[00:00] kirkland: walrus just went in with OutOfMemory [00:00] hggdh: arse [00:00] hggdh: tell me more [00:01] kirkland: I am running another 400 instances, with no reinit [00:01] hggdh: okay [00:02] hggdh: same separate topo? [00:02] kirkland: interesting enough, the CLC is surviving with ~816M of real, and the Walrus barfed with ~720M [00:03] kirkland: same separate topo, topo2 [00:03] hggdh: hrm, yeah, that's weird [00:04] kirkland: OTOH, memory growth is slower [00:05] hggdh: nope - you can use the test rig [00:05] kirkland: ^^ [00:05] kirkland: hum. This time the walrus vanished. This is new. [00:05] mathiaz: th [00:05] thx [00:06] hggdh: ? crashed? [00:06] kirkland: not sure yet, I see a "mabolo init: eucalyptus main process (1224) terminated with status 1" in the syslog [00:07] hggdh: hmm [00:08] anyway. Saving the logs now [00:17] kirkland: topo2-1.tar uploaded [00:17] kirkland: this is better than before, but no cigar yet [00:18] hggdh: okay, thanks for testing [00:18] hggdh: can you please leave feedback and pointers to your results in that bug? [00:18] kirkland: will do [00:18] hggdh: oh, and dpkg -l euca* [00:19] hggdh: is it ppa2 ? [00:19] kirkland: indeed. I will add a 'dpkg -l euca\* uec\*', but I already noted it in the bug [00:19] hggdh: cheers === dendrobates is now known as dendro-afk === dendro-afk is now known as dendrobates [00:47] RoAkSoAx: merged [00:47] kirkland, ;) [00:47] RoAkSoAx: i'm going to release 1.37 with this [00:47] kirkland, awesome [00:47] RoAkSoAx: won't make lucid, unfortunately [00:47] kirkland, i guess the next release would be the modularization? [00:47] RoAkSoAx: please [00:48] RoAkSoAx: how major are these changes? [00:49] kirkland, modularization? well i'd say major... changed various things such included the creation of the ISO list into the code, splitted the code up, some adjustments to get the same functionality [00:50] kirkland, and just the obtaining of the release codename.. I think it is SRUable [00:50] RoAkSoAx: well, I think I'm just going to SRU a one-liner that does r = maverick [00:50] RoAkSoAx: when the time comes [00:50] RoAkSoAx: or we can try to get your changes in [00:50] kirkland, wouldn;t it be better if we backport it? [00:50] RoAkSoAx: either way [00:50] RoAkSoAx: yeah, maybe so [00:51] RoAkSoAx: hmm [00:51] kirkland, we could backport when modularization is done [00:51] kirkland, or maybe try to get today's changes into lucid [00:51] RoAkSoAx: well, there's not much to backport; just grab the diff of the merge i just committed [00:52] RoAkSoAx: that might be a possibility [00:52] RoAkSoAx: would you ask about that in ubuntu-motu? [00:53] kirkland, let's see what they say [00:53] RoAkSoAx: keep me posted; we can try to get this uploaded right now if they approve [00:54] kirkland, ok we could probably get FFe because it is *major* change but simple at the same time [00:54] RoAkSoAx: right [01:00] kirkland, ScottK, could I still file a FFe for new upstream of TestDrive that adds a simple, but necessary functionality? [01:00] RoAkSoAx: I'd advise filing an FFe for that: it's not clearly bugfix. [01:00] RoAkSoAx: You could probably do an s/lucid/maverick/ upload as clearly bugfix though. [01:00] RoAkSoAx: yeah, i'm leaning toward uploading 2 1-line SRUs per year [01:01] * persia isn't actually invested with any authority, and was only offering advice [01:01] kirkland, i'd go for backporting it [01:01] RoAkSoAx: :-) [01:01] RoAkSoAx: if you're willing to fight that battle, I'll support you [01:01] RoAkSoAx: but it's your battle to fight ;-) [01:01] kirkland, I have no problem with that :) [01:02] RoAkSoAx: go for it, dude [01:02] kirkland, at the end, It would save *time* by having to SRU twice a year [01:02] RoAkSoAx: alrighty [01:02] RoAkSoAx: go for it ;-) [01:03] kirkland, however, if ScottK or anyoine else from the release team gives us the go we could jsut upload 1.37 [01:04] anyways we'll have to wait [01:04] RoAkSoAx: alrighty [01:04] RoAkSoAx: you have motu props, right? [01:05] RoAkSoAx: if you've tested this functionality very thoroughly, and ScottK approves, just go for it [01:05] RoAkSoAx: i tested it here; seemed to work for me [01:05] RoAkSoAx: retrieved lucid, used it from the cache, i manually removed that file, and touched the timestamp back a few years [01:05] kirkland, i