[09:29] <tepsipakki> I guess this won't be too much offtopic here :) http://apcmag.com/5862/intel_shows_pc_booting_windows_with_uefi_firmware
[09:29] <tepsipakki> forget the windows part
[10:07] <cjwatson> tepsipakki: I do dearly hope that we can forget the DOS partition table format in time
[10:08] <cjwatson> tepsipakki: though I'm worried that we still need the VGA BIOS to initialise many graphics cards for us; wonder if these PCs are any different in that regard than the Intel Macs
[10:11] <tepsipakki> remains to be seen.. bios-compatibility needs to be around for some time :/
[10:12] <tepsipakki> what would be the replacement for DOS partition table?
[10:13] <tepsipakki> I hear that AIX systems have a flexible format
[10:14] <cjwatson> tepsipakki: GPT
[10:14] <tepsipakki> ah, reading ->
[10:14] <cjwatson> ia64 systems and Intel Macs already use it
[10:15] <cjwatson> though for Intel Macs we still have to mirror it to DOS-style ("MBR") partition tables in a slightly crazy way
[10:25] <tepsipakki> hum, the AIX scheme is practically just LVM using the drives as PV:s.. bleh
[10:26] <tepsipakki> so nothing special there
[10:26] <cjwatson> I must admit that AIX's filesystem management is a dream to se
[10:26] <cjwatson> use
[10:27] <cjwatson> it's the only nice thing I ever saw about AIX
[10:27] <tepsipakki> heh
[10:27] <tepsipakki> I like AdvFS of Tru64
[10:27] <tepsipakki> filesets are nice
[10:27] <cjwatson> I didn't know Tru64 had the LVM-like fs grow/shrink stuff
[10:27] <cjwatson> I mean in AIX it's "please give this fs another 10GB" "OK"
[10:28] <tepsipakki> in Tru64 the layering is a bit confusing, since AdvFS has some of the features of LSM
[10:28] <cjwatson> ah
[10:30] <tepsipakki> but you could have partitioned root (root_domain), swap, and the rest (usr_domain, usually). Then create multiple filesets to the domains as needed. They still share the same storage space, but can have quotas
[10:31] <tepsipakki> like, in usr_domain we have var- and tmp-filesets, which have quotas
[10:32] <cjwatson> oh, ok, that's neat
[10:32] <cjwatson> quotas are the main reason to bother with partitioning on a server ...
[10:33] <tepsipakki> true..
[10:39] <tepsipakki> I still have the habit of using multiple partitions on my desktops, but on the most recent installation it only has root, var and home :P
[10:39] <tepsipakki> and a spare one for testing
[12:35] <jetsaredim> evand: ing
[12:35] <jetsaredim> *ping
[12:36] <jetsaredim> or cjwatson if you're around too
[12:37] <jetsaredim> I'm finally getting a chance to go over those notes that evand gave me a few weeks ago
[12:37] <jetsaredim> and I had a quick question
[12:41] <jetsaredim> actually just more questions in general
[02:43] <evand> jetsaredi1: pong
[02:43] <evand> sure thing
[02:44] <evand> though Tuesdays and Thursdays are quite busy, so if I don't respond right away, rest assured I'll respond later in the day
[04:03] <spheard> hi, Ive had my MBR wiped out by bloody vista, Im trying to use $sudo grub-install to reinstate it but its asking for a device sda0 and sda1 dont do it. Im using edgy
[04:04] <cjwatson> 'sudo grub-install hd0' is normally what you want to install on the first drive
[04:04] <cjwatson> or 'sudo grub-install /dev/sda'
[04:04] <cjwatson> (assuming your disk really is sda)
[04:04] <spheard> sata
[04:04] <cjwatson> hd0 is the grub syntax. /dev/sda is the Linux syntax. Either should work
[04:05] <spheard> then grub-install will search all the partitions and list them
[04:05] <spheard> ?
[04:05] <cjwatson> no
[04:05] <spheard> oh
[04:06] <spheard> then I have to put in each line?
[04:06] <cjwatson> what?
