--- Day changed

--- Log opened Thu Mar 17 23:59:03 2005
00:13 -!- ijuz__ [~ijuz@p54ABFE48.dip.t-dialin.net] has joined #xen
00:20 -!- ijuz_ [~ijuz@p54ABBB3C.dip.t-dialin.net] has quit [Ping timeout: 480 seconds]
02:54 -!- sh0n [~sh0n@CPEdeadbeef0000-CM000039d4cc6a.cpe.net.cable.rogers.com] has quit [Remote host closed the connection]
05:21 -!- deac [~deac@xdsl-84-44-148-142.netcologne.de] has joined #xen
07:52 -!- lilo_ [~lilo@lilo.usercloak.oftc.net] has quit [Quit: brb]
08:16 -!- deac [~deac@xdsl-84-44-148-142.netcologne.de] has quit [Ping timeout: 480 seconds]
08:28 -!- deac [~deac@xdsl-195-14-200-192.netcologne.de] has joined #xen
09:38 -!- hollis [~hollis@user-0vvde2g.cable.mindspring.com] has joined #xen
11:13 -!- hollis [~hollis@user-0vvde2g.cable.mindspring.com] has quit [Quit: leaving]
12:50 -!- monrad [~monrad@213083190130.sonofon.dk] has joined #xen
12:54 -!- ShawnWork [~c0db680a@CPEdeadbeef0000-CM000039d4cc6a.cpe.net.cable.rogers.com] has joined #xen
12:54 < ShawnWork> Can Xen use PXE boot?
13:03 < visik7> do u want to boot xen in an harddiskless machine or what ?
13:04 < ShawnWork> well install a distribution over PXE
13:05 < visik7> so you have 10 machines and u want install Xen on it via a PXE ? 
13:06 < visik7> am I right ?
13:06 < ShawnWork> well I first have to figure out how to install it with no distribution
13:06 < visik7> more or less :)
13:06 < ShawnWork> see rocksclusters.org/Rocks
13:06 < ShawnWork> but you can't install via CD-ROM since XEN doesn't emulate a bios bootup :(
13:08 < visik7> mm
13:08 < ShawnWork> but if grub can boot from CD
13:08 < ShawnWork> SYSLINUX or some other boot up 
13:08 < visik7> ok I can't figure out how you want use rocksclusters with xen
13:09 < ShawnWork> well install a frontend machine in one domain and a compute node in another domain
13:09 < visik7> ok so 1 phiscal machine
13:10 < ShawnWork> yes
13:10 < visik7> so just boot a domU kernel with the rootfs of rocksclusters
13:11 < ShawnWork> assuming I have a rootfs :)
13:11 < ShawnWork> Xen can't boot CDs correct?
13:11 < visik7> quite obvious
13:11 < ShawnWork> there is no boot up loader
13:12 < visik7> run rockscluster and copy the entire FS on a loopfs in your HD
13:12 < visik7> but
13:12 < visik7> mmm
13:12 < visik7> let me se
13:12 < ShawnWork> could be done yes dd the disk
13:13 < ShawnWork> not easy to do...
13:13 < visik7> not easy ?
13:13 < visik7> what's the problem ?
13:14 < visik7> rockclusters is a livecd based on what kind of cluster architecture ?
13:14 < visik7> mosix ?
13:14 < visik7> or what ?
13:16 < visik7> or it's a series of userspace tools ?
13:19 < ShawnWork> its a bewulf cluster
13:19 < ShawnWork> not livecd
13:20 < visik7> mmm
13:20 < visik7> ok 
13:20 < visik7> bewulf need a kernel patch am I right ?
13:20 < ShawnWork> no
13:20 < visik7> no ?
13:20 < ShawnWork> no kernel patches needed
13:22 < visik7> IIRC beowulf need applications written for it to run in "cluster mode" am I right?
13:26 < visik7> or not ?
13:31 < visik7> hei are u still there? 
13:32 < ShawnWork> yup
13:32 < ShawnWork> eatting pizza ;)
13:32 < ShawnWork> I Need to figure out how to install from a CD with XEN
13:33 < visik7> anyway
13:34 < visik7> beowulf need applications written for it to run
13:34 < visik7> or not ?
13:34 < ShawnWork> no
13:34 < ShawnWork> im talking MPI stuff
13:34 < visik7> I dunno nothing about beowulf clustering
13:35 < ShawnWork> wait can you load a CD with Xen? if its an ISO image?
13:35 < ShawnWork> mounted as a loop?