actually have tested it in, what i believe, all test cases possible [01:05] RoAkSoAx: all of those worked as expected [01:05] RoAkSoAx: what about no network connectivity? [01:06] kirkland, done too [01:06] RoAkSoAx: coolio [01:06] kirkland, if no network and cache is valid, it uses chace [01:07] though, if cache is expired it will fail to launch [01:08] wait, let me test brb [01:11] kirkland, ok, if no network, and cache is not expired works. If no network and cache is expired, it still works, however, it throws and error message saying that couldn't update the cache... [01:11] RoAkSoAx: cool [01:12] kirkland, ok then I'll file FFe [01:31] hey kirkland [01:31] I don't mean to be critial, but the lens flares on the manpages screams 1972. [01:32] jcastro: So you're celebrating references to a classic age of visual theming? [01:32] kirkland, bug #570485. I'll ping someone from the release team and keep you updated. [01:32] Launchpad bug 570485 in testdrive "[FFe] New upstream release 1.37" [Undecided,New] https://launchpad.net/bugs/570485 [01:33] persia: well, at least he didn't bust out blender. :p [01:33] I've modified .bash_aliases to include a shortcut to mounting a samba share on my machine (it's also the samba server). I'd like to convert that mount command to a line in fstab so the mount happens on startup. So far I've been unsuccessful. Any help? alias bpomount='sudo mount.cifs //localhost/BPO /home//BPO -o user=,credintials=/etc/samba/smbusers,file_mode=0777,dir_mode=0777' [01:37] With regards to the Ubuntu Enterprise Cloud, I am looking to setup 1 front end with 4 Node Controllers. I have the front end on a public addressable ip, and the node controllers on the same public. I bridged the public interface, but want to run VMs on a private network. Is there anyone that has experience with this I can talk to ? [02:11] New bug: #570494 in samba (main) "package samba-common 2:3.4.0-3ubuntu5.6 failed to install/upgrade: subprocess installed post-installation script returned error exit status 128" [Undecided,New] https://launchpad.net/bugs/570494 === samferry is now known as TheLord === TheLord is now known as samferry [04:01] New bug: #570513 in nis (main) "NIS-based autofs maps don't load on startup" [Undecided,New] https://launchpad.net/bugs/570513 === lifeless_ is now known as lifeless [04:23] kirkland: noticed the acpi fix. I'll check in the morning and let you know how I fare. Ty. [04:24] [2010-04-27 07:17:50]: uid: (1018/test.stormraidergames.com) gid: (1017/test.stormraidergames.com) cmd: test.py [04:24] [2010-04-27 07:17:50]: target uid/gid (1018/1017) mismatch with directory (1018/1017) or program (0/1017) [04:25] i see the issue but i duno how to fix it :/ [04:25] well i think its the issue [04:31] i know that 0 should not be a 0 [04:31] but my lack of experince isn't telling me how to fix it [04:38] How can I force a certain package to never be installed? I.E., I want to install other things using apt-get that use Subversion, but I already manually compiled and installed the latest version of Subversion. I don't want the older version (re-)installed. [04:39] you can make a placeholder package .. I forget the name [04:39] equivs [04:39] ziesemer_: apt-get install equivs .. read the docs.. you can make an empty package that satisfies the same dependencies [04:40] http://www.debian.org/doc/manuals/apt-howto/ch-helpers.en.html . Excellent - thanks! Just needed a place to start... [04:40] * SpamapS 's debian skills are rusty but still intact.. woot [04:40] Of course, it'd be better to not manually compile and install things [04:40] twb: sometimes the learning curve of re-packaging something inflexible and complex is higher than equivs. ;) [04:41] (I'm surprised that subversion changes enough that you even NEED a newer version.) [04:41] I used to use equivs for java.. I'm sure thats not a problem anymore. [04:41] SpamapS: granted. [04:41] * ziesemer_ Cheers to SpamapS. [04:41] ziesemer_: twb has a very valid point.. you might find it easy to just tweak the repository subversion package [04:41] SpamapS: If the effort of repackaging is more than running uscan and uupdate, or calling get-orig-source, then please file a bug about a package being hard to update. [04:41] persia: rebasing the debian/patches is usually the hardest part [04:42] twb: Yeah, I'd use the package version, if it was anywhere near up-to-date... [04:42] Yeah subversion isn't a good example of this [04:42] That and when upstream changes build system :-/ [04:42] But I will investigate the repackaging bit later! [04:42] ziesemer_: it's not "up to date" for a reason. [04:43] twb: Care to elaborate? I'm looking for specific bug fixes in the 1.6.11 release. [04:43] twb: Sure, but it *shouldn't* be hard for 99% of packages. If it's hard, then something is wrong. That something might just be that the patches in debian/patches need to get upsteam, or similar :) [04:43] ziesemer_: with subversion, its quite likely you can do just as persia says [04:44] ziesemer_: what release of Ubuntu are you running? [04:44] 10.04 RC. [04:45] Well, I guess it's a bit late in the release cycle, but a few weeks ago you could've reported this; it might be important enough that the Ubuntu subversion maintainer would backport those fixes from the newer version. [04:46] hrm.. so I didn't even know about uscan/uupdate.... [04:46] always applied the patch manually to the upstream sources.. [04:49] What's the roadmap for Ubuntu migrating to dpkg-source format 3.0 (quilt)? [04:50] Does your infrastructure already handle it, so you'll just migrate piecemeal whenever you resync a 3.0 package from Debian? [04:50] twb: yes, there are already a number of packages in the archive using it [04:50] Yay! [04:51] I don't think there's any plan for wholesale migration, just as there isn't for debian yet, while they argue it out [04:52] dh(1) is also super sexy. [04:53] sure, it gets used a bit [04:54] whats dh? [04:54] SpamapS: developer stuff [04:54] * SpamapS has been trapped in a CentOS shop for 6 years and just re-entered debian based happy land [04:55] SpamapS: it's a way of reducing boilerplate in debian/rules. Like CDBS, but without the gratuitous gmake-fu. [04:55] twb: does it have a home? I'm quite interested in developer stuff. :) [04:55] SpamapS: apt-get install debhelper; man 7 debhelper; man 1 dh dh_auto_configure [04:55] danke [05:00] twb: Debhelper 7 and source format version 3 are pretty unrelated. [05:01] by which you mean entirely ? :) [05:03] I think they are vaguely related: they both attempt to provide once-and-for-all solutions to long-standing packaging annoyances [05:04] Mind you, operationally there's no link. [05:08] persia: You can completely use one without the other, so I don't understand. [05:08] lifeless: Yes. [05:11] so.. I'm playing around re-packaging libmemcached to the latest version (0.31 in my karmic virtualbox instance, I want to have 0.40) ... the libmemcached2.symbols file has 0.31 listed as the version.. how do I update that file? I presume it is generated [05:11] ScottK: The relation would be entirely in the minds of folks looking to address issues: there's no operational link. [05:12] SpamapS: The best doc we have about that is https://wiki.ubuntu.com/stefanlsd/dpkg-gensymbols [05:14] persia: thx that helps. :) [05:17] persia: For the vast majority of packages, I'm not aware of any issues that V3 solves. [05:19] I spent about two days once unwinding issues with linda that v3 would have solved. It also solves all the annoying uuencode/uudecode issues with adding icons. [05:19] But yeah, for a clean package with a good upstream relationship, there's no benefit. [05:20] Conceptually it's more complex and raises barriers to entry. [05:41] New bug: #562852 in likewise-open5 (universe) "Please remove likewise-open5 from lucid on all architectures" [Medium,Confirmed] https://launchpad.net/bugs/562852 [05:41] ScottK: I realize dh(1) and 3.0 (quilt) are unrelated; they're just both sexy :-) [06:31] New bug: #570533 in openldap (main) "package slapd 2.4.21-0ubuntu5 failed to install/upgrade: podproces zainstalowany skrypt post-installation zwrócił kod błędu 1" [Undecided,New] https://launchpad.net/bugs/570533 [08:27] I just tried the Ubuntu self assesment on http://www.ubuntu.com/training/certificationcourses/server [08:27] and scored 9 of 10 [08:28] and can't find what would be my wrong answer [08:35] hmm.. same here.. === ttx_ is now known as ttx === sourcode__ is now known as sourcode [10:01] I have reprepo repository with slightly customized packages that I'd like to use on several of my server machines. Now I'm