[00:13] hi all... i need some assistance configuring a cluster file system... [00:13] have been trying to setup using glusterfs [00:14] any one had any success with this? [00:16] is anyone even here? [00:24] morning [00:24] o_O [00:24] anyone here installed ubuntu server in blades? [00:25] i know who did [00:25] they are XHS21 XM it is a bladecenter H [00:25] orudie: can I contact him? [00:25] yeah [00:26] irc.oftc.net #linode , talk to caker [00:26] I hate IBM because doesnt support Ubuntu for their virtualization manager agent :S [00:26] orudie: thanks [00:26] jmarsden, you should use xen [00:26] I use xen [00:27] but their agent isnot supported [00:27] it works for rhel an suse :S [00:27] jmedina, you run a company ? [00:27] orudie: yeap [00:27] this is a new customer who wants ubuntu server :D [00:28] jmedina, can i work for you ? [00:28] orudie: sure im in mexico city [00:28] lol [00:28] how about over the internet [00:29] well I really dont know, I'm only the linux guy [00:30] i thought you said you run the company, meaning you are the boss [00:30] nop sorry [00:30] hey folks... I am needing some help to set up a file server cluster [00:30] i also want to be a linux guy [00:32] orudie: haha! [00:32] mattt, what ? [00:32] i have been trying to set up using glusterfs... but keep running into problems so now I am looking at ocfs... any one have any ideas? [00:32] "can i work for you?" [00:32] :) [00:34] so ? [00:34] whats so funny ? [00:34] anyone at all.... any ideas on clustered file system using ubuntu....? [00:36] jmedina, they are they guys i bought VPS from and installed ubuntu server [00:37] orudie: ok, I didnt know about linode services [00:37] jmarsden, they are great dude , if you decide to get one let me know so that i would give you a referal code [00:37] www.linode.com [00:38] caker is the owner [00:39] capkirk, looked into GFS? [00:41] is that glusterfs? [00:41] no [00:41] Global File System [00:41] gfs-tools is in the repos [00:41] ok, then no i have not.. but I noticed that it is included with the 8.04 Server cd [00:41] yes [00:42] ok, i found gfs-tools on the cd. u think i should give it a go [00:42] orudie: thanks but I have my own rack for vps hosting [00:42] jmedina, :) [00:42] capkirk, I have never tried it, but I have clients that use it [00:42] so there you go , they are your competitors [00:43] orudie: then I want the code :D [00:43] i doubt they'll give you [00:43] you can try to buy it from them though [00:43] i'm very interested what they are gonna say [00:44] !ot | orudie [00:44] orudie: #ubuntu is the Ubuntu support channel, for all Ubuntu-related support questions. Please use #ubuntu-offtopic for other topics. Thanks! [00:44] just reading up on gfs-tools... looks like it might be the go, will do some more reading first..... [00:44] foxbuntu, dont be an asshole [00:44] !admin | orudie [00:44] Sorry, I don't know anything about admin [00:44] capkirk: well you can read redhat offfical docs about gfs [00:45] lol foxbuntu you are so miserable [00:46] orudie, move on and keep the chat out of here or I really will have and admin step in [00:46] go ahead [00:47] ScottK, you are an admin right? [00:48] what a looser lol [00:49] orudie, this is a support channel to try to help others, not your personal chat room, so as I tried to nicely point out earlier. Go to -offtopic if you want to chat [00:50] ping baffle [00:50] i'm in this channel 24/7 [00:51] you dont have to educate me [00:51] maybe u should get a life...? [00:51] idling that is [00:51] well I dont see a problem, it was ubuntu related and there is no active conversation at the moment === foxbuntu is now known as orudie_thefaggot === orudie_thefaggot is now known as foxbuntu [01:14] ok... now i am having recurring problems with apt-get [01:15] logged in as root i type 'apt-get update' and i get various errors including NODATA2 [01:16] i am trying to install gfs-tools [01:16] if i enter apt-cache search gfs it returns nothing but the prompt, even after running apt-get update [01:18] capkirk, sounds like you might need to make sure