From: Louis B. <lou...@hp...> - 2008-01-22 14:46:41
|
Hello, Here is something that I briefly discussed with Bryang and I thought I'd bounce it off all of you. Everyone here is familiar with LinuxCOE main functionality, which is to provide a small .iso image allowing a server to boot off this media and to auto-install according to instructions coming from the System Designer web interface. Now, while keeping the same functionality, let's imagine for a minute that another functionality of SysDes is to provide instructions to a back-end process that is in charge of doing the actual installation into a Virtual Machine container (either Xen, KVM, maybe VMware). For example, you select all the options as we do in the current System Designer, and then you say that you want this O/S instance installed into a KVM disk image. LinuxCOE V5 goes off, provide this information to the back-end process which creates the KVM disk container and kickstart the installation into the VM using the LinuxCOE provided install boot image (the one we currently provide). Once the installation is complete, this back-end process deletes the install boot image and the resulting KVM disk container is what is provided to the user on some FTP server. I had this idea yesterday and didn't have much time to think it over. But after all, this is what FOSS collaboration is all about, right ? Not doing everything by ourselves ? Major restrictions I can see with this : - Needs a powerful server to build all these VM - Needs a lot of diskspace to host VM container even temporarily - Sensible amount of work required to build the back-end process - Best design of this one would require to isolate it from VM technologies (i.e. use of libvirt or something similar) In the end, one would come to System Designer V5, select its O/S and options, and some cycles later, get is VM container, ready to be run. Better yet, this could be an add-on to existing LinuxCOE V5 that would continue offering the provisioning infrastructure as we know it now. Does this make any sense ? Regards, --=20 Louis Bouchard, Linux Support Engineer EMEA Linux Competency Center, Linux Ambassador, HP HP Services 1 Ave du Canada HP France Z.A. de Courtaboeuf lou...@hp... 91 947 Les Ulis http://www.hp.com/go/linux France http://www.hp.com/fr |
From: Bryan G. <Bry...@HP...> - 2008-01-22 15:38:32
|
Louis, I think this is a great idea, and I'd love to see a new backend script to generate such an "image/VM container" and I totally agree with the "libvirt" or equivalent suggestion so that it could be molded to the virtualization technology desired. bryang On Tue, Jan 22, 2008 at 02:46:27PM +0000, Bouchard, Louis wrote: > Hello, > > Here is something that I briefly discussed with Bryang and I thought I'd > bounce it off all of you. > > Everyone here is familiar with LinuxCOE main functionality, which is to > provide a small .iso image allowing a server to boot off this media and > to auto-install according to instructions coming from the System > Designer web interface. > > Now, while keeping the same functionality, let's imagine for a minute > that another functionality of SysDes is to provide instructions to a > back-end process that is in charge of doing the actual installation into > a Virtual Machine container (either Xen, KVM, maybe VMware). > > For example, you select all the options as we do in the current System > Designer, and then you say that you want this O/S instance installed > into a KVM disk image. > > LinuxCOE V5 goes off, provide this information to the back-end process > which creates the KVM disk container and kickstart the installation into > the VM using the LinuxCOE provided install boot image (the one we > currently provide). Once the installation is complete, this back-end > process deletes the install boot image and the resulting KVM disk > container is what is provided to the user on some FTP server. > > I had this idea yesterday and didn't have much time to think it over. > But after all, this is what FOSS collaboration is all about, right ? Not > doing everything by ourselves ? > > Major restrictions I can see with this : > - Needs a powerful server to build all these VM > - Needs a lot of diskspace to host VM container even temporarily > - Sensible amount of work required to build the back-end process > - Best design of this one would require to isolate it from VM > technologies (i.e. use of libvirt or something similar) > > In the end, one would come to System Designer V5, select its O/S and > options, and some cycles later, get is VM container, ready to be run. > Better yet, this could be an add-on to existing LinuxCOE V5 that would > continue offering the provisioning infrastructure as we know it now. > > Does this make any sense ? > > Regards, > -- > Louis Bouchard, Linux Support Engineer > EMEA Linux Competency Center, > Linux Ambassador, HP > > HP Services 1 Ave du Canada > HP France Z.A. de Courtaboeuf > lou...@hp... 91 947 Les Ulis > http://www.hp.com/go/linux France > http://www.hp.com/fr Content-Description: ATT00001.txt > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2008. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ Content-Description: ATT00002.txt > _______________________________________________ > Linuxcoe-devel mailing list > Lin...@li... > https://lists.sourceforge.net/lists/listinfo/linuxcoe-devel |
From: Chris S. <ch...@in...> - 2008-01-22 21:34:50
|
