[00:00] oettinger: shitty beer's fault :D [00:00] oettinger: well http://packages.ubuntu.com/search?keywords=php5 tells me that's the latest karmic version available... 5.3.2 is in lucid [00:01] Yes, i thought so. [00:01] change mirror [00:01] might solve the issue [00:01] it won't [00:01] Could be. I think the server uses the hosts internal mirrors [00:01] ls [00:01] it means he's running karmic [00:02] I know that hoster, and the mirror is in sync... [00:02] oops lol using to many keyboards at once [00:02] if he wants 5.3 he needs to upgrade again, to lucid [00:02] Do another release upgrade? [00:02] cat /etc/issue to verify [00:02] AGAIN [00:03] cat /etc/issue -> Ubuntu 9.10 [00:03] "Well there is your problem" [00:03] AGAIN [00:03] oettinger: make sure everything is fine with a reboot before upgrading (uptime is overrated compared to the trouble a faulty boot config will give you especially with a hoster where you don't have any out of bands management) [00:04] yeah, if you didn't reboot after the upgrade, you need to [00:04] * serverhorror reboots afert any and all updates [00:04] just to make sure... [00:04] you only need to reboot after kernel updates and release upgrades [00:04] Will do. Thank you for you help. Will (hopefully) post my success story a bit later [00:04] after updates too? [00:05] Just kernel updates [00:05] and if you use uptrack, that takes kernel updates out of the equation [00:05] crap [00:05] launchpad is readonly now.. doh [00:05] !language | SpamapS [00:05] SpamapS: Please watch your language and topic to help keep this channel family friendly. [00:06] lol [00:06] I meant carp! [00:06] * SpamapS loves fish is all [00:06] qman__: not exatly correct. rebooting makes sure all my init scripts are fine. everything is initialized properly. /tmp (and whatnot is cleaned out properly). There isn't some rogue pxe boot server around. The BIOS config is still fine and isn't set to PXE boot. ..... [00:07] I only reboot after upgrades [00:07] yeah [00:07] p1l0t: yes upgrades.... :) [00:07] you certainly can, but you don't need to [00:07] did I mention I'm quite conservative regarding config changes/up[(d)|(gr)]ades :) [00:08] kernel updates and distribution upgrades are where you need to [00:08] serverhorror: Rebooting when it's not needed is more risky than not rebooting. [00:08] no wait. conservative isn't the right word... paranoid (too strong)..... [00:09] my router is still running jaunty because I haven't had an opportunity for the downtime [00:09] ScottK: why? pls explain. I want to know on the earlist occasion if something failes. And unfortunately monitoring can't catch everything [00:09] Yeah some of the ancient machines I run ubuntu server on I am afriad to boot because they might turn to dust [00:09] uptrack certainly makes things easy [00:09] serverhorror: rebooting always stresses the system and so doing it when you don't need to has risks (they are small). [00:09] ScottK: of course there are exceptions (hot DB caches being an example) [00:10] There are also some updates that require services to be restarted, but if you subscribe to Ubuntu Security Notices (and you should), you'll know which those are. [00:11] ScottK: accepted. But if that stuff fails, I'd rather have it fail while I have a maintenance window scheduled anyway (and expect failure) rather than at 2am. Support Contracts will fix failing hardware, but won't give me any piece of mind during night :) [00:11] Certainly. [00:12] In that case it may make sense in your situation. [00:14] funny detail: we don't have any on-call contracts with the employer. So having things fixed (or broken) during maintenance windows is a lot better. Since stuff that is broken will stay broken and won't cause "unexpected" downtimes. (sounds harsh but I found customers react a lot better to 8h of downtime when they expect it than 5min of unexpected downtime) [00:15] Could be a month, if they expect it then its status_quo [00:16] ahh of course. The last run took me from 9.04 to 9.10 [00:17] AGAIN [00:18] :) yes yes yes. It is running. [00:19] When is the next one anyway Oct? [00:19] it's Oct and Apr or something right? [00:19] Never rebooting a system is a huge risk IMO. It makes one think twice about rebooting, and if you have a sudden unexpected power failure, you're