13:35 < visik7> probably
13:36 < visik7> dunno if it's the right way
13:36 < visik7> I mean: a CD is an Iso9660 FS 
13:37 < visik7> anyway
13:37 < visik7> try it
13:37 < ShawnWork> yes
13:38 < visik7> where can I get beowulf tools ?
13:38 < ShawnWork> rocksclusters is nice
13:39 < visik7> a part from it
13:40 < ShawnWork> gurb
13:40 < ShawnWork> grub
13:40 < ShawnWork> grub with PXE with xen 
13:47 < visik7> mm I'm lookin' around and see that softwares need to be written for Beowulf Clustering, unlike in openmosix where u can run on it any software that forking itself, MPI is a sort of library to be able to run a software on many nodes 
13:47 < visik7> or not ?
13:50 -!- lutchann [~lutchann@pool-71-240-15-110.pitt.east.verizon.net] has joined #xen
13:51 < ShawnWork> lutchann!
13:52 < lutchann> hiya
13:52 < ShawnWork> whats up
13:52 < lutchann> Just came over to see if anything interesting was happening.  :-)
13:52 < ShawnWork> heh, im trying to figure how to to PXE boot with Xen
13:52 < lutchann> ...
13:52 < ShawnWork> grub with PXE 
13:52 < lutchann> I don't think you can.
13:52 < ShawnWork> so i can do a compute node install
13:53 < ShawnWork> well Xen supports diskless systems
13:53 < lutchann> It does?
13:53 < lutchann> Oh!  You want to BOOT Xen with PXE
13:53 < ShawnWork> people have gotten it to work with pxe-grub
13:53 < ShawnWork> ya
13:53 < ShawnWork> using a CD rom initrd.gz image
13:53 < lutchann> I thought you wanted VMs to boot with PXE.
13:53 < ShawnWork> well, I dont think that can be done :(
13:54 < lutchann> Right
13:54 < lutchann> Although it'd be reeeeeel nice to have GRUB or some such loading the kernel, instead of having dom0 do all the initial image construction.
13:55 < lutchann> Probably it would have to be some totally new kind of "bootloader".  :-)
13:55 < ShawnWork> don't know if we can do that or not
13:55 < lutchann> I just don't like having my dom0 mount the VM's filesystem to get the kernel out.
13:55 < ShawnWork> so we can't somehow add PXE support ?
13:56 < ShawnWork> we don't have a bios to emulate/run
13:56 < lutchann> Well all existing bootloaders expect to run in real mode and talk directly to hardware...
13:56 < ShawnWork> Xen doesnt emulate hardware though
13:57 < lutchann> Right, so I don't think there's much in existing bootloaders that would be useful.
13:58 < ShawnWork> but could it be possible to have a Xen bootloader?
13:58 < visik7> use the iso as a loopback fs and run xen on it
13:58 < ShawnWork> that would install a server yes
13:59 < ShawnWork> but what about compute nodes which need to PXE to get their kernel?
13:59 < ShawnWork> nodes with no grub 
14:05 < schweeb> ShawnWork: only way to find out if it works is to try it
14:06 < ShawnWork> or should I use UML to do that end
14:08 < schweeb> isn't the main Xen kernel (hypervisor?) linux based?  should be able to get it working the same way as a normal kernel... although I've not done it myself... there's lots of howtos online
14:09 < ShawnWork> yes, but a compute node has no OS or kernel 
14:10 < ShawnWork> it has to TFTP/PXE boot 
14:10 < schweeb> right...
14:10 < ShawnWork> and there is no bios 
14:10 < schweeb> you have to do something to a kernel to make it PXE bootable
14:10 < ShawnWork> but the chicken and the egg
14:10 < schweeb> what exactly do you mean by "compute node"?
14:10 < ShawnWork> a empty computer that gets installed from another machine
14:11 < ShawnWork> whos purpose is to compute jobs
14:11 < schweeb> chicken and egg?  you set up a PXE server, and PXE enable the kernel... then set up the client to boot from it
14:11 < schweeb> no chicken and egg about it
14:11 < ShawnWork> how will he client get a kernel?
14:12 < ShawnWork> there will be two domains 
14:12 < schweeb> PXE (tftp?) sends the kernel
14:12 < ShawnWork> on for a frontend which has dhcp and TFTP listening
14:12 < ShawnWork> the second is a empty disk that needs to have emulated network PXE 
14:13 < schweeb> ?
14:13 < ShawnWork> think: diskless
14:13 < schweeb> right
14:13 < schweeb> which is what PXE does for you...