Louis, Sounds cool to me too. Let me know if you can make use of any of the resources at Instalinux. I have lot's of disk space and a fair bit of bandwidth... Chris On Jan 22, 2008 7:35 AM, Bryan Gartner <Bry...@hp...> wrote: > Louis, > > I think this is a great idea, and I'd love to see a new > backend script to generate such an "image/VM container" > and I totally agree with the "libvirt" or equivalent suggestion > so that it could be molded to the virtualization technology > desired. > > bryang > > > On Tue, Jan 22, 2008 at 02:46:27PM +0000, Bouchard, Louis wrote: > > Hello, > > > > Here is something that I briefly discussed with Bryang and I thought I'd > > bounce it off all of you. > > > > Everyone here is familiar with LinuxCOE main functionality, which is to > > provide a small .iso image allowing a server to boot off this media and > > to auto-install according to instructions coming from the System > > Designer web interface. > > > > Now, while keeping the same functionality, let's imagine for a minute > > that another functionality of SysDes is to provide instructions to a > > back-end process that is in charge of doing the actual installation into > > a Virtual Machine container (either Xen, KVM, maybe VMware). > > > > For example, you select all the options as we do in the current System > > Designer, and then you say that you want this O/S instance installed > > into a KVM disk image. > > > > LinuxCOE V5 goes off, provide this information to the back-end process > > which creates the KVM disk container and kickstart the installation into > > the VM using the LinuxCOE provided install boot image (the one we > > currently provide). Once the installation is complete, this back-end > > process deletes the install boot image and the resulting KVM disk > > container is what is provided to the user on some FTP server. > > > > I had this idea yesterday and didn't have much time to think it over. > > But after all, this is what FOSS collaboration is all about, right ? Not > > doing everything by ourselves ? > > > > Major restrictions I can see with this : > > - Needs a powerful server to build all these VM > > - Needs a lot of diskspace to host VM container even temporarily > > - Sensible amount of work required to build the back-end process > > - Best design of this one would require to isolate it from VM > > technologies (i.e. use of libvirt or something similar) > > > > In the end, one would come to System Designer V5, select its O/S and > > options, and some cycles later, get is VM container, ready to be run. > > Better yet, this could be an add-on to existing LinuxCOE V5 that would > > continue offering the provisioning infrastructure as we know it now. > > > > Does this make any sense ? > > > > Regards, > > -- > > Louis Bouchard, Linux Support Engineer > > EMEA Linux Competency Center, > > Linux Ambassador, HP > > > > HP Services 1 Ave du Canada > > HP France Z.A. de Courtaboeuf > > lou...@hp... 91 947 Les Ulis > > http://www.hp.com/go/linux France > > http://www.hp.com/fr > > > Content-Description: ATT00001.txt > > ------------------------------------------------------------------------- > > This SF.net email is sponsored by: Microsoft > > Defy all challenges. Microsoft(R) Visual Studio 2008. > > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > Content-Description: ATT00002.txt > > _______________________________________________ > > Linuxcoe-devel mailing list > > Lin...@li... > > https://lists.sourceforge.net/lists/listinfo/linuxcoe-devel > > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2008. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > Linuxcoe-devel mailing list > Lin...@li... > https://lists.sourceforge.net/lists/listinfo/linuxcoe-devel > |
From: Bruno C. <Bru...@hp...> - 2008-01-23 10:58:55
|
Louis Bouchard said on Tue, Jan 22, 2008 at 03:46:27PM +0100: > For example, you select all the options as we do in the current System > Designer, and then you say that you want this O/S instance installed > into a KVM disk image. I'm doing that right now, but just using LinuxCOE the way it is, downloading the ISO image generated and providing it as an input to my Project-Builder tool, using the newvm option. Details: http://trac.project-builder.org/wiki/NetPerfExample (Look for newvm) > Major restrictions I can see with this : > - Needs a powerful server to build all these VM I do that on a Dual Core2Duo server with 2 GB of RAM. No pb. > - Needs a lot of diskspace to host VM container even temporarily Indeed: du -sh /users/qemu/ 118G /users/qemu/ > - Sensible amount of work required to build the back-end process No !! > - Best design of this one would require to isolate it from VM > technologies (i.e. use of libvirt or something similar) No again. IMO, the best approcah is to use the external commands provided by each tool to handle the VMs. For the moment, I work with QEMU/KVM so that's what is supported, but placeholders exist to add support for other techno (Xen, VMW* if they provide the right tools, I'm less concerned by non Open SOurce solutions myself) > Does this make any sense ? A lot :-) Bruno. --=20 Linux Profession Lead EMEA / Open Source Evangelist \ HP C&I EMEA I= ET http://www.mondorescue.org / HP/Intel Solution Center \ http://hpintelco.n= et Des infos sur Linux? http://www.HyPer-Linux.org http://www.hp.com/lin= ux La musique ancienne? http://www.musique-ancienne.org http://www.medieval.o= rg |