your sources.list is alright [01:19] this is a fresh clean install of ubuntu server 8.04... should sources.list need udpating? [01:20] capkirk, nope [01:20] capkirk, you sure you have a web connection then? [01:20] if i type host google.com, it returns all the right numbers.... [01:21] capkirk, try: ping 4.2.2.2 [01:21] ping works as expected [01:21] ok [01:21] is it possible that our company proxy server is sending cached info to the apt-get command [01:22] ah, did you setup a proxy? [01:22] I mean, the proxy connection [01:22] yes, i am running ipcop [01:23] no need, ipcop is running transparent proxy mode [01:23] gotcha [01:24] you said you install 8.04? or is it 8.04.1? [01:25] im fairly certain it is 8.04, is there some way i can check [01:25] lsb_release -a [01:26] can some one help me with port forwarding [01:27] ok, it says description ubuntu 8.04.1 release 8.04 [01:29] capkirk, ok well perhaps the repos for that have been moved...lets have a look [01:40] capkirk, I think you can point to old-releases.ubuntu.com to get those repos, however there is major security vulnerability in 8.04.1, you would do best to: sudo do-release-upgrade [01:40] ok, will do upgrade now [01:40] foxbuntu: 8.04 is supported. [01:41] kees, hasn't it been moved to old-releases? [01:41] 8.04(.2) is a long term support release and should be supported for a further 4 years [01:41] hmmm says no new release found [01:41] 7.10 is off the repo (and in old-release) but 8.04 (Hardy) will be supported for a long time [01:41] kees, ah [01:41] yes, that is why i am using 8.04...LTS [01:41] capkirk: just sudo apt-get update && sudo apt-get dist-upgrade and you can get any available updates. [01:41] kk, will try that now [01:42] capkirk: if you *want* to upgrade to intrepid or jaunty, you'll need to tell it so. [01:42] kees: that's if -updates and -security is enabled :-) [01:42] Nafallo: very true, I did make that assumption. [01:42] u mean in the sources.list [01:42] Nafallo: we need apt-get fix-yo-self [01:42] lol [01:42] kees: it is by default so... :-) [01:43] apt-get dist-upgrade is underway.... [01:48] hmmm... update seems to have stalled.... 23% [Waiting for headers] [01:50] its away again... must have needed a coffee break [01:51] (or a timeout...) [01:51] (...and retry) [01:53] yes, maybe, it is doing it again.... different site but seems to have stalled [01:53] working again... [01:54] this repo site is slow.... 15kb/sec [01:54] Which repo? [01:54] au.archive.ubuntu.com [01:56] Which ISP do you use? [01:56] bigpond adsl2+ [01:57] Hm, bigpond doesn't have its own mirror AFAIK. [01:57] no, i dont think so.... [01:57] internode's is externally visible, so I guess you could put it in there as well and see if it's faster than au. [01:57] well, at least it is working..just a bit slow [01:57] According to a TPG user I know, internode's is faster than everything else even for him :-) [01:58] ok... ill look into it [01:58] deb http://mirror.internode.on.net/pub/ubuntu/ubuntu hardy main [01:58] do i just edit the sources.list file and replace the au..... [01:58] or add it [01:59] capkirk: better to leave the au. entries there, at the bottom, so if internode isn't available or is out of date, it will fall back on using them. [01:59] On Debian there's a package called apt-spy which automatically puts the fastest repo in your sources.list.d, but it seems that Ubuntu doesn't have that. [01:59] ok, so sources.list works from top to bottom order of preference? [01:59] Yeah [02:00] i feel like such a linux newbie trying to achieve complex things.... [02:10] phew... dist-update complete [02:11] now running 8.04.2 [02:11] now back to gfs-tools [02:11] What of it? [02:12] i am trying to set up a cluster file system [02:14] gfs or ocfs... any opinions? [02:19] how do I upgrade 8.10 server to 9.04 server? (and should I?) [02:20] sudo do-release-upgrade [02:20] (and if you want to_ [02:20] Iceman_B1SSH: you may want to wait it out a few days [02:20] servers are hammered [02:20] or does do-release-upgrade now use apt-p2p? [02:21] apt-p2p? sounds awesome === Iceman_B1SSH is now known as Iceman_B|SSH [02:21] apt-cache show apt-p2p [02:21] meh, I can wait a few days [02:21] or maybe I should just wait until support runs out on 8.10 [02:21] so far it's doing everything I want [02:21] eh, why wait that long? [02:22] critical server? [02:22] nah [02:22] it's my home server [02:22] then upgrade in a few days or something [02:22] functions as a bittorrent client and I have a samba share for the local network here [02:22] oh and an irssi client of course [02:22] :) [02:23] doing upgrades on a release day is rarely a good idea -- it hangs if it can't get the package, etc -- nasty stuff, at least in previous upgrades [02:23] I'll wait a few days then [02:24] hey folks [02:24] you;s think that in this day and age, Canocial would have scalable bw [02:24] I did the upgrade and am paying the price :D [02:24] Iceman_B|SSH: it has little to do with canonical [02:24] Iceman_B|SSH: all the mirrors are hammered [02:24] broke SSH and netatalk -suspect kerberos but not sure how do diagnose this one [02:24] SpaceBass: be more specific [02:24] log into the machine locally and read logs, to start [02:24] when I try and SSH in, it says "connection closed" [02:25] giovani, looking for the sshd log still, syslog is no help [02:25] /var/log/auth.log [02:26] says it cannot find pam_foreground [02:26] but then issues an auth [02:27] issues an auth? [02:28] after the error, there is a new line: "authorized to jdoe, krb5 principal...." [02:29] doing some google-fu on that error now [02:31] when I go dpkg -l I get lines like "rc bsd-mailx 8.1.2-0.20071201cvs-3 A simple mail user agent" [02:31] does that mean, the rc at the beginning [02:32] +what [02:32] some brilliant admin put a note about jaunty in the header of the ubuntu forums so every single post is a match [02:32] Iceman_B|SSH: read the header/manpage [02:33] Desired=Unknown/Install/Remove/Purge/Hold [02:33] | Status=Not/Inst/Cfg-files/Unpacked/Failed-cfg/Half-inst/trig-aWait/Trig-pend [02:33] |/ Err?=(none)/Hold/Reinst-required/X=both-problems (Status,Err: uppercase=bad) [02:33] cryptic message ahoy :/ [02:34] not cryptic at all ... [02:34] very clear [02:34] r = remove [02:34] c = cfg-files [02:34] because it wasn't purges [02:34] purged* [02:34] ....oh, I have to read it like that [02:35] they should have underlined the first letter of every state and made the descriptions all lowercase [02:35] I said to read the header [02:35] it's clear from the capitalization [02:35] I dont know what you meant by "header" [02:35] that's why it's "trig-aWait" for example [02:35] then you're in trouble ... [02:35] why? [02:36] sigh [02:36] because "look at the header" should be clear [02:37] then I must have glossed over a linux tutorial that explains that that term should be standard issue knowledge under serveradmins ._. [02:37] header is an english word [02:37] but that's cool, I'll look into it soon, sleep first [02:37] it's not linux-specific [02:38] I know, but header can mean a lot of things [02:38] it always means the stuff at the top [02:38] anyone doing kerberos/ldap auth in 9.04? [02:38] okay [02:38] or, the "first" stuff, more specifically [02:38] SpaceBass: I will be in a few days, 9.04 was just released today [02:38] thanks for your input, Ive gotten a bit wiser [02:38] later [02:39] giovani, will be curious to hear your results [02:39] SpaceBass: 9.04 ldap may work fine [02:39] upgrades are messy [02:39] and can break non-standard configs [02:39] thats what I thought... [02:39] I understand about upgrades... [02:39] so, ldap auth working in 9.04 won't help ya with your broken upgrade :) [02:40] doesnt mean that its unrealistic to try and get it working again [02:40] yeah, sometimes it's easier to start with a fresh install, when possible [02:40] not a great option for me...but if I have to rebuild from scrach then I will [02:41] ok [02:42] would prefer to troubleshoot what I got [02:43] but it does all point to pam now that I dig in [02:44] yep, it's probably your pam config, or any changes you made to pam when enabling kerberos [02:44] since that's non-standard [02:44] probably new pam version doesn't like the config, or something along those lines [02:44] well for that matter anyone who adds a user is non-standard [02:44] we should all run the live CD [02:46] except that the format of /etc/passwd never changes [02:46] so no [02:47] upgrades of a distro can never correct config files that were edited by the user [02:47] other than replace them, ify ou want [02:47] but that'd hard break your system [02:47] giovani: if you're talking about conffile conflicts, usually dpkg will offer to drop you to a shell to manually merge them. [02:48] twb: nope, that's not what I'm talking about [02:48] I'm talking about a potential config option in an old version of program X, which is no longer compatable with new version Y [02:48] Ah. [02:48] during an upgrade, if you want to keep your old config ... it may not work as expected with new version Y [02:48] That would need to be handled specially in the .postinst. [02:49] twb: maybe you want to help SpaceBass troubleshoot his non-working kerb auth in a 9.04 upgrade [02:49] I've NEVER gotten kerb working [02:50] like I said ... non-standard :) [02:50] SpaceBass: are you trying to fix a kerb client, or the kerb server? [02:52] kerb client, I believe [02:52] client [02:52] yeah [02:53] worked in every release since 6.x (when I started using ubuntu)... usually pretty straight forward... config kerberos, add a key file, config the pam files and boom, it works [02:53] SpaceBass: well, there are two parts. You need to make sure nss is working (/etc/nsswitch.conf), and once getent agrees (e.g. "getent passwd fred" for a Kerberos account "fred"), you need to get pam working. [02:54] does anyone know about ATAoE...? [02:54] twb, thanks! have confirmed nss is working [02:55] can still log in locally as network users + getent passwd returns what I expect [02:55] think its pam ... although all that appears in my auth.log is a successful auth [02:55] I am trying to set up a shared cluster storage arrangement with disk redundancy [02:56] SpaceBass: OK, now log in as a kerb user and confirm that you have received a valid TGT. [02:56] roger that [02:56] locally I can kinit, and if I try and ssh in, I get the approprate ticket (and then it says "connection closed") [02:56] appreciate the help btw - thanks [02:56] Does auth.log say why you were kicked out? [02:57] no - one line for the ssh transaction... auth [02:57] Try to work out how to turn on debugging for the pam_krb module(s) and for sshd. [02:59] got sshd in debug mode now... checking [02:59] "access denied for spacebass by PAM account config" [03:04] So now you need to grovel through pam debugging, which is enabled on a per-module basis. [03:09] :( [03:10] seems to be a lot of infor re red hat and cluster file systesm.... anyone know anything about ubuntu and file system clusters? [03:10] i can spell, just can't type ^^^^ [03:11] too many years on windows .... clicking a mouse :P [03:15] You could try wiki.ubuntu.com [03:16] yes, seems i have been all over the web... just cant find anything that explains it? [03:16] i know what i want to achieve, just not sure what tools or packages to use to make it happen [03:18] twb, thanks for the hand holding [03:18] think I got it - ssh is at least working [03:18] Cool. [03:18] i'll recompile netatalk tomorrow [03:18] i have a number of older pc's which i want to configure in an array of some sort so that the disks can be configured for redundancy and shared as a single disk share on the local net [03:21] Hi. I currently try to play with the ICU library. it already came shipped with my distro, so I don't need to install it myself. however, I don't have the header files needed. is it normal for them not be shipped with the library? and if I download them from the icu website, where should I place them? [03:21] anthony1x: you need to install the associated -dev package. [03:21] ok, so i am trying to set up a SAN array using old PCs... [03:22] twb, I already searched for it. none found for libicu [03:22] anthony1x: then you are searching wrong. [03:23] twb, [03:23] twb, how should I search correctly then? [03:23] aptitude install ~nlibicu.