14:14 < schweeb> PXE downloads a kernel image to RAM and executes it
14:14 < ShawnWork> but you need a network card
14:14 < ShawnWork> and Xen doesnt have a network 
14:14 < ShawnWork> there is no bios to boot
14:14 < schweeb> are you trying to PXE boot from the same machine?
14:14 < ShawnWork> yes
14:15 < schweeb> why?
14:15 < schweeb> that makes no sense
14:15 < ShawnWork> because i am testnig frontend to compute node installation
14:15 < ShawnWork> Host { Frontend (PXE, TFTP), Compute node}
14:15 < schweeb> you could accomplish the same thing with NFS root
14:16 < ShawnWork> see: www.rocksclusters.org/Rocks/
14:16 < schweeb> all PXE does for you is grabs the kernel... after that, the kernel takes over
14:16 < ShawnWork> but what will do the PXE
14:16 < schweeb> you need real hardware for that
14:16 < ShawnWork> but the real hardware is in use by the host
14:17 < schweeb> if you need to test the PXE portion, use VMWare or something
14:17 < ShawnWork> I did use Vmware
14:17 < schweeb> I think you're overthinking things
14:17 < ShawnWork> I was hoping Xen could let me do something like this yet
14:17 < ShawnWork> perhaps UML I should try
14:17 < schweeb> UML won't do PXE either...
14:18 < ShawnWork> argh
14:18 < schweeb> you need a complete x86 virtualized system, that handles PXE... VMWare... QEmu or BOCHS might do it, dunno
14:20 < ShawnWork> so we cant emulate a PXE boot setup in Xen 
14:20 < schweeb> it may be possible, but not that I could think of
14:20 < ShawnWork> where is the xen hypervisor itself could do PXE
14:20 < plars> ok, I've managed to do this before in a stupid way, but what's the "right" way to make xen use eth1 in dom0?
14:20 < schweeb> you'd need a userspace bootloader
14:21 < schweeb> I think
14:21 < plars> if I don't do anything, I just get xen-br0 and my guests don't work
14:21 < ShawnWork> userspace bootloader 
14:21 < ShawnWork> hmm interesting
14:22 < ShawnWork> well Kexec
14:23 < visik7> ShawnWork for nodes don't need pxe to get their kernel
14:23 < schweeb> which is kinda what I was trying to say
14:24 < visik7> as soon as you run a domU the kernel is there
14:24 < visik7> and u just need the rootfs that u can get from nfs
14:24 < ShawnWork> but you miss something
14:24 < schweeb> which is what I said :D
14:24 < visik7> ShawnWork what ?
14:24 < ShawnWork> the kernel is there, but what about the domU's IP and the initrd.gz ?
14:25 < visik7> u can pass parameter to the kernel boot
14:25 < visik7> like initrd
14:25 < ShawnWork> what parameter?
14:25 < schweeb> the xen utils take care of IPs and initrd type stuff
14:25 < ShawnWork> we dont know anything at that stage
14:26 < ShawnWork> do you know how Red Hat kickstart works?
14:26 < visik7> to pass the initrd to a domU configure it
14:26 < schweeb> you do understand that kernels must be modified to work under Xen, right?  so you're not using the exact same kernel you would be using to boot real hardware
14:26 < ShawnWork> yes
14:26 < schweeb> so I don't see your point in testing out the full PXE process in the first place then
14:26 < ShawnWork> but dom0 and domU need to send data to each other over network
14:26 < schweeb> root-NFS does the exact same thing
14:27 < ShawnWork> you cannot use root-NFS
14:27 < schweeb> except the xenU kernel is started from xen0, rather than PXE downloading it on real hardware
14:27 < schweeb> why the hell not?
14:27 < visik7> ShawnWork if u want andvice run a normal domU play with it and when u understand how the whole things work u'll be able to solve this problem 
14:27 < ShawnWork> because Red Hat kickstart wont work
14:28 < schweeb> xen exposes a network interface to the kernel, so it can mount the fs...
14:28 < schweeb> I've not used kickstart, but I'm pretty sure you're mistaken
14:29 < visik7> ShawnWork do u have a Xen Box up and running ?
14:29 < ShawnWork> not yet
14:29 < visik7> so bring it up play with it and than everything will be clear
14:29 < ShawnWork> compiling it now
14:29 < schweeb> you should try A) Xen, and B) LTSP
14:29 < ShawnWork> LTSP with xen?
