=== nealmcb changed the topic of #ubuntu-server to: Ubuntu Server discussion and support || For general (not server specific) support visit #ubuntu || Seriously good guide to asking questions on IRC: http://www.sabi.co.uk/Notes/linuxHelpAsk.html || Be patient - the right person to answer your question may not be available now, but it is best to just ask your specific question including relevant information || server guide: https://help.ubuntu.com/7.10/server/C/ || https://wiki.ubuntu.com/Server === nealmcb changed the topic of #ubuntu-server to: Ubuntu Server discussion and support || For general (not server specific) support visit #ubuntu || Seriously good guide to asking questions on IRC: http://www.sabi.co.uk/Notes/linuxHelpAsk.html || Be patient - the right person to answer your question may not be available now, but it's best to just ask your specific question including relevant information || server guide: https://help.ubuntu.com/7.10/server/C/ || https://wiki.ubuntu.com/ServerT === nealmcb changed the topic of #ubuntu-server to: Ubuntu Server discussion and support || For general (not server specific) support visit #ubuntu || Seriously good guide to asking questions on IRC: http://www.sabi.co.uk/Notes/linuxHelpAsk.html || Be patient. It's best to just ask your specific question including relevant information || server guide: https://help.ubuntu.com/7.10/server/C/ || https://wiki.ubuntu.com/ServerTeam [02:34] Hey, im in the process of installing Ubuntu server 7.10, it didnt find a kernel matching my CPU, its a VIA x86 so im installing it without a kernel, how can i install a kernel when its install or under the install process? [02:35] ... without a kernel? [02:35] ... [02:35] the box was running debian etch so a default kernel should work [02:35] ... [02:35] Kamping_Kaiser, yea, if the install program does not find a kernel matching something it does not install a kernel [02:36] oooookay. [02:36] You need the kernel to boot. [02:36] i know [02:37] so i was thinking if i could enter a shell and install it afterwards [02:37] from the cd [02:37] What did you do to convince it to let you install without a kernel? [02:37] Fujitsu, just selected yes [02:38] i also got a CPU vendor unkown error when the cd bootet [02:39] but it is a VIA CPU in a Vortex86 mini box [02:45] it will be funny to see what my boot options are going to be :D [02:52] hmm it seems like neither LILO or GRUB will install without a kernel [02:52] funny that [02:53] [02:53] how can i install a package in the cd shell? [02:55] * nealmcb watches intently, hoping to see a rabbit pulled out of a hat [02:56] * somerville32 pats bardyr on the head... with a shovel. [02:56] somerville32, hmm chroot is enough [03:00] i love linux shells [03:02] victory :D [03:02] cd -> shell -> chroot -> bash -> apt-get install linux-image-generic [03:03] its probaly the wrong kernel for my single core CPU but it is a kernel [03:06] when i did chroot /target how can i chroot back to / ? [03:07] ^D [03:07] aka, exit the chroot [03:07] thanks :) [03:08] np [03:18] w00t, got a kernel and grub loaded :D [03:23] booted? [03:24] yep [03:24] doing a boot now, its a slow 200mhz box [03:25] hmm what will be the best kernel for this CPU http://pastebin.com/m7b26e609 [03:41] whats the difference between linux-image-generic and -368 ? [03:41] They used to have optimized kernels [03:41] but no longer === joerlend_ is now known as XiXaQ [08:52] If apache was configured to use ssl to serve a cgi script why would firefox pop up a download window for that cgi script rather than run it? [08:52] is apache configured handle cgi scripts? [08:52] whatever the scripts extention is specifically too? [08:52] Oh, is that another module that should be enabled? [08:53] I've just followed 5 different howtos and hope I'm almost to the end. :) [08:54] yes, the scriptalias line is in the virtual site file that points to /usr/lib/cgi-bin [08:56] you'll need to make sure mod-cgi is loaded. i'm not much use with cgi apart from rough suggestions ;\ [08:56] i'd advise #apache for better help [08:58] Alright, thanks for the point in the right direction. [08:59] good