trying to figure out how to make those servers prefer my packages over ubuntu's, even if there's a newer version in ubuntu's repos. Is this something I'd do with apt preferences? [10:02] You probably want pinning [10:02] It's a major pain in the arse to get right. [10:04] hmmm [10:04] Well, I have a theory that might work... [10:04] I'm building these packages myself, so maybe I can add an Origin field in the control file of each package I build, and then perhaps: “Pin: origin "something"” [10:04] would that work? [10:11] stephank: Something like that, yes. [10:12] stephank: And you want the pin-priority to be > 990. [10:12] stephank: See the apt_preferences(5) man page. [12:03] there are still some issues with upstart and bonding [12:03] it starts network services before bonding [12:04] argh [12:05] * twb feels better about the relatively staid startpar(8). [12:26] New bug: #570636 in samba (main) "package samba-common-bin 2:3.4.0-3ubuntu5.6 failed to install/upgrade: subprocess installed post-installation script returned error exit status 2" [Undecided,New] https://launchpad.net/bugs/570636 [12:51] New bug: #570655 in vsftpd (main) "package vsftpd 2.2.0-1ubuntu2 failed to install/upgrade: subprocess new pre-removal script killed by signal (Interrupt)" [Undecided,New] https://launchpad.net/bugs/570655 [12:56] New bug: #570657 in openldap (main) "package slapd 2.4.21-0ubuntu5 failed to install/upgrade: subprocess installed post-installation script returned error exit status 1" [Undecided,New] https://launchpad.net/bugs/570657 [13:02] morning kirkland. Updating servers now to see if the acpi fix works. Yer awesome btw :) [13:02] rgreening: howdy; you applied that patch? [13:03] kirkland: got the 2x750gb disks, installing right now [13:03] cemc: cool, good to hear [13:03] cemc: keep me posted [13:03] kirkland: not yet. did the package get published or is waiting in queue? Do you have a ppa with deb? If not, I can patch... and test easy enough [13:03] smoser: ping [13:03] rgreening: not published; i just uploaded to lucid-proposed [13:04] rgreening: it was rejected for release, pushed to SRU [13:04] ok. cool. [13:04] rgreening: it's a one-line patch, though [13:04] rgreening: you don't even have to build the package [13:04] rgreening: just patch the source file on the OS [13:04] rgreening: /etc/acpi/powerbtn.sh [13:04] kirkland: on another note, it seems I cannot create vm's using vmbuilder since upgrading to lucid [13:04] rgreening: [13:04] -. /usr/share/acpi-support/power-funcs [13:04] +[ -r /usr/share/acpi-support/power-funcs ] && . /usr/share/acpi-support/power-funcs [13:05] kirkland: we'll need you to cover the UEC tests in ISO testing on your local rig (from ISO), as I don't have my cloud with me here [13:05] kirkland: ok, I'll manually fix and test [13:05] rgreening: talk to soren about vmbuilder [13:05] ttx: okay [13:05] k. [13:05] ttx: candidates are hot? [13:05] kirkland: yes, the current ones are expected to be final now [13:05] ttx: i'll sync and test after breakfast [13:06] I can do KVM based testing for the rest, but for UEC and the cloud images on UEC, we'll need you [13:06] soren: vmbuilder? any known issues? I'm getting "AttributeError: 'Libvirt' object has no attribute 'vm'" when attempting to build a vm [13:06] UEC images need a respin: don't bother testing yet (see -devel) [13:07] rgreening: bug 562189 ? [13:07] Launchpad bug 562189 in vm-builder "AttributeError: 'Libvirt' object has no attribute 'vm'" [Medium,New] https://launchpad.net/bugs/562189 [13:07] persia: they are not on the tracker anymore [13:07] thanks ttx. I'll review bug. [13:07] persia: but the UEC tests can be run on the karmic image [13:07] and the lucid images can be tested separately [13:07] ttx: me, and hggdh, right ;-) [13:08] kirkland: how many local machines does hggdh have ? [13:08] ttx: OK. Just wanted to make sure that slangasek's message from 10 minutes ago requesting a respin wasn't missed in creating extra testing. [13:08] ttx: none; lab machines [13:08] persia: I'm currently one meter to the left of Steve :) [13:09] Then you're entirely informed :) [13:09] kirkland: then he can do the UEC image validation [13:09] * persia feels a bit redundant when people have physical proximity [13:09] kirkland: we still need you to cover the UEC ISO tests [13:09] ttx: okay [13:09] at least the topology 1 ones [13:09] and if you have time the two optional topologies on amd64 [13:11] kirkland: you can keep the optional ones for tomorrow, since they take quite some time when run from ISO [13:12] * ttx should have taken his cloud with him, retrospectively [13:14] kirkland: ty ty ty ... acpi fix confirmed and working. [13:17] soren: Thanks for the pointers. It appears to be working nicely. :) [13:18] ttx: okay [13:19] * ttx covers random testing in KVM [13:21] kirkland: now that this work, If I perform a shutdown on the host server, will the guests be properly shutdown or do I need to manually issue shutdown to each vm (I dont recal reading anything on that) [13:26] ttx: not enough, one old i386, and two amd64 with 1G [13:26] and my bloody laptop [13:28] hggdh: but you can test t [13:28] arh [13:28] hggdh: but you can cover the UEC cloud images tset on the test rig [13:28] hggdh: when they will be made available [13:28] ttx: I can run them all in the rig, yes [13:29] ttx: certainly [13:31] kirkland: BTW, just the daily, or daily+PPA? [13:33] hggdh: just the current ISO [13:33] hggdh: which is proposed for GA [13:34] hggdh: the UEC tests on the ISO tracker [13:34] kirkland: install complete, and it boots, no problems [13:34] cemc: okay, well thanks for all your help [13:34] cemc: i don't know what's wrong with my disks, my scenario [13:35] kirkland: don't know either, maybe you can doo a step-by-step detailed list of what you're doing and I'll try again, if you want to [13:36] I was searching for the ubuntu server debootstrap but I was unable to [13:36] cemc: that's okay; i need to focus on UEC today [13:36] maybe I should install one of these: http://archive.ubuntu.com/ubuntu/pool/main/d/debootstrap/ and then change something afterwards? [13:38] astroboy: There's no difference in debootstrap especially for server. Why do you want deboostrap? [13:38] ttx, here now. [13:38] persia: well I have to install it on a remote server [13:39] * persia digs into kirkland's blog harder [13:41] astroboy: is the remote host networked? Do you have local waldoes? Do you have KVM or serial over IP? [13:41] smoser: no, just slangasek looking for you :) [13:41] persia: so I should install the debootstrap for 9.10 and then pass some flag to debootstrap? [13:42] * persia can't find it, unfortunately [13:42] twb: I have a netboot with access to the disks, I did that many times with debian. I was only wondering where to get the server version. [13:42] Appendix C installs are a pain in the arse, and should be avoided where possible. Preseeding, at least as far as the sshd udeb, is vastly preferable. [13:42] ttx, ok. so there is a uec respin and then population of iso tracker with amis [13:42] and then presumably need that tested ? [13:42] astroboy: the server version of what? [13:42] twb: ubuntu-server [13:42] astroboy: Just use the netboot if you're comfortable with netboot. You can choose the server tasks, and end up with a server. It's all the same software. [13:43] astroboy: ubuntu-server is always the server version. [13:43] oh ok. I thought there was a ubuntu with specific server packages [13:44] The only differences between the server and alternate CDs are the preseed files and the debs in the local pool. [13:44] so the the packages are exactly the same? aren't there something like gentoo profiles [13:44] mhm ok... [13:44] Nope. [13:44] smoser: yes, we'll need the EC2 image tested [13:44] Everything is exactly the same. [13:44] All variants of Ubuntu back onto the same package archive. [13:44] twb: "flavours" [13:44] smoser: any chance you can run your magic script ? [13:44] yeah. i can do that. [13:44] and of course there will be no bugs [13:45] yeah but usually I want bleeding edge on the desktop and stable software on the server, so I thought that the server version had different repos that installed more stable packages & compiled with different ./configure, something like that [13:45] e.g. gentoo [13:45] Nope. Everything is designed to be stable. [13:45] Gentoo doesn't belong on a server in the first place. [13:45] ok... thanks for the clarification [13:45] Some folks only upgrade their servers for LTSs though. [13:45] (as LTSs are *more* stable) [13:45] persia: I do that for end user desktops, too. [13:46] yeah I think I'll do