*-dev [03:24] capkirk: is this more of a learning exercise or something you want put into production? [03:25] capkirk: stickystyle has a point; this is not the kind of thing you want to put production data on. [03:26] You would be many times better served just by trowing all those drives in one box and learning mdraid. [03:26] twb, that did the trick. how come sudo apt-get install libicu and then hitting tab to get a list of all available packages does not list the dev package? [03:26] anthony1x: probably because you have not loaded /etc/bash_completion. [03:26] twb, thanks anyway. [03:26] stickystyle: or even better, by buying three or four brand new 1TiB drives and putting them in a dedicated file server, RAID5'd. [03:27] Using old disks fills me with horror, especially since SATA is ridiculously cheap now. [03:27] twb: very true [03:27] For a 3×1TiB SATA array you're looking at AUD500 for the disks. [03:28] Make that 450. [03:28] capkirk: but if your just looking to learn and play (which is always good) check out glusterfs [03:28] For a 4×1.5TiB SATA array, AUD860. [03:29] Plus the case, motherboard and CPU, of course, but you don't need anything special for a file server (except dual gigabit NICs). [03:30] hitting the sack - thanks again tbw, appreciate your help and patience :D [03:30] Actually, I haven't run the numbers. If the internal PCIe or SATA bus is the bottleneck, you could just use one gigE NIC. [03:31] ok, i take your point as very valid... this is for production use... but budget is tight....(as usual) [03:32] what is this mdraid you speak of? [03:32] mdraid == software raid [03:33] ok, so this is for a network attached file service, shared by samba [03:33] capkirk: if you can come up with a grand, you can have a file server that will kick the arse of anything you build from hand-me-down workstations -- faster, more reliable and able to store more by several orders of magnitude. [03:33] if its for production, you need to always consider your time as a cost in the budget...how many hours will you have to tweek something to get it going, and how many times will you have to go back. [03:34] too true... could probably find a grand.... [03:34] shoot, for a grand you could build a pretty rocking bit of storage, removable drive trays and everything. [03:34] we have an old server running w2k that i want to retire and move to ubuntu server and samba. [03:35] You just say to your boss: "look, I can spend a week setting this up, and it won't work well, and I'll have to go back to it every month or two to fix something... or you can front a grand now, and we will have a state-of-the-art system that you can leave running for ever." [03:35] stickystyle: depends on the currency; I was working in AU$1000 :-) [03:35] twb is dead on. [03:36] For US$1000 you could throw in removable bays and a nice rackmount case. [03:36] You'd end up with about 3TiB of storage and the ability to lose one of the four drives without problems. [03:37] If you count up all your 200GiB hand-me-downs, they probably won't add up to 3TiB even BEFORE you take into account the loss from parity. [03:37] And parity loss is higher the more disks you have. [03:39] parity loss is lower the more disks you have [03:40] (as a percentage of the array) [03:40] i.e. 1 parity for 3 drives is 33.33% loss [03:40] 1 for 4 drives is 25% loss, etc [03:43] ok... i am liking this advice... makes sense... [03:44] so an external unit with 4x SATA drives in some kind of raid array for redundancy connected via gigabit ethernet to a ubuntu server shared via samba... have i got it right? [03:44] giovani: you're right; sorry. [03:45] is only for local file serving and application data store so ultra high speed not really required [03:45] giovani: I guess I was assuming you'd scale up the number of parity disks proportionally to the number of disks in total. [03:45] With only four drives you could easily find a case to hold them all in rather than doing