stuck with systems that may not come back up in your worst crisis. [00:20] After hitting the "this FS hasn't been checked for X days, lets mark it dirty and fsck for 3 hours" time bomb a few times, I learned my lesson and started scheduling server restarts every 120 days. [00:21] p1l0t: next what? [00:22] Next release [00:24] Yeah it's Maverick Meerkat (10.10) October 10th [00:26] Is it better to stick with LTS then to upgrade to something with an earlier EOL? [00:27] depends on what you plan to do [00:27] The next LTS will be out around the same time maverick is EOL [00:28] Success! Thank you again. [00:28] Most server types would rather have a stable known platform and therefore run LTS. [00:29] That's what I'll do at work, maybe at home I'll be risky [00:29] Because my home server is just for messing around and learning [00:29] Have a nice timeofday(/you) [00:30] You too oettinger [00:31] yeah its a good idea to keep your "messing around box" up to date.. you'll be ready for any changes that come in the next upgrade from LTS -> LTS [00:33] SpamapS, that's what UPS and generators are for ;) [00:33] any idea how to have an init script run a daemon as a specific user? [00:33] My UPS is only good for twenty minutes :( [00:34] sudo apt-get install generator [00:34] haha [00:34] qman__: I refer you to livejournal's experiences with UPS and generators :) [00:35] where I'm at, power failures are expected [00:35] the electric company doesn't exactly have a good reputation [00:35] I expect every single piece of my systems to fail. [00:36] Plus UPS keeps the supply stable... This is especially important if you run a server at a machine shop where people power up whatever whenever with no warning [00:36] yeah === shade\ is now known as shade} [01:07] thank god NYC has reliable power === shade} is now known as shade\ [01:49] when I install from disk by hand there should be a seed file someplace correct? where is that file? [01:58] maek: I don't believe one is generated [01:59] giovani: ok, I wasnt sure. find / -name *seed* didnt turn up anything but the docs have a good example [01:59] maek: I just did the same thing to make sure there wasn't one :) [01:59] giovani: thanks [01:59] I want to automate network installs. is FAI or rolling my own the only option I have? [02:00] well d-i should handle the install [02:00] what features are you looking for? [02:01] well Im used to cobbler for redhat [02:01] it lets you add a system with some info like mac and network info, what profile to use etc etc then it generates a kickstart file from a template === shade\ is now known as toshade [02:01] d-i is what you use in preseeding right? debian-installer? or is that a seperate tool? [02:02] ah, I haven't used cobbler [02:02] we write our own pxe configs, and use kickstart at work for redhat [02:02] yeah, debian-installer is for preseeding [02:02] cobbler maybe over load. I only really have 1 kind of system to install so only 1 preseed file [02:02] giovani: you should check out cobbler for managing kickstarts its slick [02:03] it manages pxe and all the dhcp entries for hosts [02:03] maek: what does it buy me over plugging the machine in and sleecting which image I want in the pxe menu? [02:03] I mean, dhcp is already handled -- no need to reinvent the wheel [02:03] if you installs are that clean cut then not much, just some resuable ness in your kickstarts because they are templates [02:03] you can use snippets [02:04] ah [02:04] yeah, we just use kickstart to get the machine usable and on the network [02:04] then rdist out to it [02:04] I just liked that I only had to type in 1 place to add a machine to kickstart and it did everything else [02:04] no point in reworking the kickstart constantly [02:04] yeah true [02:05] yeah, so cobbler is server-side handling [02:05] so sorry to being dumb, is debian-installer a tool or just the d-i bit in the seed files [02:05] s/to/for [02:07] maek: you can use kickstart for ubuntu [02:07] debian-installer is the installation tool, like anaconda for redhat [02:07] giovani: do I want to do that? [02:08] giovani: thanks [02:08] maek: I don't know [02:08] I don't mass-deploy ubuntu, so I'm not up on which is superior [02:08] giovani: I mean is it legit, can I do what I could do with preseed? [02:08] I'm sure the wiki has