14:29 < schweeb> no
14:29 < schweeb> separately
14:30 < ShawnWork> I'll try Xen first
14:30 < schweeb> then you'll understand more completely how Xen's virtualization works, and how PXE and root-nfs work
14:30 -!- ShawnWork [~c0db680a@CPEdeadbeef0000-CM000039d4cc6a.cpe.net.cable.rogers.com] has quit [Quit: CGI:IRC]
14:30 -!- ShawnWork [~c0db680a@CPEdeadbeef0000-CM000039d4cc6a.cpe.net.cable.rogers.com] has joined #xen
14:31 < ShawnWork> oops
14:31 < visik7> kernel oops ? :)
14:31 < ShawnWork> cgi:irc 
14:31 < schweeb> redhat's kickstart should just be an executable... so you should be able to mount the fs via nfs (like kickstart prolly does)  and execute their binary
14:32 < ShawnWork> another app stole the webpage
14:32 < ShawnWork> I need to see that first
14:32 < ShawnWork> a compute node in Rocks is a dummy PC
14:32 < ShawnWork> no partition nothing
14:32 < schweeb> and isn't kickstart just their system to populate the installer with answers?  I didnt' think it had anything to do with PXE
14:33 < schweeb> you could /run/ it over PXE, but I didn't think PXE was a part of it
14:33 < ShawnWork> when you boot it, you use PXE to get an IP, kernel and initrd.gz which then installs the OS on to the system remotely.
14:34 < schweeb> http://www.redhat.com/docs/manuals/linux/RHL-7.3-Manual/custom-guide/ch-kickstart2.html
14:34 < schweeb> that page says differently
14:34 < ShawnWork> this is modifed ;)
14:34 < ShawnWork> see www.rocksclusters.org/Rocks/ Manual
14:38 < ShawnWork> http://syslinux.zytor.com/archives/2004-July/003813.html
14:38 < ShawnWork> [syslinux] Loading Xen with PXELinux
14:39 < ShawnWork> UNDI driver?
14:46 < schweeb> it's just as I thought
14:47 < schweeb> it doesn't load the kickstart file until after the kernel is loaded from PXE/CD/Floppy-PXE
14:47 < schweeb> it's the same as RH's regular kickstart...
14:49 < ShawnWork> so I can use Xen?
14:50 < schweeb> yea, should be able to
14:51 < ShawnWork> now i need to figure out how
14:51 < schweeb> I mean, you'll have to do a bit of work to get it working
14:51 < schweeb> but it'll work
14:51 < schweeb> well
14:51 < schweeb> I can't say that for certain
14:52 < schweeb> cause their kernel is likely modified
14:52 < schweeb> most clustering kernels are afaik
14:52 -!- deac [~deac@xdsl-195-14-200-192.netcologne.de] has quit [Quit: Leaving]
14:52 < schweeb> so you'll have to see if rocks+xen patches play well together
15:12 < ShawnWork> any GUI frontends for xen?
15:15 < schweeb> there's a web interface
15:37 -!- hollis [~hollis@pixpat.austin.ibm.com] has joined #xen
15:41 -!- deac [~deac@xdsl-213-196-206-164.netcologne.de] has joined #xen
15:47 < ijuz__> ShawnWork: did you allready try with grub? imo it should just work
15:49 -!- ijuz__ is now known as ijuz
15:54 < schweeb> ijuz: he's trying to boot guests via PXE, which makes not so much sense
15:59 < atari> hm...
15:59 < atari> is it possible to attach xen directly to an interface without a bridge?
16:00 < ijuz> oh, guests
16:00 < ijuz> ShawnWork: you have to port grub to xen :)
16:02 < schweeb> or some form of PXE bootloader
16:34 -!- anticw is now known as cw
16:35 < cw> booting Xen guests via PXE doesnt make much sense
16:35 < cw> im not even sure how/why you would go about this
16:35 < ijuz> well, this way everything could come from a central server...
16:35 < ijuz> on the other side you can also get the config files from there
16:36 < cw> none of that requires PXE
16:36 < cw> we are talking about booting domU PXE not dom0 right?
16:37 < ijuz> i think so
16:37 < cw> i really have no idea why anyone would want to do that, it's overly complicated and not needed
16:37 < cw> once dom0 is up you can boot domU's in an manner of sane ways
16:51 < atari> hm...
16:56 < ShawnWork> ugh
16:56 < ShawnWork> Xen sorta works in 2.6.11.4
16:56 < ShawnWork> however, i cant get console to any domain
16:56 < ShawnWork> ;-)
16:57 < ShawnWork> Im gonig to wait until Xen goes into 2.6 mainline
16:57 < ijuz> ShawnWork: the unstable xen seems to have a 2.6.11 patch
16:58 < ShawnWork> hmm
16:58 < ShawnWork> I wish a XEN howto was out there for 2.0
16:59 < ShawnWork> its confusing ;)
16:59 < ijuz> you could write it! :)
16:59 < ijuz> actually i can't get network working
16:59 < ShawnWork> neither
17:00 < ShawnWork> i can get a dhcp lease
17:00 < ShawnWork> but i can't ping anything from the host
17:00 < ShawnWork> cw: a usermode bootloader?
