See http://www.vmware.com/vmtn/vm/community.html Maintaining a vmware image could be a useful way to get people into gentoo by providing an easy way to try the distro out. With the recent release of a free VMWare player, more and more people are going to be trying out various OS's with it.
I think it would be a good Idea to do so! Btw: Is there an ebuild around for VMWarePlayer 4 Linux?
Steffen: emerge vmware-player The only issue I see with offering a VMware image is what do we include? What *is* Gentoo? Do we include all of GRP? Gnome? KDE? What about server packages? Do we keep the portage tree? distfiles? Where would we host it?
note that the VMware Technology Network is already making available images of other unix or linux operating systems like fedora, etc. It's frustrating to login the vmtn pages and see fedora, etc listed but not gentoo. We probably should look into what the other vendors are including in their vm's. Also, I'm not sure if we'd be producing a virtual machine that designed to run under just workstation or esx/gsx? I'm sure we could find something that is mostly compatible with either.
Just make sure that you create the virtual disk as a SCSI disk and I think that {E,G}SX can import the image. I would recommend you double check it as it has been a little while since I used those.
OK. Well, let's try to put together a specific plan here. I am willing to help with building and testing an vmware image for gentoo whenever a new release is made. I have to do that anyway to speed up deployment of new virtual machines here. However, I would be building the image primarily for vmware esx. Of course, if someone else from releng wants to handle building images. that's fine. After ESX testing is done, I can also test under workstation and GSX, but my main development/deployment environment is ESX and the resulting image would reflect that. I'd say the vmware images would lag the livecd images by 1-2 weeks generally. On another note, where would the images be hosted? What specifications for the image would we require? Others might not agree with my selections, all though - gentoo makes it trivially easy to change stuff.
Since nobody has answered a single question from my comment #2 I would say that we should *not* be even working on this. We have no indication of what Gentoo is exactly. With Fedora, et al. someone can just do a "full" install and get everything that is Fedora. There is no definition on what Gentoo really is to be able to do this effectively. Do we want a server installation? Desktop? Do we want it to be the same or similar to the LiveCD (not InstallCD)? Were you to ask me, I would say that we absolutely should not rush into providing this, nor should we jump into providing something such as this without information on how this will impact on our infrastructure. We have nowhere to host such a beast currently. I am looking at doing quite a lot with our releases that is applicable to all of our users, whereas these images would be quite large and only useful for a very tiny subset of users in comparison. I also feel that this falls squarely in the realm of Release Engineering and should not be taken by any other project, especially when we are unable to get help in Release Engineering for our actual releases when we ask for it.
Ok. Well, if release engineering would have to manage many aspects of the virtual machine and arent interested in it, than that definitly causes a problem with proceeding. I suppose i could go ahead and create an "unofficial" virtual machine and make it available via my own bandwidth resources (datacenter with rate-limits applied at the router for ftp traffic for the image), but then that is quite unoptimal. Or I could volunteer to help release engineering specifically with this task, if release engineering was agreeable. Assuming that was taken care of, that leaves the questions of: what should be on the virtual machine? Where would it be hosted? Do we keep portage tree? etc? I will make the following recommendations: - Make the virtual machine be as minimal as possible, and where we must include something, attempt to either mimic livecd or go with current recommendations from the releng team. Lets look at the issues one by one: Hardware: Set it with a max 10GB SCSI harddisk, using lsilogic controller, where disk space is allocated in files that can be up to 2GB, and the files/space is allocated only when used. That should mean we have the smallest distribution file. Memory would be set to 512MB as this really is the minimum required for compiling some software. Network card would be standard vlance 100Mbps. There would be no cd/floppy. Users could add it if they want. Network: eth0 via dhcp, otherwise hostname is 'gentoovm' with nothing else specified Disk Partitioning: 512MB swap partition (isnt this the minimum people allocate these days?), remainder as a single rootfs Portage tree: it would contain the same snapshot as the livecd copied over to /usr/portage along with just updates from the initial emerge sync involved with creating the image using standard stage3 install steps. Since the vm would be created rather shortly after release, we should still have a rather good tree. Distfiles: same as portage tree, just what was put on the universal livecd + updates at the time of building profile: default gentoo profile for x86 arch (thats what releng recommends for x86 arch). I dont think its possible to use any arch but x86 at this point since vmware doesnt really support anything but x86 well across its entire product line. make.conf stuff: minimal stuff, mostly copied over from default livecd image, use.defaults also would be minimal and really close to the default gentoo profile packages: just essentially the stage3. I could include whatever syslog/cron stuff is default these days if releng wants. Where possible boot loader would be grub, but I've seen that lilo sometimes works better for handling vmware stuff as grub gets easily confused. Hosting: We can find a way to handle this if its the sole remaining issue. There are alot of technical solutions, and worst case we can rate limit or limit it to bittorent/etc. Now, I know that the resulting cd would be rather minimal - but thats fine, users can customize it however we want. other projects could customize and distribute their own versions. The only issue then would be documentation = I can write a draft up but I dont know how it would get into the handbook. As for user demand for a virtual machine, I think its underestimated. Furthermore, it is only growing exponentially as virtualization becomes mainstream. For all we know, no one will want livecd's 5-10 years from now, they may only want virtual machines (xen/esx replaces pc bios's). I'd like gentoo to be able to support the user base that is leading the way here. I mean, if people dont want to be leading, they generally stick with windows or redhat.