that. 8.04 should be fine. [13:46] I hear bad things about non-LTS releases, like "oops, did you want libpam-ldap to *work*?" [13:46] astroboy: you should target 10.04 for a new deployment. [13:46] and anyway, gentoo is not bad as a server :P [13:47] astroboy: Or try 10.04. It releases in a couple days, so should be mostly good (or you might want to wait until ~june to have had the post-release bugfixes happen) [13:47] twb: is nothing critical. I just want something that works and doesn't get too much cluttered [13:47] mhm ok.... [13:47] Don't cannonical say you should wait till 10.4.1 to push into production? [13:47] I have been handed gentoo servers to fix. Gentoo on a server is Wrong. [13:47] I know maybe it's not the right channel to ask [13:48] gentoo as server worked peachy for me for many years [13:48] um... system-config-kickstart is broken [13:48] At least if you expect to maintain the server. [13:48] but it's running ubuntu server nowadays [13:48] If you adopt an OpenBSD-style "fire and forget" approach to deployment, I suppose it might work until the next CVE. [13:48] Which is usually when I get called in to un-bugger the box [13:48] but would you suggest debian or ubuntu for a simple server (some stupid websites and proxying basically...) [13:49] astroboy: yes. [13:49] if you want STUPID websites, run it on windows ;) [13:49] :) [13:49] twb: yes to what :P? [13:49] smoser: works for me [13:49] astroboy: I would suggest Debian or Ubuntu for pretty much any server. [13:50] zul: please cover some regular ISO tests. My laptop is awfully slow at it* [13:50] twb: no I meant "which between the two" [13:50] astroboy: doesn't matter much unless you are into support contracts. [13:50] ttx: doing the i386 tests now [13:50] twb: ok. then I guess I'll keep using debian [13:51] I mean I wanted to try something new on a new server :P [13:51] as I understand it, ubuntu server has newer packages than debian [13:51] astroboy: well if you already have a conclusion and just want to rationalize it... [13:51] twb: yeah eheh [13:52] Zider: IMO the non-LTS release are about as flaky as testing. By that view, they're not significantly older or newer. [13:52] twb: I don't know what LTS is [13:52] Zider: I actually like old and working software in servers... if I really want something new I compile it [13:53] Zider: Then you probably shouldn't deploy servers :-P [13:53] astroboy: I like new and security-fixed :P [13:53] twb: it's just a home server I'm running [13:53] Zider: those are mutually exclusive [13:53] new features implies new bugs, including new security bugs. [13:53] but not as well known as old security bugs [13:54] Security patches are backported to stable release. [13:54] *releases [13:54] yeah exactly.. it's not like debian is going to keep unsecure software unsecure.... [13:54] <_kettle> hey. looking for pointers to info on how to build my own ec2 ami. or should I just use ubuntu's images and customize those? [13:55] astroboy: not deliberately, anyway. *cough* ssh-keygen *cough*. [13:55] zul: ok I'm on amd64 now, but trying to run two KVM tests in parallel slows down my laptop so badly I prefer to only run one at a time [13:55] twb: ahaha yeah that was bad. seriously bad. [13:55] ttx: [13:55] ttx: "slows down" as in the non-VM environment? [13:55] twb: no [13:55] ttx: try nicing/ionicing the VMs? [13:55] ttx: you should get a macbook ;) [13:55] Oh, OK. Never mind, then. [13:56] zul: I should get a SSD drive and 2 more Gb of mem [13:56] zul: maybe when they go back to shipping a decent architecture... [13:56] ttx: that works as well [13:56] the main difference is that on the desktop I want to assume the risk of choosing what to use [13:56] on the server I trust the packagers more generally :P [13:57] "choosing what to use" means using the package manager, not funroll-looping. [13:57] --omg-optimizations [13:58] no I meant using development versions of almost everything :P === Barre_ is now known as Barre [13:59] Let's not disparage other distributions: we can do better to just explain how to do things in Ubuntu. [13:59] persia: I'm not disparaging a distro, but a mindset :-/ [13:59] astroboy: If you want the newest and sharpest (and don't mind getting cut once in a while), you can always run the Ubuntu development release whilst it's being