some hints though [02:09] I think preseed is the best supported method [02:10] giovani: makes sense. [02:10] configs are simple === jjohansen is now known as jjohansen-afk [02:19] preseed also isn't so fugly [02:24] Hi, i have just installed ubuntu server and configured SAMBA, im about to copy some files over but the troughput is horrible. Its now 17KB/sec. it takes 15min to copy 12MB over :S and i have gigabit network cards in both machines. [02:25] anyone awake? [02:35] gundehest: i dont know the issue with your system, but FYI: we arent always awake and will answer you when we get to it :P [02:36] win 65 [02:38] There's no 6.5 release; they're only up to 6.1 [03:19] What's the current "best practice" for a small office (5 users) migrating to linux workstations & server (file & print)? [03:20] I'm guess CIFS for file sharing over NFS, but I'm curious about "disconnected operation." (laptops, etc.) [03:23] Either will do if it's homogeneously unix [03:26] Is there a generally accepted method of handling the disconnects & reconnects that come with laptop use? [03:27] I'm not necessarily looking for synchronization, just a way to easily mount/umount... [03:27] NFS gets really hairy when the server goes down, I assume frequent disconnections would be the same [03:28] CIFS would be the better option there, that or sshfs/sftp [03:28] that's what i was thinking too. [03:29] when my NFS server goes down unexpectedly, I have to `sudo umount -l` my directories [03:29] and nautilus usually crashes, requiring a reboot [03:30] I've been looking for something that can mount/umount some shares,maybe tie that into laptop suspend or something... [03:31] thinking about a workgroup or projects share, not home directories. [03:33] this isn't my home directory [03:33] nautilus just completely hangs, even if you bring the share back up [03:34] have to kill -9 it, and if I just launch it again, desktop icons are missing, and a few other annoying bits [03:34] until I reboot, or at least log off/restart X [03:36] in any case [03:36] sftp is quite a good solution there [03:37] if you use something like nautilus anyway [03:37] because you can mount your shares through the 'connect to' menu, then add them to the favorite places [03:37] and a simple click mounts it, clicking the eject button dismounts [03:37] I suppose CIFS would work the same way [03:38] I guess the point is, easiest depends a lot on client software [03:38] That's a interesting idea. Don't even try to set up a "smart" script. Be dumb an let the user attach when they need to. [03:39] the downside to that is, at least with nautilus, I don't really understand where/how it mounts them, so you can't just interact with scripts or the command line easily [03:39] sort of an all-CLI or all-GUI choice to make [03:40] Yup. I don't understand that either. I've gone hunting, and there's obviously some Gnome-VFS magic going on, but I can't track it down. [03:43] one thing when doing this, make sure each user has a unique UID/GID [03:44] if you just install all the systems, and add one user, they'll have the same UID [03:45] it won't make much difference with sftp and cifs, but if you use nfs it will [03:45] not sure about sshfs, probably not [03:45] nfs is intended to be used with networks that have directory services [03:46] yup. that's true. that's another nail in the NFS coffin. I'm trying to do this without building up a full infrastructure. [03:55] qman__: are you hard or soft-mounting your NFS? [03:56] I mean NFS and CIFS are horrible protocols [03:57] I'm not sure what the difference there is [03:57] it's in /etc/fstab and I'm using the kernel client [03:57] qman__: if you don't specify either, then it's hard-mounted [03:57] giovani: what would you use instead? [03:57] hard-mounting NFS hides the lower-level errors/disconnections from the upper layers [03:58] this is helpful if you expect frequent disconnects [03:58] but if you want the higher-level apps to fail properly, then soft-mounting is what you want [03:58] MarchHair: AFS [03:58] 192.168.1.8:/home/public /home/public nfs auto 0 0 [03:58] soft mounting assumes the high-level app will correctly deal with the error [03:59] twb: indeed [03:59] he's getting nautilis hangs with hard-mounts [03:59] Yeah, well, that's it working as advertised :-) [03:59] it