17:00 < ijuz> ShawnWork: like on my box
17:01 < ShawnWork> cw: well im trying to debug a cluster 
17:01 < ShawnWork> rocksclusters.org/Rocks is a Linux distribution (Red Hat) that builds a cluster network for HPC 
17:01 < ShawnWork> problem is, it uses PXE to bootstrap compute nodes from a frontent/central server
17:02 < ShawnWork> sure, vmware works fine but slow dogs
17:02 < ShawnWork> UML cant do PXE booting so thats out
17:02 < ijuz> anyway, if it uses PXE there is also a kernel and a initrd
17:02 < ShawnWork> yeah but you need a way to get the initrd
17:03 < ijuz> is it dynamically created?
17:03 < ShawnWork> no
17:04 < ShawnWork> its in /tftplinux directory on a frontend/central server
17:04 < ijuz> so put it there where the kernel is
17:04 < ShawnWork> put it where though?
17:04 < ShawnWork> in my xen config file?
17:05 < ijuz> ... probably
17:05 < ShawnWork> I can boot from a CD ISO no?
17:05 < ShawnWork> make it a 'disk'
17:06 < ijuz> yes, look at the template in /etc/xen
17:06 < ijuz> it can use a initrd
17:06 < ShawnWork> so I'll extract the initrd.gz from the CD
17:06 < ijuz> booting from iso doesn't help so much, because you need a special kernel
17:07 < ShawnWork> hmmm
17:10 < ijuz> but first you have to get networking working, otherwise it's useless anyway
17:12 < ShawnWork> let's see if the CD's initrd works
17:12 < ShawnWork> it ignored it?
17:13 < ShawnWork> oh
17:13 < ShawnWork> i didnt boot it yet
17:14 < cw> ShawnWork: i dont know about rockslinux
17:15 < cw> ShawnWork: basically inside dom0 get the conf/details to boot the domUs and poke them into an initramfs image or whatever and boot domu that way i guess
17:15 < cw> ShawnWork: i assume these are all diskless?
17:16 < ShawnWork> disk
17:16 < ShawnWork> but initially a empty disk
17:16 < ijuz> cw: is there any trick to get networking working? i followed the user manual step by step to set up a debian virtual machine... and network doesn't work
17:16 < cw> so you want to boot dom0 or using PXE then?
17:17 < ShawnWork> dom0 is the host correct?
17:17 < cw> ShawnWork: yes
17:17 < ShawnWork> dom0 wont have DHCP/PXE booting 
17:17 < ShawnWork> its the frontend/compute node inside dom0 that will
17:17 < ShawnWork> so its a small isolated cluster inside a host
17:17 < cw> ShawnWork: well, dom0 boots all other domains, so that should be easy
17:18 < cw> ShawnWork: you're making a cluster inside a single machine?
17:18 < ShawnWork> yes
17:18 < ShawnWork> testing frontend/compute node
17:18 < ShawnWork> as vmware can do currently, but slow
17:19 < ShawnWork> the problem is if Xen can do this or not (yet)
17:19 < cw> ShawnWork: do 'what' exactly?
17:19 < ShawnWork> allow me to:
17:19 < ShawnWork> 1) install from a CD ISO image a new distribution to a flat disk file image
17:20 < ShawnWork> 2) boot that new distribution and have it become a central install server (kickstart server)
17:20 < ShawnWork> 3) bootstrap a compute node (another domain) from the central install server using PXE or some mechanism to get the initrd.gz image across
17:21 < ShawnWork> I dont think xen can do it nor is designed to 
17:21 < cw> (1) is booting which domain? dom0?
17:21 < ShawnWork> dom1
17:21 < ShawnWork> the host is debian 
17:22 < ShawnWork> we dont touch the host (dom0)
17:22 < ShawnWork> host [ frontend | compute ]
17:22 < cw> ok, so dom0 boots of disk, dom1 ideally from an ISO image, dom2...n over the network from dom1 ?
17:22 < ShawnWork> yes
17:22 < ShawnWork> bingo
17:22 < cw> shouldn't be a problem
17:22 < ShawnWork> i dont know if Xen can do this?