[04:06] <cjwatson> just use either of the commands I suggested
[04:06] <spheard> I take it then I will have to tell grub where all my operating systems are
[04:07] <cjwatson> no, your old /boot/grub/menu.lst will still be there
[04:07] <cjwatson> it's not stored in the MBR
[04:07] <cjwatson> grub-install will just switch it back on
[04:09] <spheard> oh of course
[04:09] <spheard> sweet
[04:09] <spheard> so $grub-install /dev/sda should make it all good then
[04:10] <spheard> although the partition with grub on is is sda5
[04:11] <spheard> thnks, Im going to give it a go
[06:28] <wienczny_> Hi I heard that there a plans for a new automatic installer for feisty+1 but I did not find a spec yet. Does a spec already exist?
[06:28] <cjwatson> I think that's overstating the case slightly
[06:28] <cjwatson> you're probably thinking of https://blueprints.launchpad.net/ubuntu/+spec/ubiquity-automation
[06:29] <cjwatson> which is basically adjusting ubiquity to make automation a bit easier
[06:29] <cjwatson> I suspect many people doing automated installs will still want to use d-i though
[06:30] <mark> is there anything new in that respect in feisty?
[06:30] <cjwatson> the main use cases for ubiquity automation are (IMO) people who are already doing customised live CDs and automatic testing
[06:30] <cjwatson> mark: very little, just minor tweaks
[06:31] <mark> ok
[06:31] <cjwatson> partman-auto/disk is optional now if you have only one disk and have partman-auto/method preseeded to regular
[06:31] <cjwatson> that's about the only significant thing
[06:31] <cjwatson> it's been a slow cycle because Debian's been frozen
[06:31] <mark> hmm oh well
[06:31] <mark> at least I won't have many problems adapting to feisty ;)
[06:32] <wienczny_> There are no plans beyond that?
[06:34] <cjwatson> wienczny_: do you have specific requests?
[06:35] <mark> automated LVM/raid would be really nice
[06:35] <cjwatson> right, but that's for d-i
[06:35] <cjwatson> and being worked on in Debian already
[06:35] <mark> ok
[06:35] <wienczny_> Its just that I wrote my own small scripted-installer for debian and ubuntu and I'd like to see a better solution in the distro
[06:36] <cjwatson> wienczny_: what did you find you couldn't do with the existing d-i preseeding facilities?
[06:36] <cjwatson> or with Kickstart in the case of Ubuntu
[06:36] <wienczny_> Installing software-raid. lvm. etc.
[06:36] <cjwatson> ok, then what I said to mark
[06:36] <mark> heh
[06:36] <cjwatson> ubiquity-automation isn't going to help with that
[06:37] <cjwatson> and you don't want a "new automatic installer", you want extensions to the existing one :-)
[06:37] <cjwatson> no point reinventing a perfectly good whell
[06:37] <cjwatson> wheel
[06:37] <mark> no, d-i seems pretty good
[06:37] <wienczny_> I want an installer that is easier to use that can do raid and lvm ;-)
[06:37] <cjwatson> ubiquity doesn't do lvm or raid yet, although it's on a sort of tenuous long-term roadmap
[06:38] <mark> d-i raid/lvm works ok once you learn about its quirks
[06:38] <cjwatson> a lot of the point of doing the new ubiquity advanced partitioner in feisty was to make it more practical to fit advanced partitioning methods like lvm and raid into it in the future
[06:38] <mark> it's a REAL pain on 9600 bps though ;-)
[06:38] <cjwatson> this was basically not practical with gparted/qtparted
[06:39] <cjwatson> once those are fitted into ubiquity and ubiquity-automation is done, it'll be possible to use exactly the same preseeding as becomes available in d-i for LVM and RAID

[06:40] <cjwatson> however, there's enough stuff backed up from previous cycles for ubiquity that I'm not sure when lvm and raid will happen
[06:40] <wienczny_> Will it possible to use ubiquity for automatic network installation?
[06:41] <cjwatson> ubiquity's very much designed for live CDs at the moment, although the LCA2007 guys did work out a hack to get it to install over the network
[06:41] <cjwatson> it would probably be a smallish patch from someone interested
[06:42] <mark> is this for some automated desktop deployment thing? :)
[06:42] <cjwatson> make sure casper's support for netbooting works properly, and teach ubiquity where to get its source filesystem
[06:42] <mark> because if you were installing servers like me, you wouldn't care about ubiquity probably ;)
[06:43] <wienczny_> mmh. My point is this: Ubuntu tries to be a server-distribution (some sort of that). If you've got a look at SuSE or Redhat you will see that its a lot easier to do a unattended automatic installation of complete networks. This is still missing in ubuntu
[06:43] <mark> why would you want ubiquity for servers?