doesn't happen very often, only when the server crashes [03:59] twb: heh [03:59] qman__: sure [03:59] but it's quite annoying when it does [03:59] use a better network fs [03:59] Coda, AFS, anything decent [04:00] giovani: a better network filesystem won't hang *nor* return an error when the backend store disappears? [04:00] Or are you just talking about having a distributivity/caching, so it happens less often? [04:00] twb: well, a better-engineered distributed network fs won't, no [04:00] AFS handles caching on the client properly [04:01] It's no good if the file isn't already cached [04:01] and with a distributed fs you're less likely to encounter the downtime in the first place [04:01] caching is only one component, like I said [04:02] Coda and AFS are both distributed [04:02] Distributed in the sense that the same data is stored on multiple hosts? [04:02] I don't remember reading about that when I looked into AFS [04:02] sure [04:03] it's configurable [04:03] Coda's better at that [04:03] It was more like /afs/foo.org is always stored on foo.org, and nowhere else. [04:03] Coda is optimal for high-latency high-downtime links [04:03] Coda and AFS are also harder to set up and less widely tested than NFS/CIFS, of course. [04:04] If this is just for a LAN, I'd recommend qman__ to instead work out why the server's crashing and fix that [04:04] and yet they're much better [04:04] servers crash, it's inevitable [04:04] it's not a regular thing [04:04] giovani: sure; you can either have the thing nobody likes, or the thing nobody uses :-) [04:04] had a NIC fail a couple months ago, etc [04:04] twb: you'd be demonstrating extreme ignorance [04:04] giovani: yes, but if the server crashes for five minutes each year, he probably doesn't give a shit [04:04] to claim that "nobody uses" AFS or Coda [04:05] twb: but he'd learn to use a better protocol [04:05] so when he wants to implement something at work [04:05] he knows something more than NFS [04:05] I don't have the budget for multiple servers at this point, so distributed wouldn't provide any advantages [04:05] but a more graceful handling when it crashes would be nice [04:05] one that doesn't require rebooting [04:05] qman__: they have advantages besides the ability to be distributed [04:05] qman__: why are you having to reboot? [04:06] because all of gnome gets messed up after nautilus crashes [04:06] giovani: it's just that you sound like a weenie [04:06] desktop icons missing, other weird bugs [04:06] qman__: so maybe just restart X? [04:06] effectively the same, I still have to close all my programs [04:06] twb: just someone who's used NFS, and knows it sucks, and encourages people to learn the better options out there rather than stick with an old standard [04:07] I still haven't rebooted since my server last crashed, about two weeks ago [04:07] qman__: technically if the NFS server is brought back up, your session should be resumably [04:07] resumable* [04:07] * MarchHair quietly ducks under the table as the fight he innocently started kicks into high gear... [04:07] it hard locked, never figured out exactly why [04:07] and twb, these symptoms he's describing are EXACTLY what caching prevents [04:07] ran zip on it and it just choked [04:08] Shrug. [04:08] hard locks from a network outage are a classic issue with NFS [04:08] we've probably lost countless dollars to them at work [04:09] just saying, when you use NFS in the real world, this stuff bites you in the ass [04:09] giovani, it usually takes a few hours to bring the server back up, since if it crashes, it's configured to fsck the RAID, and if it's a hardware failure, well [04:10] qman__: yikes -- do you know what the cause of the crashes has been? [04:10] qman__: buy a UPS [04:10] even if they're infrequent [04:10] the most recent one, no idea [04:10] I have a UPS [04:10] a UPS? [04:10] it'll last about an hour and a half [04:10] how does that prevent server crashes? [04:10] Oh, you mean a software crash [04:10] yes [04:10] What the fuck are you running on your NFS server that it EVER crashes? [04:10] well if it's his home network, presumably if he lost power, his desktop would also lose power [04:10] the one before that, the NIC failed and I had to shut it down and replace, forgot to umount my filesystem first [04:10] so