luck [08:59] I'm looking to see if I'm supposed to have a certain login.html file or something else in /var/www [08:59] what are you trying to setup? [09:00] chillispot [09:00] Everybody's howtos are different and I'm trying to piece together what exactly I need to do. [09:01] the authentication system? *twitch* [09:01] dd-wrt+freeradius+chillispot [09:01] +mysql [09:01] *more twitch* [09:01] haha [09:02] I want a more secure wireless signal without the lower throughput because of wpa. [09:02] * Kamping_Kaiser lives in a faraday cage [09:03] not as such, but the walls work like that :) [09:03] that works too [09:03] painted the walls with that paint that blocks wireless? [09:03] lots of corregated iron. [09:04] not sure my walls are structurely sound enough for that paint - 1kg m2 or something :| [09:05] well also by logging to chillispot that won't be stored in somebody's keyring so that computer can't get back on the wireless unless the right user is on it. [09:08] anyway, this is taking much longer than expected. I better get back to the battle. [09:09] gl === paul____ is now known as pschulz01 [11:43] you know what we should implement by default? [11:44] a fail-safe grub :) [12:07] cluefull admins? ;-) [12:07] :) [13:25] Hey, how can i get the server to connect to a wireless network at boot? [13:55] heh [13:56] now, where are all those people claiming that there's not place for wireless tools on servers :) [13:57] hans: you have to edit and set up /etc/network/interfaces [13:57] ivoks: yea just found out about "wireless-mode", "wireless-network", etc [13:58] and it should work now, rebooting :) [13:58] there are examples in /usr/share/doc/ifupdown/examples/network-interfaces.gz [15:24] hi [15:25] whats that timestamp in the syslog? [15:25] is it a timestamp? [15:25] oh, sorry! I meant dmesg not syslog [15:26] [42959878.420000] Foo... [19:32] how was it you could install a task from the command prompt again? [19:32] I'm pretty sure there is something to prefix tasks with to make them aptitude installable [19:37] tasks? [19:47] Nafallo: apt-get install taskname^ [19:48] soren: aye. figure it out after a while. thought the $CHAR was supposed to be in the beginning first :-) [19:48] Nafallo: Nope :) [19:48] soren: yes, I did. [19:48] soren: :-P [19:49] Nafallo: How did you find out? It's not in the apt-get man page. Odd. [19:49] soren: tasksel uses apt-get :-) [19:50] Nafallo: Indeed it does :) [19:50] :-P [19:50] should be in the man though :-P [19:50] Quite. [19:50] or at least the changelog === dantalizing is now known as dantalizing_away [20:29] Hmm, what could: [140018.890000] Failure registering capabilities with primary security module. [20:30] Refer too? [20:30] Its from dmesg === dantalizing_away is now known as dantalizing [21:41] nealmcb: how's it going? [21:41] nealmcb: there was some discussion at UDS and afterwords on the doc ML about moving the Server Docs into their own area/package [21:42] nealmcb: I'm not sure the end result of the discussion, but I believe the Doc Team wanted some solid reasons to make the move. [21:43] nealmcb: the way things are configured now, the server docs build for each *buntu flavor without having to add extra commands [22:54] sommer: thanks. I've gotten the bzr branch out and have sketched out an idea for mail.xml. but I still don't know which version the main google result ( http://doc.ubuntu.com/ubuntu/serverguide/C/index.html ) actually is.... [22:57] nealmcb: that's actually a very dated version [22:57] yeah - seemed like it.... [22:57] very dated, and undated.... [22:58] the "server guide" has been deprecated in favor of integrating the server docs under the Advanced topics section [22:58] where do i find the buglist for 7.10 server ? [22:59] hatter: do you mean the 7.10 docs? [23:00] nealmcb: I also have another patch adding a "Mail Filtering" section to mail.xml... currently being reviewed by lamont and ScottK [23:00] sommer, i found one naughty bug in 7.10 and have another networking problem with two 7.10 servers that i cant track down, i want to see if others have the problem [23:01] it appears i made