[06:43] <cjwatson> for unattended installations ... what mark said
[06:43] <cjwatson> we already implement compatibility with RH's Kickstart tool
[06:44] <cjwatson> the main pain is that the underlying preseeding facilities for LVM and RAID aren't quite there, but when they are, we can extend that to them
[06:44] <czr> do ks lvm/raid setups work in feisty d-i?
[06:44] <cjwatson> and then I think it's merely a documentation problem
[06:44] <mark> I imagine it hard to explain ;)
[06:45] <cjwatson> czr: no, due to the problem I just mentioned
[06:45] <wienczny_> ^^ Kickstart would be great if I could just take the redhat kickstart docu and do whats documented there ;-(
[06:45] <cjwatson> it's documented as an omission in the installation guide
[06:45] <cjwatson> wienczny_: aside from LVM and RAID, is there anything you can't do?
[06:45] <czr> cjwatson, I thought so, was suprised by your "we already implement compatibility"-thingy. nm :-)
[06:46] <cjwatson> Our installation guide includes errata of what we don't implement relative to RH
[06:46] <czr> are pre and post-scripts supported?
[06:46] <cjwatson> yes
[06:46] <wienczny_> I'm need to doing initial configuration that goes beyond the di
[06:46] <czr> cool. x setups?
[06:46] <cjwatson> they may only be shell scripts
[06:46] <mark> wienczny_: like what?
[06:46] <czr> they mostly are anyway, shell restriction is not a problem
[06:46] <cjwatson> czr: to some extent, yes, though the need for that is gradually going out the window as X gets smarter
[06:47] <cjwatson> we translate xconfig options into xserver-xorg/blah preseeds
[06:47] <wienczny_> deploying exchanging cfengine-keys
[06:47] <cjwatson> wienczny_: post scripts should let you do all of that I'd've thought
[06:47] <mark> that's hardly a d-i thing is it ;)
[06:47] <mark> yeah
[06:48] <wienczny_> You need to get it over a secured channel...
[06:48] <czr> cjwatson, well, I normally only set the resolution through ks, and let it decide and autodetect the rest automatically
[06:48] <mark> I never understand what "secured" means on a newly installing box
[06:48] <czr> haven't really figured out how to do that with d-i yet (worked with edgy though)
[06:49] <cjwatson> xconfig --resolution is translated into xserver-xorg/config/display/modes
[06:49] <wienczny_> Well, I'm currently copying all of those files using ssl and a kerberos-ticket.
[06:49] <cjwatson> in a slightly hackish way although I think the original is hackish to begin wish
[06:49] <cjwatson> with
[06:49] <cjwatson> wienczny_: if you install the relevant packages in %packages, you should be able to do just the same thing with a chrooted %post script
[06:50] <czr> I'll take a stab at creating lvm stuff with d-i next week maybe when I have the time. I need a setup which will install on any hard disk (not depending on the device name).
[06:50] <wienczny_> cjwatson: do you run tests on automatic installtions of complete desktops? I ran into great problems doing that.
[06:50] <czr> and will be testing with feisty, so I guess I'll be counting the installs again. took 82 installs to get edgy in the way I wanted :-)
[06:51] <mark> sounds familiar ;)
[06:51] <cjwatson> wienczny_: not personally yet, though I believe our support team do
[06:51] <cjwatson> I have unit tests of some components
[06:51] <cjwatson> automatic installation is a relatively small part of my job, so I'm quite reliant on people filing bugs about their problems
[06:52] <czr> how many other active hackers are there working on d-i?
[06:52] <czr> d-i in ubuntu I mean
[06:52] <wienczny_> cjwatson: There are some quirks you have to get around. Some basic libs are broken, when trying to preseed them.
[06:53] <czr> wienczny_, in feisty?
[06:53] <cjwatson> czr: fabbione and evand do the odd bits
[06:53] <wienczny_> cjwatson: did not try that in feisty yet.
[06:53] <cjwatson> what *libraries* need to be preseeded?
[06:53] <czr> wienczny_, I have edgy in full automatic install over pxe, no problems here.
