M * CosmicRay I haven't touched the master server since it was first isntalled 1105656111 M * CosmicRay which would have been June 21 :-) 1105656127 M * CosmicRay yes, I've been pleased with it. 1105656133 M * CosmicRay we ran the same setup on a Dell 2650 previously 1105656148 M * CosmicRay had some trouble, but tracked that down to a buggy tg3 driver, nothing to do with vserver 1105656201 M * CosmicRay I would have no qualms about using it in all but the most demanding production environments 1105656207 Q * chrish01 Quit: Leaving 1105656222 M * CosmicRay even there I wouldn't necessarily run away 1105656248 M * CosmicRay (but in those cases you are probably having less of a need for it anyway) 1105656362 M * JonB in what cases ? 1105656372 M * CosmicRay maybe you run ebay.com or something 1105656385 M * JonB no no 1105656391 M * CosmicRay in those cases you are more likely to use the "other" vserver project anyway (where you have multiple machines doing the same thing in a fault-tolerant situation) 1105656398 M * JonB just too many users and processes for too little a machine 1105656402 M * CosmicRay (so things keep working even if a box crashes) 1105656434 M * Tbery how use vproc?? 1105656475 M * CosmicRay Tbery: www.linuxvirtualserver.org 1105656484 M * CosmicRay it's essentially the opposite of vserver 1105656498 M * CosmicRay JonB: a cvs server shoudl be no problem with vserver 1105656505 M * CosmicRay JonB: I think you'll be quite happy. 1105656561 M * JonB CosmicRay: well, we also have this self coded java cms system 1105656587 M * CosmicRay JonB: sounds scary :-) 1105656591 M * JonB CosmicRay: it is 1105656597 M * JonB CosmicRay: i dont run it 1105656602 M * CosmicRay JonB: though probably less scary than jnode.sourceforge.net :-) 1105656613 M * JonB i pushed that to the developpers 1105656628 M * CosmicRay really? you encouraged them to run jnode? 1105656648 M * JonB CosmicRay: no no, i mean the running of their own app 1105656654 M * CosmicRay ah 1105656668 M * JonB CosmicRay: what is jnode anyway? 1105656677 M * CosmicRay JonB: an operating system written mostly in java 1105656688 M * JonB oh 1105656732 M * lithium 1105656864 J * chrish01 ~chrish01@69.90.131.10 1105657071 M * lithium Just out of interest...what distros do you guys run inside and outside of your vservers? 1105657125 M * CosmicRay lithium: debian and debian. 1105657136 M * CosmicRay tried suse inside one very briefly. 1105657138 M * CosmicRay painful. stopped. 1105657143 M * JonB debian 1105657164 M * lithium Lots of Debian I see. I run Slackware and Gentoo inside and usually Gentoo outside. 1105657171 M * JonB CosmicRay: oh no, i thought of doing suse because then the setup can mimic the production server hosted at an ISP 1105657185 M * CosmicRay JonB: probably would be easier for someone more accustomed to suse 1105657193 M * JonB CosmicRay: i dont know it 1105657193 M * CosmicRay it really didn't like being able to bring up hardware devices itself 1105657203 M * CosmicRay I did get it running OK 1105657212 M * CosmicRay anyway, time to run... 1105657215 Q * CosmicRay Quit: Client exiting 1105657278 M * lithium I've found Gentoo to be difficult at times inside because so many services need to be turned off to keep it from complaining. Slackware is trivial...just comment out a few lines in rc.M and you're ready to go. 1105657292 M * chrish01 Bertl_8|, you still here? 1105657315 M * JonB lithium: i use deb_newserver.sh 1105657335 M * lithium JonB: I have no idea what that is. 1105657338 M * chrish01 lithium, i feel all distros have issues with rc =/ ... thats why im very excited to see apples LaunchD system in Tiger 1105657376 M * JonB lithium: it's Paul Sladens? script for making a new debian vserver 1105657409 A * sladen looks around 1105657410 M * lithium JonB: Nice. I still make all my vservers manually... 1105657488 M * lithium In the past I've used the vskel (I think it's called) scripts as a hint in terms of what to copy and what to leave out...especially in terms of /dev. 1105657521 M * JonB lithium: http://www.paul.sladen.org/vserver/faq/ 1105657531 M * JonB lithium: Q: Where can I get decent Debian/Woody stuff 1105657572 M * chrish01 mirrors.kernel.org 1105657577 M * lithium I think that question should be directed at everyone...I know nothing of Debian. 1105657580 M * chrish01 they have *fast* debian mirror 1105657674 M * lithium JonB: Oh...that was from the site. 1105657702 M * lithium Hmmm...references to the old vserver site. 1105658502 J * serving ~serving@213.186.182.244 1105658615 M * JonB lithium: yes it was from the site 1105658617 M * JonB zzz 1105659109 Q * JonB Ping timeout: 480 seconds 1105660186 P * lithium Leaving