the mistake of deploying 7.10 too early... [23:01] heh. [23:01] sommer: my main goal right now is to figure out what I need to do to get my postfix-using-a-smarthost config to do smtp-auth to the smarthost. and the mail.xml section doesn't distinguish between postfix working as a server, and postfix working as a client [23:01] We're still in 7.04 as core here. [23:02] hatter: https://bugs.launchpad.net/ubuntu/ [23:02] hatter: What's the "naugty bug"? [23:02] naughty, even. [23:04] nealmcb: sounds like a great addition, another idea would be to create a wiki article first (I don't think there is one covering that) [23:04] nealmcb: then translate the aritcle into DocBook... at least that's the approach I've started doing [23:04] soren, if a nic is changed, udev remembers the old nic and keeps the driver for it, which then tells you the new nic cant be used, until you go and delete the nics address from 70-persistent-net-rules [23:05] with a wiki article you can also include instructions for past releases as well [23:06] the other bug is worse, though i havent found where to point the finger yet, after a short period of time the network lags for up to a minute then comes good again [23:06] hatter: Where does it tell you this? [23:06] hatter: I thought that was an /etc/iftab issue, not udev [23:06] soren, when you put a different brand nic in the same pci slot, [23:06] fujin_: We ditched iftab back in feisty. Quit living in the past :) [23:07] hatter: Not when. Where? [23:07] I'm living in Feisty, because it works. [23:07] xD ;) [23:07] fujin_, i am thinking udev because its a problem with the 70-persistent-net rules [23:07] cool [23:07] Haven't completed my test-phase of Gutsy. Probably won't roll it out till at least next year. [23:07] hatter: I still haven't seen what the problem is. Could you add a bit of detail? [23:08] soren, after you replace the nic, then try to install the driver module you get the error as if installing the wrong driver [23:08] 5 or 6 lines that say the nic cant be found or module cant be loaded something similar (from memory) [23:09] but then after you manually delete the old nics details from 70-net-persistent-rules it works [23:09] lamont: any chance the postfix dpkg-reconfigure could step up to the task of letting folks configure postfix with a smarthost that requires a password? are other packages involved? how might that work? [23:09] hatter: I really need more info than that to have even the slightest chance to look into it :) [23:10] I'm installing RoR on Ubuntu 7 Server, on "sudo gem install rails -y" an error is returned "SSL is not install on this system", why is SSL a requirement? [23:11] soren, it is easy to duplicate, just change nics [23:11] hatter: No. [23:11] soren, what more info could i provide you > [23:11] ? [23:11] hatter: You could show me the actual error message you get. [23:11] hatter: I've changed nics several dozen times. it works ifne. [23:12] ok, i have the box here, i will do it and let you know shorty [23:12] hatter: Please don't think that we accidentally put that stuff into persistent-net-rules. [23:12] hatter: It's quite on purpose. [23:13] nealmcb: are you looking for configuration steps, or do you already have them and are looking to document them? [23:13] soren, i understand, i know its purpose, but currently its not allowing a nic change [23:13] kshah: That's really a question for the ruby developers. [23:14] kshah: You're using gem, which fetches all sorts of stuff that's not from Ubuntu. [23:14] k [23:14] hatter: Yes, it does. [23:14] hatter: I have done this many times. [23:15] sommer: the former. I've seen other web pages about it, but am wondering how they relate to ubuntu, and the server guide left me wondering if it was addressing my use-case (unlikely) or not [23:16] nealmcb: gotcha, ya it probably doesn't... the current thinking about the guide is to document the most common configuration [23:17] nealmcb: if you think it needs to be documented I'm up for it [23:17] or if you have a patch just submit it to the doc list, and I'll review it [23:18] sommer: I thought a smarthost config was pretty