[06:54] <wienczny_> my problems came up with libnss-ldap
[06:54] <cjwatson> czr: for d-i I'm increasingly trying to keep in sync with Debian now that they have a pretty decent single-stage configuration
[06:54] <cjwatson> though we still have a non-trivial delta
[06:54] <wienczny_> and more than one installed desktop.
[06:54] <mark> cjwatson: and now d-i is used less in ubuntu anyway, I imagine?
[06:54] <mark> you can do all the customization in ubiquity
[06:54] <cjwatson> ok, I can't be familiar with every possible installation scenario :)
[06:54] <czr> cjwatson, seems sane thing to do. I'm just starting poking around d-i now, there seems to be pretty scarse documentation on it (mostly out of date).
[06:54] <cjwatson> mark: to some extent
[06:55] <cjwatson> certainly there's less pressure for the simplest-possible install
[06:55] <cjwatson> which created some difficult tensions early on
[06:55] <mark> I'd rather want more control / dialogs than less ;)
[06:55] <cjwatson> czr: it's mostly in doc/ in the debian-installer source package, but better to look in svn.debian.org/svn/d-i/ than in Ubuntu
[06:56] <czr> cjwatson, k, thanks
[06:56] <cjwatson> and wiki.debian.org/DebianInstaller
[06:56] <czr> ah, first time I hear about the wiki on it. damn :-)
[06:57] <czr> thanks
[06:59] <wienczny_> I had a look at libnss-ldap sources. They are still as they were for edgy.
[07:00] <cjwatson> although I realise it impinges on automatic installations, I don't consider libnss-ldap part of the installer
[07:00] <cjwatson> it would be more effective to talk to folks who know about that for preseeding improvements
[07:01] <wienczny_> who is that? IMHO the debconf part of this package is terribly broken and I already send a patch to the upstream maintainer to get it fixed.
[07:01] <cjwatson> I do not know; bug reports are as good a way as any to contact maintainers
[07:02] <cjwatson> the upstream maintainer wouldn't typically be involved with debconf; do you mean the Debian maintainer?
[07:02] <cjwatson> I would suggest a Debian bug report
[07:02] <cjwatson> http://www.debian.org/Bugs/Reporting
[07:02] <wienczny_> and then create a link in launchpad?
[07:02] <cjwatson> if you like
[07:03] <cjwatson> though it may just create work since I don't think anyone's particularly actively paying attention to that package in Ubuntu at the moment
[07:05] <wienczny_> This package is just important for those who use ldap-server for authorization
[07:06] <czr> which is I think majority of larger environments where linux is used
[07:06] <wienczny_> ;-) And its in universe -> no support
[07:10] <czr> this is still something which is somewhat lacking in ubuntu. support for larger scale environments (and servers too).
[07:10] <czr> but I guess it will change with time and volunteers
[07:11] <wienczny_> ^^ Thats what I need to do with it.
[07:11] <czr> well, drop me a note when you've fixed all the problems ;-)
[07:12] <wienczny_> *grml*
[07:12] <czr> iscsi/aoe/luks/ldap-tls/nfs4 integration would be nice. and the tools to go with those :-)
[07:13] <czr> ah, and gfx and xen too. maybe I still forgot something :-)
[07:13] <czr> gfs even.
[07:13] <wienczny_> I hate gfs ;-) It just broke out test installation every 24h
[07:14] <czr> heh
[07:14] <wienczny_> I did a package for iscsi an posted it to revu some time ago.
[07:14] <czr> I've yet to set it up properly
[07:14] <czr> although I did try to fix some brain damage in gfs2 kbuilds
[07:15] <wienczny_> We wanted to use cyrus-imap with gfs. The dlm just hang after some time
[07:17] <wienczny_> nfs4 integratoin would be great. some department migrated from nfs3 one month ago.
[07:17] <czr> I've yet to debug it properly
[07:17] <czr> it seems a bit on the complex side with all the gss-stuff leaking into it
[07:18] <wienczny_> They are using opensuse for their desktops though.
[07:19] <wienczny_> One of my favorite ubuntu bugs has been fixed for feisty ;-)
[07:20] <wienczny_> You can use the openafs-version in the repository and get it working with the default ubuntu-kernel. this has never worked since hoary...