17:23 < cw> yes, but not exactly the way you want to
17:23 < ShawnWork> hehe
17:23 < ShawnWork> <- confused
17:23 < cw> in fact, using Xen for this should be easier than mucking about other ways
17:24 < cw> boot dom0 from disk, for dom1 it depends if it will boot/install into domU/xen cleanly, if not, you will need to fix the distro
17:24 < ShawnWork> seems harder 
17:24 < cw> once you've done that you can strap dom1 from dom0 and for dom2...n use initramfs and nfsroot/whatever from dom1 is you like
17:24 < ShawnWork> well, you cant install a distribution with an ISO image in dom1?
17:24 < cw> nope
17:24 < cw> well, probably not... it's not desirable either
17:25 < ShawnWork> since we dont have a bios couldn't xen virtualize a userspace bootloader?
17:25 < ShawnWork> so grub or syslinux could do the magic?
17:25 < cw> you need a modified kernel to run inside Xen which the distro probably doesnt have and you need a userland which doesn't have lots of fucked up assumptions (FC has lots of weird crap that makes installing inside Xen a pain)
17:25 < ShawnWork> since xen runs before a kernel is loaded
17:25 < cw> no
17:25 < cw> you dont need grub or a bootloader
17:26 < cw> you have dom0 ... it can construct a kernel image and fs and boot that as another domain
17:26 < cw> xen runs before the OSs yes, under them in fact
17:26 < ShawnWork> right, but with xen can you install a fresh distribution
17:26 < ShawnWork> ro you need an existing one to create a dom1
17:26 < ShawnWork> or do you
17:27 < ijuz> ShawnWork: are you about doing an test setup or a generel new way for installing clusters?
17:27 < ShawnWork> testing setups 
17:27 < cw> ShawnWork: either, but again, most distros out of the box wont install under Xen
17:27 < ShawnWork> so i can debug the cluster
17:27 < cw> ShawnWork: so you need to fix that first
17:28 < ShawnWork> something tells me this will not be worth doing for a few years
17:28 < cw> ShawnWork: also, why *install* into a domain?  why not just have a network image/fs you clone and use that everywhere?
17:28 < ShawnWork> because the image you install i dynamic
17:28 < cw> ShawnWork: seriously, this is gonna be easier to do under Xen than just about anything .... the problem is you are trying it make it work with ISO CDs and crap which assume "I really have a fucking PC in front of me"
17:28 < cw> and Xen != a PC
17:29 < cw> Xen doesn't emulate a BIOS like VMWARE, it doesn't need to or want to ... it's complexity that offers nothing really
17:29 < ShawnWork> well you cant install the OS without CDs 
17:29 < ShawnWork> i know
17:29 -!- plars [~plars@pixpat.austin.ibm.com] has quit [Quit: ran screaming into the woods]
17:29 < cw> you can install an OS w/o CDs
17:30 < cw> ive not done a CD install for years for any of my machines, x86, ia64, xscale or ppc
17:30 < ShawnWork> this one needs CDs or ISOs
17:30 < cw> no it doesnt
17:30 < cw> you just think it does
17:30 < cw> the CD is just a collection of files, you can pull it apart
17:30 < ijuz> ShawnWork: sound like Scali... ;)
17:30 < ShawnWork> how do I insert a 'roll' then?
17:30 < cw> what's a roll?
17:30 < ShawnWork> Scali?
17:30 < ShawnWork> an ISO image with RPMs inside it
17:31 < ijuz> ShawnWork: just kidding, a shitty cluster linux
17:31 < cw> ShawnWork: you dont, that paradigm is for installing by switching CDs
17:31 < cw> why would you wnat to do that when you can have all the files on a network somewhere and crap those as needed?
17:31 < ShawnWork> well, with a central install you dont need to switch CDs
17:32 < ShawnWork> it will send the RPMs over a network
17:32 < cw> do you know what a CD is?
17:32 < ShawnWork> an ISO 9660 image 
17:32 < cw> it's a hunk of plastic with bits on it, you can put those bits on a hard disk or whatever too you know and crap them any number of ways
17:32 < ShawnWork> yes yes
17:32 < cw> anyhow, rockslinux im 99% sure wont work out of the box with Xen
17:32 < cw> you need to fix that first or you another distro
17:33 < ShawnWork> I know, but Xen is confusing to me since its a different way of looking at machine virtualization
17:33 < cw> stop thinking vmware
17:33 < cw> vmware is a gross ugly hack that windows lusers like
17:33  * cw tries to insult as many people as possible
17:33 < ijuz> vmware is binary patching?