common, but I guess in most cases they don't require any auth. but on my laptop I want to use fastmail as my smarthost, rather than my cable company, and fastmail (of course) requires auth for relay [23:19] sommer: great. I'll try to get the facts and steps straight first - thus my query to lamont :-) [23:19] mmmmm... ya a wiki article may be more appropriate, but it would probably be a good idea include something about smarthosts in the main docs [23:19] soren, i have replicated the problem [23:19] nealmcb: cool, just let me know if I can help [23:20] but just for fun I'll send you my little bundle with all its warts and incompletions now [23:20] (my first doc patch) :-) [23:20] nealmcb: party! [23:20] :-) [23:20] hatter: I need all the info I can get. syslog, dmesg, pci ids of nics.. [23:21] hatter: ..which driver they're using.. Everything, really. [23:21] scenario : one intel nic. 70-persistent-net-rules shows PCI device (e100). intel card removed. realtek card installed. 70-persistent-net-rules shows PCI device (e100) with ETH0, and new detected 9139too) as eth1 [23:22] '8139' [23:22] make sense ? [23:22] Yes. What's the problem? [23:22] there is now only one nic. the system detects it and allocates it as eth1 [23:22] Er... That's on purpose. [23:23] Once a name as been assigned to a nic, it keeps it. [23:23] Trust me.. the alternative is worse. [23:23] so if a nic is removed, the system decides the removed nic still belongs to eth0 ? [23:23] Yes. [23:23] seriousy ? [23:23] Yes. [23:23] This is a good thing. [23:24] how is this a good thing ? [23:24] How is it bad. [23:24] ? [23:24] It's good because your interface names don't change all the time. [23:24] because if you change nics you have to go into 70-persistent-rules and remove the old one. [23:24] Why? [23:24] and if you dont know about this, you waste a few hours trying to find out [23:25] No. [23:25] You waste time, because my make false and useless assumptions. [23:25] Say you've got 4 nics. [23:25] you remove eth0. [23:25] What should happen? [23:25] reference to the old nic should be removed, allowing new nic to get eth0 [23:26] Why? [23:26] if old nic is not there anymore, why should it want eth0 ? [23:26] Why should the new one want eth0? [23:26] shorewall for instance references nics by eth(?) [23:27] Yes. [23:27] Look.. [23:27] NIC's can be connected to a variety of different busses. [23:27] i must be missing something. if i remove a nic from a machine, you are saying i should then want my new nic to be eth1 ? [23:27] Several of them support hotplugging. [23:28] At which point should we allow a given nic name to be reused? [23:28] What if 5 seconds after that timeout it detects my old eth0? [23:28] of course. if a nic is removed, why should there be reference to a something that isnt there ? [23:28] hatter: Why not? [23:28] hatter: It might show up in a few seconds? [23:29] hatter: And if not, why does it matter that your only nic is called eth1? [23:29] because of shorewall for one reason [23:29] Heck, I know some people who would probably prefer them to be 1-indexed, but let's not get into that. [23:29] hatter: So because shorewall is broken, we should break it for everyone else? [23:29] hatter: Or should we fix shorewall? [23:29] hmm.... [23:30] so its not reasonable to say the only nic in a system gets eth0 ? [23:30] i am still missing something. why would you want something not in your machine to still be referenced ? [23:30] If you install a new system with just one nic, it'll get eth0. If you plug in a new one, it gets eth1. If you remove the first one, the second one keeps eth1. [23:31] if you plug in a third one, it gets eth2. What's the logic in the third added nic gets eth0? [23:31] Sorry to break into this but I recently had to mess with changing a NIC. If you edit that 70-persistent file and make the new nic manually the same device as the old one won't shorewall just work? [23:31] sure, and if you remove the first, the second one gets eth1 also [23:31] hatter: Because it might show up again! [23:31] Centaur5, yes, this is what i am saying [23:31] hatter: I didn't test the server