[07:23] <czr> eww. afs :-)
[07:24] <wienczny_> ?
[07:28] <czr> it's evil :-)
[07:29] <wienczny_> why is afs evil? I think it should be rewritten from scratch - but the concepts are great
[07:34] <czr> the pag passing implementation at least was pretty evil
[07:35] <czr> it's been some time since I read the codebase, and yes, it definitely should be rewritten
[07:35] <czr> anyhow, this is getting very OT :-)
[07:40] <wienczny_> I'll have to install one of our afs-fileservers now. We should talk about it later. Maybe we should start a group for large scale installations?
[07:46] <wienczny_> bye
[07:46] <czr> wienczny_, would be nice
[07:46] <czr> damn :-)
[07:46] <wienczny_> how could I contact you about that?
[07:49] <czr> wienczny_, privmsg
[07:49] <czr> but note that I'm not affiliated with ubuntu in any way :-)
[07:50] <czr> and also I'm quite busy unfortunately right now, I'd think ubuntu has a global "corporate" team that you might want to join?
[07:50] <wienczny_> I'm not, too
[07:50] <wienczny_> We'll see ;-) bye
[08:48] <spheard> hi, Im trying to reinstate grub aftr a vista reinstall. $sudo grub-install /dev/hda gives me the error :Could not find device for /boot Not found or not a block device. sda is there along with sda1->sda6. anyone any idea whats going on?
[08:49] <czr> spheard, try #ubuntu (this is d-i development channel as the topic says)
[08:51] <spheard> oh I been trying in there
[08:51] <cjwatson> hang on a second
[08:52] <cjwatson> spheard: are you in rescue mode or something?
[08:53] <cjwatson> or chrooted from some other system?
[08:53] <spheard> nope
[08:53] <cjwatson> ls /proc/cmdline
[08:53] <spheard> 6.1 live disk, with pci=nomsi
[08:54] <cjwatson> oh, you need to mount your installed system and chroot into it
[08:54] <spheard> oh okay
[08:54] <spheard> where can I find a howto?
[08:54] <cjwatson> sudo mount /dev/sda1 /mnt; sudo chroot /mnt
[08:54] <cjwatson> (assuming your root filesystem is /dev/sda1)
[08:54] <cjwatson> then grub-install /dev/sda in there
[08:55] <spheard> I tried that, but it said there was no entry in fstab
[08:55] <cjwatson> or in fact https://help.ubuntu.com/community/RecoveringUbuntuAfterInstallingWindows
[09:02] <spheard> thanks man
[09:02] <spheard> Ill give it a go
[10:30] <tepsipakki> czr: what do you think is lacking from nfsv4-integration?
[10:31] <czr> tepsipakki, I don't know, never tried it on ubuntu yet :-)
[10:31] <czr> it was more of a "wishlist of all the things I'd like to get working on ubuntu". not complaints per se.
[10:32] <tepsipakki> I've been using it for a year
[10:32] <tepsipakki> with ubuntu
[10:32] <tepsipakki> and we also do fully automated installations with d-i since two years
[10:33] <tepsipakki> you really don't need to get everything preseedable if you have the infrastructure to pull in a tarball with common customizations and apply it during the install
[10:33] <czr> tepsipakki, how can you easily select the target autopart disk if the same preseed needs to support ide/sata and scsi?
[10:34] <tepsipakki> d-i     partman-auto/disk               string  /dev/sda /dev/sde /dev/hda
[10:34] <tepsipakki> that seems to work :)
[10:34] <czr> and that would select the first one that exists?
[10:34] <tepsipakki> I guess so
[10:34] <czr> interesting, will try. thanks :-)
[10:34] <tepsipakki> we only have one for the desktops
[10:35] <tepsipakki> on some machines the usb-card reader gets /dev/sd{a-d}
[10:36] <czr> yeah
[10:36] <czr> firewire disks also go into /dev/sda. and I guess with libata all the IDE will go there too
[10:36] <czr> joy++.
[10:37] <czr> the only things that aren't going into /dev/sd* so far are the integrated mmc/sd-readers and the floppy drive.