17:34 < cw> nfi, probably that and more
17:34 < ShawnWork> yes, but its the only thing that works for testing :(
17:34 < cw> actually, vmware is pretty clever for what it does ... but that said it's not a useful paradigm really unless you want to run windows under linux
17:34 < ijuz> so use a stack of machines for testing
17:34 < ijuz> or probably vmware esx/gsx
17:34 < ShawnWork> UML also doesn't help me
17:35 < ShawnWork> it cant do PXE booting either
17:35 < cw> ShawnWork: as far as linux is concerned, it does care where the kernel comes from so long as it can access the root filesystem
17:35 < ShawnWork> right
17:35 < cw> ShawnWork: so with Xen you make a suitable rootfs (or method to get to one) and boot other domains with that
17:35 < ijuz> ShawnWork: you just can't test the installation, but isn't that a minor part?
17:36 < cw> ShawnWork: UML isn't useful either
17:36 < ShawnWork> actually, testing Anaconda is a big part ;)
17:36 < cw> ShawnWork: i really think im failing to explain this, because this sounds *way* more complicate than it should be
17:36 < cw> bbiab
17:36 < ShawnWork> its complicated :)
17:37 < ijuz> ShawnWork: aren't 2 or 3 real boxes enough for testing the installation?
17:38 -!- hollis [~hollis@pixpat.austin.ibm.com] has quit [Quit: leaving]
17:39 < ShawnWork> I have 3 boxes, but @ home they run Debian/XP/LFS
17:39 < ShawnWork> I dont wanna wipe them out ;)
17:39 < ShawnWork> vmware is total crap in Linux crashing too often
17:39 < ijuz> ebay.ca :)
17:40 < ShawnWork> Xen isn't like CoLinux?
17:40 < ijuz> ShawnWork: what is it using to boot with PXE?
17:40 < ShawnWork> in some aspects?
17:40 < ijuz> isn't CoLinux just a UML port for windows?
17:40 < ShawnWork> oh
17:47 < ShawnWork> oh
17:47 < ShawnWork> etherboot?
17:48 < ijuz> netboot? :)
17:48 < ShawnWork> netboot?
17:49 < ijuz> another netbooting thingy
17:49 < ShawnWork> could that work?
17:49 < ijuz> for xen?
17:49 < ijuz> no
17:51 < ShawnWork> interestnig
17:51 < ShawnWork> someone wants to add etherboot support to xen
17:52 < ijuz> hm... ok, great so you can do a kernel image with mknbin
17:52 < ShawnWork> mknbin
17:52 < ShawnWork> ?
17:53 < ijuz> did you ever use ethetboot?
17:53 < cw> ShawnWork: xen is like xen
17:53 < cw> ShawnWork: it's similar to vmware and uml in some ways, and different in others
17:53 < ShawnWork> never used etherboot
17:53 < cw> ShawnWork: xen is probably more like the VMM used on zSeries and POWER5 in some ways than either vmware or uml
17:53 < ijuz> ShawnWork: you have to give the kernel a special blessing with mknbin or something like that
17:54 < cw> ShawnWork: who wants to add etherboot support? i dont see that as useful or desirable at all
17:55 < ShawnWork> I'd like to be able to boot a virtual machine with another one
17:55 < ShawnWork> 'boot' i mean install inside itself
17:56 < cw> you can do that now
17:56 < cw> that's exactly how xen works
17:57 < ShawnWork> but from inside a domain it expects to see the outside as a PC though
17:58 < cw> no it doesnt
17:58 < cw> well, define 'it' i guess
17:58 < ijuz> now irc would need a whiteboard
17:59 < cw> there are a couple of options, i tried to find some that worked well but alas they all sorta suck
18:00 < cw> i think there is considerable confusion as to what "linux" really needs and expects to see
18:03 < cw> ShawnWork: why do you have to install from CD?
18:05 < ijuz> the interface of my domU loks like this (as seen from dom0)
18:05 < ijuz> vif2.0    Link encap:Ethernet  HWaddr FE:FF:FF:FF:FF:FF
18:05 < ijuz>           UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
18:05 < ijuz>           RX packets:0 errors:0 dropped:0 overruns:0 frame:0
18:05 < ijuz>           TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
18:05 < ijuz>           collisions:0 txqueuelen:0
18:05 < ijuz>           RX bytes:0 (0.0 b)  TX bytes:0 (0.0 b)
18:06 < ijuz> from inside the domU:
18:06 < ijuz> XE1:~# ifconfig eth0 up 192.168.3.20
18:06 < ijuz> XE1:~# ifconfig
18:06 < ijuz> eth0      Link encap:Ethernet  HWaddr AA:00:00:1E:9B:03
18:06 < ijuz>           inet addr:192.168.3.20  Bcast:192.168.3.255  Mask:255.255.255.0
18:06 < ijuz>           UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
18:06 < ijuz>           RX packets:0 errors:0 dropped:0 overruns:0 frame:0
18:06 < ijuz>           TX packets:1 errors:0 dropped:0 overruns:0 carrier:0
18:06 < ijuz>           collisions:0 txqueuelen:1000
18:06 < ijuz>           RX bytes:0 (0.0 b)  TX bytes:42 (42.0 b)
18:06 < ijuz> but when i ping from domU anythign nothing comes back or reaches anything, where could be my error?