he wouldn't care [04:11] giovani: not if it was a laptop :-) [04:11] presumably :) [04:11] I've got a UPS on that too [04:11] good point [04:11] my network stays up when the power goes out :) [04:11] NFS home directory on a laptop doesn't sound like a good idea [04:11] kind of defeats the purpose of it being a laptop :) [04:12] pretty much everything else just uses CIFS, it just wasn't adequate for this desktop [04:12] lots of lag trying to load music files [04:12] That shouldn't happen [04:12] what's the CIFS server? [04:12] samba? [04:12] yes [04:13] well there's your problem ;) [04:13] Yeah, is this NAS running Ubuntu LTS? [04:13] and windows clients have no issues using it [04:13] yeah, 10.04 now [04:13] samba is junk [04:13] sorry to say [04:13] had to upgrade from hardy because my replacement NIC didn't have drivers [04:14] qman__: I'm in that position; it's fucking me up because I have to port all this in-house juju to 10.04isms [04:14] it had been running great until that NIC failed [04:14] almost a year of uptime [04:14] but a NIC failure shouldn't have required an unclean shutdown and therefore a fsck [04:15] Nod. [04:15] no, but I had to order one [04:15] took a week to get here [04:15] what were the issues that caused unclean shutdowns? [04:15] I have lots of spare PCI NICs, but it didn't have any free PCI slots [04:15] haven't figured it out yet [04:15] I ran 'zip' on the server [04:15] and it just choked, out of the blue [04:15] hmm [04:15] disk failure? [04:16] filesystem failure? [04:16] what fs are you using? [04:16] disks are all good, filesystem checked clean [04:16] ext3 [04:16] I'm wearing of filesystem checks [04:16] weary* [04:16] giovani: switch to btrfs, then, where they're not supported [04:16] you still have the logs? [04:16] SMART is happy, mdstat is happy [04:17] there was nothing logged [04:17] twb: we're evaling btrfs actually [04:17] qman__: nothing? [04:17] no oom? [04:17] like it never crashed [04:17] no watchdog? [04:17] ouch [04:17] I don't know if there's anything special I can configure to help it [04:17] serial console :) [04:17] but it was so far gone there was no kernel panic messages on screen, sysrq commands did nothing [04:18] sounds hardware-related then [04:18] if the kernel doesn't panic but sysrq is no good, 90% chance it was hardware [04:18] not good news, but good to know anyway [04:19] only a few things could cause that [04:19] CPU/Motherboard/RAM [04:19] run some burn-in tests [04:19] it's been operating hiccup free since then [04:19] of any of them, I suspect the motherboard [04:19] who made it? [04:19] but it's all DDR equipment, old hat [04:20] if it failed I'd replace all three [04:20] foxconn [04:20] ah [04:20] yeah, cheap taiwan stuff [04:20] guess that one goes to the top of the 'to-buy' list [04:21] I personally prefer Intel [04:21] for my home systems that don't need something fancy [04:21] every Intel board I've used has been rock-solid [04:21] but they're never on the cutting-edge [04:21] I'm a bit of an AMD fanboy [04:22] I'll probably buy a gigabyte board, though [04:22] you must be hurtign then :) [04:22] since AMD's market share is dropping like a stone [04:23] it needs an upgrade anyway, it's a single core 2.2GHz, 4x512MB in it [04:23] sounds more than adaquate for a home server [04:23] That's faster than most of my machines put together [04:24] My home server is a 16MB/32MB 200MHz MIPS system. [04:24] heh [04:24] NFS hosted on that? ;) [04:25] hey guys [04:25] i have a web server at home [04:26] that i am trying to access. but i can't seem to find a way to access it through the same name both from within the network and from outside the network [04:26] i have a dynip domain that works when i'm not at home, but from home i never works [04:26] an internal DNS server is usually the solution [04:26] well, if you want to access it by any name from outside the network, you need to register a name [04:27] oh, I see [04:27] qman__: i already have a name [04:27] the problem is your router [04:27] by default, they will not route packets back into the network [04:28] qman__: but how come i use the machines actual name and i do access it from within the network [04:28] so, you'd need to either configure it to do that (bad idea security wise) or set up an internal DNS to answer that