[10:37] <tepsipakki> it's been discussed to release our tools (lkprop: a machine metadata db, altdb/rfist: cfengine/puppet killer :) and make them OSS
[10:38] <czr> well, the problem in this case is that our machine metadata can't be part of the ubuntu-install system
[10:38] <tepsipakki> not all IDE-controllers use libata
[10:38] <tepsipakki> yet anyway
[10:38] <czr> not yet.
[10:38] <czr> it's getting worse all the time.
[10:38] <tepsipakki> heh
[10:39] <tepsipakki> we generate everything from the db
[10:39] <czr> yeah, that would be nice
[10:39] <tepsipakki> from sudoers to nfs-exports
[10:39] <tepsipakki> I mean, everything :)
[10:40] <tepsipakki> which involves object groups etc
[10:40] <tepsipakki> if a computer is added to the db, it's installable in 5min
[10:41] <czr> sounds nice
[10:42] <czr> our environment is a bit different though. we reinstall different setups/configs into the same sets of computers all the time. same set in this case consists of about 30 class rooms with varying different hardware. and obviously we install not just linuxes. mostly windows infact. with about 200 diff setups.
[10:42] <tepsipakki> during that time it has a krb machine principal, DHCP reservations etc. NFS access is the only thing that needs manual intervention, but that's just paranoia
[10:42] <tepsipakki> this system only handles all our unices
[10:42] <czr> yeah, I wish the env I'm working in would be more homogenic, but it's somewhat spaghetti :-)
[10:43] <czr> so I'd just be content of piggybacking d-i ubuntus over pxe in the current system.
[10:43] <czr> although I still need to figure out a scalable way of getting the hostname into d-i so that it will use it for dhcp-requests ;-)
[10:43] <czr> (scalable meaning not having separate pxelinuc.cfgs for each target host)
[10:44] <tepsipakki> the windows guys are always trying to sell their proprietary tools which suck
[10:45] <tepsipakki> hah, we have separate pxeconfigs for every host :)
[10:45] <tepsipakki> because they are generated on the fly
[10:46] <tepsipakki> and because they contain some settings that need to be given on the cmdline
[10:46] <czr> well, I don't want to generate them on the fly. I want one config to represent one target setup (edgy/x86, feisty/x86, etc)
[10:46] <tepsipakki> like the preseed-url
[10:46] <czr> sure
[10:47] <czr> but my preseed is not machine-specific.
[10:47] <czr> so I can pass the "same url" to all the d-is that are installing the same config. only the hostname needs to be different.
[10:47] <czr> but my aims are quite different from what you have.
[10:47] <tepsipakki> hostname and IP here, dhcp is only used for netbooting
[10:48] <tepsipakki> since we rely on fqdn as the hostname, and there is a bug which makes the hostname short even if it is preseeded as full
[10:49] <czr> heh. in our env the hostname appears into dns based on the dhcp hostname in the dhcp request ;-)
[10:49] <czr> it's evil.
[10:49] <tepsipakki> we use dhcp only for laptops, for now
[10:50] <tepsipakki> although there are no linux-laptops installed by this system, only macs
[10:51] <tepsipakki> the nss/ldap-situation on linux is pathetic.. the padl tools should be forked
[10:51] <tepsipakki> and made into something much better
[10:51] <tepsipakki> and integrated
[10:52] <czr> indeed
[10:52] <tepsipakki> a coworker (mac-enthusiast) has toyed with the idea
[10:53] <tepsipakki> he says something like what OSX has would be cool
[10:53] <tepsipakki> a single daemon which handles cacheing, authentication etc
[10:53] <tepsipakki> ncsd is a joke
[10:56] <czr> vintela decided to go with the single daemon solution
[10:56] <czr> they just provide nss hooks to it
[10:56] <czr> and that's the only way to keep the whole linux/ad integration clean
[10:56] <czr> otherwise you end up breaking ad-security or breaking ldap local user security
[10:59] <tepsipakki> you use vintela?
[10:59] <tepsipakki> it's not certified for ubuntu, and that's the only valid reason why the windows-guys are pushing redhat here..
[11:00] <tepsipakki> thank god they are having problems with it, so I can have the 250 desktops on U
[11:00] <tepsipakki> and more coming
[11:01] <tepsipakki> I'd hate to maintain them using a windows-desktop, or via rdesktop
[11:05] <tepsipakki> also it would mean bye-bye NFS and using CIFS
[11:05] <tepsipakki> hell no :)