18:09 < ShawnWork> cw: initially, i need to have the OS somewhere
18:09 < cw> so install it w/o xen
18:09 < ShawnWork> im goin home so im gonna play more, and try qemu w/ etherboot and see what works
18:09 < cw> ijuz: vif2.0's MAC address looks wrong
18:11 < ijuz> ShawnWork: just tar up some install at work
18:12 < ShawnWork> if i could use a vmware disk image 
18:13 < ShawnWork> then i could do this with Xen (using etherboot to boot the node)
18:13 < ijuz> # Optionally define mac and/or bridge for the network interfaces.
18:13 < ijuz> # Random MACs are assigned if not given.
18:13 < ijuz> #if = [ 'mac=aa:00:00:00:00:11, bridge=xen-br0' ]
18:13 < ijuz> if = [ 'mac=aa:00:00:00:00:11, bridge=xen-br0' ]
18:14 < ijuz> buuut python doesn't seem to be happy about the usage of "if"
18:15 < cw> for some recent the python code started blowing up for me unde debian a while ago
18:15 < cw> i need to recheck that
18:15 < cw> ShawnWork: why would you do either of those things?
18:15 < cw> ShawnWork: why not install it and just use tar?
18:16 < ijuz> cw: so i can't set the MAX for vif*.0?
18:16 < ShawnWork> just dump whats in the vmware disk into a tarball?
18:17 < ijuz> yes, we install our cluster with a tar, needs <10 minutes per node
18:17 < ijuz> i did this setup faster than this scali idiot installed his shit on another testcluster :)
18:18 < cw> ShawnWork: what? are we still talking about rockslinux?
18:18 < ShawnWork> ya
18:18 < cw> install it somewhere yes, and then tar it up
18:18 < cw> ideally before you do that, fix it to work under xen
18:18 < cw> that might require several reboot/hack cycles
18:18 < ShawnWork> it runs 2.4 kernel
18:19 < cw> mostly init-script stuff
18:19 < ShawnWork> RHEL 3
18:19 < cw> well, you need to replace the kernel obviously
18:19 < ShawnWork> sure, i can drop in a 2.6 kernel or a 2.4 with xen
18:19 < cw> well, once that's working and all happy as a xen instance, ideally a domU instance tar it up
18:20 < ShawnWork> when I go home in a few mins I'll try
18:20 < cw> and use that as a base to work from
18:20 < ShawnWork> im not sure how the actual installer will like that
18:20 < cw> it wont
18:20 < cw> so you install it on the metal and put xen under it
18:20 < ShawnWork> i'll try this weekend
18:21 < ShawnWork> time to go :)
19:42 -!- sh0n [~sh0n@CPEdeadbeef0000-CM000039d4cc6a.cpe.net.cable.rogers.com] has joined #xen
19:59 -!- sh0n [~sh0n@CPEdeadbeef0000-CM000039d4cc6a.cpe.net.cable.rogers.com] has quit [Remote host closed the connection]
20:59 -!- cw [cw@adsl-67-124-116-142.dsl.snfc21.pacbell.net] has left #xen []
21:00 -!- cw [cw@adsl-67-124-116-142.dsl.snfc21.pacbell.net] has joined #xen
21:00 -!- mode/#xen [+o cw] by ChanServ
21:03 -!- visik7 [~ciao@host155-36.pool80182.interbusiness.it] has quit [Quit: did you grep /dev ?]
22:43 < Beirdo> what's the status of xen for x86_64?
22:47  * mikegrb hides
22:52 < Beirdo> it isn't compiling for me (from schweeb's package)
22:52 < schweeb> Beirdo: I'm guessin the problems are probably fixed in testing or unstable
23:24 -!- deac [~deac@xdsl-213-196-206-164.netcologne.de] has quit [Ping timeout: 480 seconds]
--- Log closed Fri Mar 18 23:59:02 2005