(In reply to comment #6) > ... that is Fedora. There is no definition on what Gentoo really is to > be able to do this effectively. Do we want a server installation? Desktop? what about creating a minimal Image with just the base-system and providing some packages as binary builds. You could offer the user then 3 scripts vor example something like: "install server-packages" "install desktop-packages" "install games" or something similar wich merges the precompiled packages. I think this would show the user how easy and comfortable it is to install software with portage.
(In reply to comment #7) > Ok. Well, if release engineering would have to manage many aspects of the > virtual machine and arent interested in it, than that definitly causes a > problem with proceeding. Here's the problem. We do not have an x86 Release Coordinator. I have asked time and again for assistance with this, and nobody has stepped up. This means that I have to not only do my normal duties in leading Release Engineering, but now also have to buld the entire release for the most widely varied architecture we support. I will not support adding *any* extra release materials of *any* kind so long as this situation does not change. > I suppose i could go ahead and create an "unofficial" virtual machine and make > it available via my own bandwidth resources (datacenter with rate-limits > applied at the router for ftp traffic for the image), but then that is quite > unoptimal. You are more than welcome to do this, but be sure to mention that it is *not* anything official from Gentoo and any bugs regarding it will be immediately closed by Release Engineering. > Or I could volunteer to help release engineering specifically with this task, > if release engineering was agreeable. We would definitely not be, for reasons mentioned above. > As for user demand for a virtual machine, I think its underestimated. In comparison to the CD media, which is where my comparison comes from? Quite frankly, I would not only oppose this, but would actively seek to block it from being done in any manner by anyone until such time as x86 had a Release Coordinator. > Furthermore, it is only growing exponentially as virtualization becomes > mainstream. For all we know, no one will want livecd's 5-10 years from now, > they may only want virtual machines (xen/esx replaces pc bios's). I'd like > gentoo to be able to support the user base that is leading the way here. I > mean, if people dont want to be leading, they generally stick with windows or > redhat. The simple fact that we cannot even provide support for our CD media tells me that the absolute *last* thing that we need to do is to add yet another barely supported piece of release media to our repetoir of not-quite-working-or-supported releases. Perhaps once we get some people on board to actually help with the current official releases, we can revisit this, so I am marking it LATER.
> Here's the problem. We do not have an x86 Release Coordinator. ... > Perhaps once we get some people on board to actually help with the current > official releases, we can revisit this, so I am marking it LATER. OK. Thanks for being clear. I'll do the same. One of the critical advantages of opensource is that developers can scratch their own itch, and because everyone has different itches, a comprehensively good set of software eventually emerges. My itches are in vmware, apache, and enterprise stuff. Unfortunately, I cant really do much in any of those areas because lack of other developers either in the associated herds or issues/lack of manpower in other herds that we would be dependent on. I am willing to assist as much as I can now with vmware herd, learn catalyst/etc to eventually help out releng where desired and possible, etc --- but there isnt really anything I can do to help find the x86 arch coordinator, so I guess this bug might be stalled for quite awhile. Thanks.
Gentoo virtual machine is currently being discussed in the vmware player community forums: http://www.vmware.com/community/thread.jspa?messageID=337745#337745
My opinion: Get a build of Gentoo that "fits like a glove" for a set normal computer (one that has a network, 10gig harddrive, 128 megs of ram, USB, etc.) Get it to work perfectly with the hardware, and then give them a link to the installation guide on how to continue from there. In general I wouldn't add any packages, of course portage and other necisary stuff, but nothing like X or anything like that. That allows the user to customize it for themselves, but keeps the boot times quick.
I was also looking for a gentoo Virtual machine, but since couldnt find any, checked out the forums, which directed me here. The debate here is ragin on "what is gentoo". MY suggestion is create a virtual machine with X, Network and basic stuff like portage and emerge, maybe even give either KDE or Gnome(I would go with KDE). Then after the user installs he/she can simply do emerge and get all the pacakges needed. I am a gentoo user, and I went to XP because I need it for certain tasks. So currently I an a dual booter. With a gentoo virual machine I can easily work on gentoo from inside my XP too.
Hi, I have created a simple (minimal) Gentoo VM based on the inputs that i gathered from here and at the VMware forums. I currently have a torrent hosted at my site and the is information available at http://gentoovm.blogspot.com/ The VM is pretty bare, it basically has the Linux Kernel, bash shell and ethernet netowkring and ofcourse emerge. I believe this will allow the user to customize it for themselves. Thanks!