constructed. [13:59] I'm not dispawhatever anything, cause I don't know what it means ;) [14:00] * persia tends to even upgrade non-critical servers come beta [14:00] persia: I personally run archlinux on my desktop [14:00] also because it's annoying to have a system with so much stuff installed [14:00] can't help you with that, but we've a nice desktop offering as well :) [14:00] You can also roll your own debs from VCS snapshots for critical packages, which is what I do for codebases I contribute to upstream. [14:00] and because of aur, mainly [14:01] Arch has zero Q/A, at least for Haskell packages. [14:01] I know but for example I don't use a de [14:02] twb: Q/A=question and answer? [14:02] Quality assurance [14:02] ah ok [14:02] you are right [14:02] expecially on haskel-xft and company [14:02] they broke down half of the time [14:02] but anyway [14:03] it's just that I choose what to install from the beginning [14:06] ah yes...ricers [14:09] rgreening: I'm working on a fix for that. [14:11] <_kettle> is anyone running ubuntu on Amazon EC2? I tried the instructions on https://help.ubuntu.com/community/EC2StartersGuide but "ec2-run-instances ami-2fc2e95b -k ec2-keypair" just fails with this error: Client.InvalidAMIID.NotFound: The AMI ID 'ami-2fc2e95b' does not exist === ogra_ is now known as ogra [14:23] could anyone kindly lead me in the right direction? I am searching for some decent documentation on the stability (or possibly lack of stability) of the EXT4 fs. I have been faithfully using JFS for years due to it never letting me down, but using JFS is a hassle so I am wanting to use another. I have read many opinions stating people's dislike of EXT4, but want to find out more. [14:23] mugwort13: ext4 is some minor improvements, plus extents, tacked onto ext3. [14:24] mugwort13: if you avoid extents, it's backwards compatible with ext3, so you shouldn't get any surprises there. (But extents are on by default.) [14:26] I suspect most of the annoyance is 1) people who got extentful ext4 before they were ready, and couldn't back out; or 2) XFS/JFS/btrfs/whatever weenies who value functionality/performance over stability. [14:27] using Btrfs in production is cool [14:27] only for the tough guys [14:27] btrfs isn't production-ready. [14:27] I'm joking :) [14:27] but XFS/JFS is stable [14:28] As at a month ago, there were severe data loss problems with /var/lib/dpkg on btrfs. [14:28] I'd say both are better tested than ext4 [14:28] Er, and that's on Debian, so it'll probably take longer to sync to Ubuntu. [14:28] doesn't surprise me - nobody said Btrfs was stable [14:29] RoyK: well, I worry that the noobs won't realize you were being sarcastic [14:29] :) [14:29] I wouldn't run XFS without a UPS, or in combination with any dm-based tech (including md RAID and LVM), though. [14:30] give Btrfs another year and it might catch up with early ZFS [14:30] twb: why? [14:30] Due to XFS' tendency to eat data after an outage, and the maintainers' tendency to tell anyone without write barriers to FOAD [14:31] (dm doesn't support write barriers.) [14:31] I imagine the same applies to other post-ext filesystems, but I haven't actually had reports of it. [14:32] I guess I'll keep my storage on ZFS until Btrfs shows signs of growing up [14:32] btrfs couldn't even delete subvols prior to .33 and btrfs-tools .19+4 [14:33] I was looking at btrfs a year ago and last summer a post said raid[56] was in the works [14:33] still no cigar, though [14:33] I ended up with opensolaris, which works [14:34] Basically, I am looking for peace of mind if I decide to move my company's server from JFS to EXT4. Has anyone heard of any real disasters using it? I simply like JFS because of my personal history of 0% data loss with it. I am not married to JFS, I simply don't want to be risky. EXT4 could possibly make my life easier since support is built in by default. [14:35] if it's not a truckload of data, moving it should be trivial [14:36] No, a little less than 1T, not much at all yet, but in the next year it will increase alot because of new utilities we will be using. But still, no more than like 4T total. [14:37] * RoyK pats his 30TB opensolaris box [14:38] what is a good webmail client? I currently have postfix+dovecot installed. [14:38] squirrelmail is nice and lightweight [14:38]