before I left but I just assumed it would just work. I didn't hear from them so I guess it did. :) [23:31] hatter: If shorewall can't deal with not having an eth0, shorewall is broken. [23:32] That's no reason for us to break everything else. [23:32] Centaur5, i am not seeing why the default behaviour is to alocate the next eth() instead of removing reference to a device that isnt there [23:32] hatter: You're not listening. [23:32] hatter: Answer this: [23:33] hatter: When should we free eth0? [23:33] eth0 should be freed if there are no nics. [23:33] And how do you determine that? [23:34] or if a nics mac address is changed [23:34] Well windows does the same thing. If you add a nic then it creates local area connection 1 or higher and if the old one was removed you just don't see regular local area connection anymore. [23:34] hatter: NO!!!! [23:35] hatter: How do you tell if it's the same nic if it's changed its mac? [23:35] hatter: Hwo do you determine if a nic has been removed? [23:35] but when a new nic is added its detected [23:35] hatter: You might think it's a stupid question, but please answer anyway. [23:36] probing the pci location i expect [23:36] What if they just haven't been detected yet? [23:36] No. [23:36] You can move nics around in your machine. [23:36] There are also other buses than pci. [23:36] It would have to store information about the card and check it against every slot? [23:36] hatter: Seriously: If you moved two nics around, would you expect them to switch names? [23:37] soren, i expect the first nic is eth0, or if two are put in, the first device detected is eth0 [23:37] hatter: That's the way it is. [23:37] then if one is removed, the detection happens again, and the first one gets eth0 [23:38] hatter: What if the first one is still there, buthasn't been detected yet? [23:38] isnt the detection happeing anyway ? [23:38] hatter: Have you heard of usb nics? [23:38] when is nic not detected ? [23:38] hatter: Have you heard of hotpluggable pci? [23:38] hatter: Stuff takes time. [23:38] soren, no, i havent heard of it [23:38] hatter: Which one? [23:38] hotpluggable [23:38] pci [23:38] hatter: hotpluggable pci? It's quite common in mid- to high end servers. [23:39] hatter: surely you're familiar with usb? [23:39] yes [23:39] is usb referred to as hotpluggable pci ? [23:39] hatter: So... don't you find it reassuring that a usb nic keeps its name between plugins? [23:40] hatter: No. USB supports hot plugging. So does PCI if you've got the hardware to support it. [23:40] ugh [23:40] that's cringeworthy [23:40] What is? [23:40] hotplug pci, I don't think I'm that brave. [23:40] even on servers that support it I don't do it. [23:40] fujin_: I'm not talking about yanking crap out of your home pc. I'm talking about hardware that was meant to do this. [23:40] fujin_: Bah. [23:40] I'm not a retard, this is #ubuntu-server [23:41] You wouldn't catch me dead running Ubuntu on my home PC, either. [23:41] fwiw [23:41] meh [23:41] :S [23:41] hatter: So... When should the name be recycled? [23:41] soren, ok. so the usb example has me thinking. [23:42] hatter: We could for instance wait a minute from when we've started detecting for nics.. [23:42] hatter: If it hasn't show up yet, recycle the name. [23:42] hatter: ...but that would mean that the newly added nic can't get a name until after a minute has passed. [23:42] isnt there some uniquie identifying sequence ? [23:42] hatter: No. [23:43] hatter: Well, there could be, but we really don't want to do that. [23:43] Besides: [23:43] There's *no* point. [23:43] Who gains anything at all from recycling interface names? [23:43] but if a nic is detected, then the it is identified for the correct module to be put in [23:43] hatter: Huh? [23:44] when a nic is installed in a new system, it is detected and the correct module loaded to suit it. (hopefully) [23:44] fujin_: btw.. #ubuntu-server has seen its share of retards. :) check the logs. :) [23:45] hatter: Yes. [23:45] hatter: And? [23:46] ergh [23:46] sorry :) [23:46] fujin_: If only calling it #ubuntu-server would scare them away.. /me looks to the skies [23:46] :) [23:47] so if a nic in inserted and its not the same one, and its the only one detected, why not remove reference to the old one [23:47] If only. [23:47] * somerville32 pokes soren with a stick. [23:47] hatter: I know people that have universal docking stations for laptops that link USB ethernet, audio, and such through one port. I could see that being a problem if the eth name kept changing everytime they went home. [23:48] Centaur5, and each time it is plugged in to the same device, in which case the reference should stay there [23:48] hatter: There could be many reasons: a) it might be here, just not detected yet (again: how long should we wait, and what should we do about the new nic until then?) or b) it might show up the next time (docking stations are a good example). [23:48] hatter: You're trying to fix a problem that doesn't exist. [23:48] hatter: There is *no* reason why we should recycle those names. [23:49] soren, Centaur5, ok, I will give these things some good thought. [23:49] hatter: Actually one of the most popular problems people complain about is when they keep taking their laptop home and plugging their printer in and all the sudden they can't print cause the default goes to usb1 with they plugged it into usb2 and it created a copy printer. [23:49] hatter: I honestly wouldn't bother. [23:49] hatter: Unless. [23:49] hatter: If you can give me just *one* use case where it would be useful to have those names recycled. [23:50] hatter: If not, you're trying to solve a problem that does not exist and that's really a waste of time. [23:50] (by definition) [23:50] I haven't tried it but has Gutsy fixed the problem of plugging printers into different ports and the jobs not being sent to the right one? [23:50] :) [23:50] Centaur5: Depends on the printer, I think. [23:51] Centaur5: For some, the connection is defined using the serial number of the printer in which case everything works fine. [23:51] I'll have to test that out cause that confuses the crap out of people. :) [23:51] Centaur5: In other situations this sort of things break *for* *exactly* the same reasons as interface naming. [23:51] soren, no point wasting time. this is the first time i have experienced this with a distro and it took me little while to realise why the nic wasnt working. [23:52] hatter: IMO *everyone* ought to be doing it our way. :) [23:52] hatter: The world is a better place this way. [23:52] soren, so this a ubuntu specific concept ? [23:52] hatter: Not sure. [23:52] hatter: I don't think so, though. [23:52] i dont think it does it on etch [23:52] but i would have to check to be sure [23:52] unlikely, the machine I had to do it with a month ago was using Debian Etch. [23:53] We get it from udev upstream. [23:53] I don't know if Debian use it, too. [23:53] Since I don't know anything about shorewall I had to quickly google how to make the new NIC use the old eth3 name. [23:53] apt-get remove shorewall && profit [23:54] We had a similar problem with mrtg. [23:55] Er.. no. [23:55] apt-get remove mrtg && apt-get install cricket [23:55] &&profit && done [23:55] That was differnt. I'm confused. [23:55] ugh [23:55] mrtg makes me cry [23:55] i still have another issue that is plaguing me that is hard to communicate. on 2x 7.10 servers with different hardware, the samba share is slowing down to a crawl every little while for a minute or so. what these 2x boxes have in common is : 7.10, vmware-server, hardware is completely different [23:55] i like shorewall, why dont you like it ? [23:56] Because I know how to iptables. [23:56] fujin_, good reason. i am lazy, shorewall makes iptables easy [23:57] Wasn't Gutsy supposed to include winmodem support out of the box? [23:58] Centaur5: It does. [23:58] soren: Only on certain chipsets? [23:58] Centaur5: Well... Sure. [23:58] Centaur5: Quite a few, though. [23:58] Centaur5: You need to enable it through the restricted manager thing. [23:58] It's not entirely free. [23:58] soren: I've tried conexant, agere, and smartlink and hylafax doesn't seem to find them. [23:59] We use slmoded [23:59] slmodem, I mean. [23:59] I suppose that's smartlink?