query as a local IP instead of your external IP [04:28] wouldn that count as routing packets back? [04:28] isn't there a redirect feature that some routers use? [04:28] no, because when you use the internal name, it gives you the internal IP [04:28] giovani: no, but HTTP and sshfs is [04:28] i see [04:28] LAN-side requests for the WAN/DMZ IP get icmp-redirected back to the LAN ip [04:28] when you use the internet name, it gives the external IP [04:28] how i have a setup a dns server [04:29] and your router is not routing traffic back to the server [04:29] The files themselves wouldn't fit into the 16MB of nonvolatile memory; it has a USB key hanging out the back [04:29] is a dns server with pretty much no traffic heavy to run? [04:29] not at all [04:29] memory and cpu-wise [04:30] Hilikus: do you mean a local DNS cache of the internet's A records, or do you mean hosting your own A record for the internet to get? [04:30] I use BIND because that's what I know, but you probably want dnsmasq [04:30] Hilikus: what router do you use? See if it supports icmp-redirect. [04:30] in my laptop i had to come come up with a script that altered /etc/hosts depending if the SSID was my home's or not [04:30] but now with my phone i can't do that [04:31] twd i have no idea, don't know much about dns [04:32] i just want to be able to use the same name to access my server from the lan and wan, because i have bookmarks, settings and stuff like that that have the name of the server stored, so when i go out and try to access it they don't work [04:32] Hilikus: why not just use split-horizon DNS for the router's A record? [04:32] well, if your router supports it, the icmp-redirect will be the easiest [04:32] Sorry, bad completion [04:32] MarchHair: why not just use split-horizon DNS for the router's A record? [04:33] qman__: didn't you say that would be dangerous or you were talking about something else? [04:33] something else [04:33] MarchHair: my router is pretty crappy, but i'll look [04:33] twb: if he's using something like hilikus.dyndns.org, would split-horizon make the rest of dyndns.org unaccessible from inside the LAN? [04:34] MarchHair: I don't believe so; you'd only be splitting hilikus.dyndns.org, not the parent domain [04:34] even what I was thinking of is not inherently dangerous, it just goes against the damage-control ideal should you get a malicious user or program on your network [04:34] twb: you can do that without being authoritative for the parent? cool. gonna go look. [04:34] MarchHair: I don't know. [04:35] so the three options i have is either to run a dns server, to configure icmp-redirect and this split-horizon thing? [04:35] $coworker won't let me do any significant split-horizoning due to him being a fuddy duddy [04:35] Hilikus: they all require your router to not be shit [04:35] FSVO router = dhcp/dns server [04:35] uhm [04:35] no, if he has internal DNS resolve the external name to an internal IP [04:35] twb running a dns server would work anyway, no? [04:36] his router doesn't need anything special [04:36] twb, hilikus: just read. yeah. split-horizon will do the trick too. [04:36] dnsmasq has a nice option for split-horizon where it'll just magically work out the Right Thing based on interfaces and the requestor's IP and the list of A records in /etc/hosts [04:37] sounds like the split-horizon dns is a good option. [04:39] Can't find the option now, though :-/ [04:40] Ah, localise-queries [04:40] Doesn't work for IPv6 [04:41] twb: what does? IPv6 support needs to hurry up. [04:42] MarchHair: dnsmasq's split-horizon (--localise-queries) is IPv4 only at present. [04:42] I wouldn't be surprised if I'm telling that to my grandchildren [04:42] qman__: tell me about it :-/ [04:42] If a person were to modify /etc/init/samba.conf to tweak samba's initialization, could I expect updates to samba to over ride my alterations? Or would my customized /etc/init/samba.conf file be in danger of being overwritten? [04:42] errr would it NOT be in danger rather. [04:43] Lord_Devi, all the times I have run into that, it asked me what to do [04:43] qman__: yup, as we all huddle around fires talking about how the CIDR block shortage has returned us to the stone age.... [04:43] of course, those were during release upgrades [04:45] Hrm, yeah ok [04:45]