external. [03:46] yet again you speak wisdom.... [03:46] capkirk: not "an external unit" -- just an ordinary rackmount/tower case and motherboard [03:46] and the raid array can be setup *inside* ubuntu with mdraid [03:46] so all you need is a regular old computer + 4 drives + ubuntu [03:47] pretty much every motherboard comes with GigE built-on now. [03:48] s/old// [03:48] you can swap the 'regular old computer' get fancy and buy a rack mount computer, with removable drive trays if you desire [03:49] twb: yeah, didn't really meen *old* [03:49] There's no point in doing so unless you already have a rack to put it in, of course :-) [03:49] Excellent point! :D [03:50] ok, well i do have a rack, but only as a storage cabinet, no rack mount servers, all shelves and stuff [03:50] plenty of 'old' computers quite capable of running ubuntu and 4x disks.... just a little problem but.. [03:51] it i install 4x disks and no cd rom... how do i install os? [03:51] capkirk: for the cost, I would buy a $50 motherboard and $100 CPU anyway, just because. [03:51] capkirk: there are a number of ways -- you can install from a hard disk, or you can install from the network. [03:51] capkirk: you could use the eSATA port to connect a SATA CD drive. [03:51] *old* pcs... no sata, only ide [03:51] capkirk: or you could set up the array to start of degraded, then latter connect the third disk. [03:52] capkirk: don't use PATA for a new system. [03:52] PATA costs will only go up. [03:52] maybe i should bust open some old boxes... never know, might find some sata .... [03:53] HD's are dirt cheap http://www.pricewatch.com/hard_removable_drives/ $44 for 500GB [03:54] And a full tower size case should fit four drives with a CD/DVD-ROM no problem [03:54] is there such a thing as pci - sata cards [03:54] capkirk: yep [03:55] capkirk: yes, but I'd still say just buy a new motherboard. [03:55] capkirk: it wil still cost less than a grand. [03:56] For commodity hardware, trying to reuse gear that's three or five years old and a generation behind will screw you in the long run. It's a lesson I learnt the hard way. [03:56] what about pwr supply wattage ? [03:56] capkirk: you'd have to calculate that. [03:56] If you get a decent case, it will include a PSU. [03:57] so let me summarise to see that i understadn.... [03:58] 1x pc with 4x sata drives ... average cpu and ram sufficient with ubuntu and mdraid [03:59] a second machine which is the acutal 'server' with ubuntu and samba sharing the drives on the 1st pc.... [03:59] or just run everything from 1 computer [03:59] Your over engineering it, just one computer [04:00] k, was still thinking of scalability... [04:00] how many users? [04:00] we have an average use of about 30 users connected at any one time [04:01] and it wouldn't be a scalable solution as your bottle necking everything through the network [04:01] what kind of data? is this their home folders that they are mounting from their local machines? [04:01] i was thinking allong the lines of SAN network... [04:01] i.e. they are running off the server [04:02] no roaming profiles or anything like that [04:02] just docs, spreadsheets etc and some shared application data files [04:03] I worked at a place where we had a 600MHz P3 serving 220 users shared data [04:03] I don't think you have to worry about scaling just yet [04:03] true ok, i will put that concept aside then. [04:03] you don't want to make the problem harder than it is [04:03] there is one next issue thought... virtual machines. [04:05] capkirk: http://paste.lisp.org/display/79118 [04:05] that's another chapter in server design. [04:05] That's with an overspecced CPU and an expensive case, though. [04:06] I also didn't bother to check for gigE NIC. [04:06] twb: I havn't seen a board that didn't come with GigE in a while. [04:06] For a file server, a 64-bit Celeron D (if they still make those) would be sufficient. [04:08] current server we running is p4 3.2G 3G ram with 1x250Gb drive W2K server [04:08] i loaded vmware 2.0 onto it and killed it... [04:08] so i took it off again [04:08]