Linux Fatware: Distros That Need To Slim Down

snydeq writes "We need bare-bones Linux distros tailored for virtual machines or at least the option for installs, writes Deep End's Paul Venezia. 'As I prepped a new virtual server template the other day, it occurred to me that we need more virtualization-specific Linux distributions or at least specific VM-only options when performing an install. A few distros take steps in this direction, such as Ubuntu and OEL jeOS (just enough OS), but they're not necessarily tuned for virtual servers. For large installations, the distributions in use are typically highly customized on one side or the other — either built as templates and deployed to VMs, or deployed through the use of silent installers or scripts that install only the bits and pieces required for the job. However, these are all handled as one-offs. They're generally not available or suitable for general use.'"

Really?

[Anonymous Coward]

Got that. It's called Debian Net Install.
Done.

Ubuntu Core

[simonbp]

Ubuntu core distribution is ~34 MB, and available for x86, amd64, and ARM. It's more than suffcient to bootstrap a lean OS.

Re:Ubuntu Core

[ilikenwf]

It's also nonstandard in terms of all the stupid patches and daemons it comes with.

TurnKey Core

[americamatrix]

I always like to use TurnKey Core for such things http://www.turnkeylinux.org/core [turnkeylinux.org]

It's small, lightweight and runs very quickly even on older hardware. It does a great job.


-americamatrix

Archlinux, Slackware, Gentoo

[ilikenwf]

If you really want lightweight and have a specific purpose in mind, just use something that only gives you what you want/need based on what you install. Then, localepurge.

#! Linux

[Tyler R.]

I'm really liking Crunchbang lately! It's very fast, very stable, and it's based on Debian so it works pretty well with mainstream software. It also comes with non free repositories, and codecs.

task-*.rpm

[hduff]

For RPM-based distros, it's easy enough to set up a task-*.rpm to install a minimal subset of the entire repository for a specific purpose, like a LAMP server. I'm sure .deb-based distros have something similar, so I'm really not seeing the problem here, just a lack of understanding the power of FOSS by the OP.

Re:Agree -- issues w/ VirtualBox...

[Sparticus789]

PEBKAC

I have Fedora 18 running in VBox with a Windows 7 host at this exact moment.

Re:RHEL/CENTOS minimal

[BitZtream]

thats the base install? Hell my full Raspian install is smaller than that!

Ubuntu Core is 34MB.

Whats better ... if the submitter of the story had bothered to even google for it ... on the Ubuntu Core page ...

https://wiki.ubuntu.com/Core [ubuntu.com]

About half way done the page, under Deploying Ubuntu Core, it links to the documentation for an x86 VM running ubuntu core ...

Re:Really?

[Freshly Exhumed]

TFA was a complete exercise in BS. Here's another example of how to do a slim Linux install: during a Mageia or Mandriva install, select the Custom option, deselect everything, click through to proceed but when it stops to check if you really, really want to have such a sparse choice select "truly-minimal-install" and you will get exactly what it says, without X or even man pages.

Re:RHEL/CENTOS minimal

[gbjbaanb]

CentOS minimal [mirrorservice.org] is 342Mb, which isn't as small as the Ubuntu, but I guess it comes with more "what you'd install anyway" packages.

There's the netinstall too, which is 230Mb. Nowadays if it can fit on a CD, its considered insignificant in size.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Re:It's already there...

[volkerdi]

2GB for a full Slackware install? Try nearly 8.

And yeah, I'd like to put it on a diet, but once something is already included it becomes quite entrenched. It's extremely difficult to remove anything large enough to make a difference without causing rioting in the streets with torches and pitchforks. I suspect it's the same for any Linux distribution.

Re:Really?

[Anonymous Coward]

A huge reason why is the hodgepodge of INIT you have if you are running 12.04... there is no mechanism to tell whether the packages are you installing use SysV or upstart style-init. You have to go looking for it depending which package you installed.

Other reasons...oh where to start...
1) Your ubuntu-only gnome3 UI? (eg unity). Did you remove it? If so, wtf are you using ubuntu for again?
2) resolvconfd, another ubuntu-introduced joke
3) disparate dependency tracking mechanisms (eg, are you using synaptic? it doesn't play with aptitude's dependency tracking, and vise versa)
4) ufw is a dependency of most network daemons, another huge fail (you imply you are savvy, so you should find ufw particularly offensive)
5) kernel hardening? WTF are you talking about? Look at the 12.04 kernel sources, they are the opposite of hardened. Bonus question: how do you harden those 3rd party binary BLOB drivers?

I could go on...but it's not challenging and it probably isn't even news to you. Great you use ubuntu...but please don't imply it isn't loaded with an above-average amount of crap, whether you tried to strip it down or not.

"Tuned"?

[Guspaz]

What exactly need be "tuned" for virtualization in a VM? I start my VMs with ubuntu-minimal [ubuntu.com], which is pretty darned minimal indeed. I think "eject" is about the only package in there that a VM wouldn't want.

What about PuppyLinux or DamnSmallLinux?

[hillbluffer]

What about PuppyLinux or DamnSmallLinux?

http://puppylinux.org/ [puppylinux.org] http://www.damnsmalllinux.org/ [damnsmalllinux.org]

Both are tiny, and boot in less than a minute.

Re:Really?

[the_B0fh]

Because some of us have used both, and know people who are the release managers for both, and know what kind of shortcuts Ubuntu takes (things that will screw you over).

debian testing is far more stable than ubuntu stable.

Re:Your problem has already been solved.

[zachary.grafton]

I've been using it for almost 6 years as my primary desktop and laptop OS. Never had an issue like that, but then again, I take the time to search for critical bugs before I update, and considering this situation is supposed to be used for multiple VMs, it's not like rolling back to a previous snapshot is hard, minimal testing before deploying is assumed.

Re:Really?

[RevSpaminator]

Ubuntu didn't always suck. I've used it since the Flatulent Badger release and, for years, it was pretty standard Debian with a bunch of stuff preconfigured for new users. Over the last 3 or 4 years I've watched it become more and more "user friendly" and it seems like every release breaks a bunch of things I had manually installed/configured. Now when I go into familiar /etc files I see, more and more, "# Do not edit this file. Some new mysterious daemon will screw up all your hard work." Unity wasn't why I gave up on Ubuntu, but it certainly didn't help. I don't appreciate any GUI that presumes I want to do everything full screen mode. (I could save the cycles and not load any GUI for that.) I've now switched to Arch Linux. I'm learning a lot of things I never had to deal with before. I still don't have it the way I want it, but the rolling releases make it worth the effort. I particularly appreciate the fact that the Arch website regularly notifies users when an update needs special attention or of major architectural changes.

Re:Really?

[kernelpanicked]

I'm not in this fight, as I care for neither Ubuntu nor Debian. However, I have a bone to pick with #2.

No, you don't get security updates until 2017. You get security updates only on packages that Canonical hand picked for that particular release. Hence, your dwm (or really it could be any WM/DE other than Unity) and any other packages that stray from that line, are absolutely left in the cold and unpatched, unloved as soon as next-new-shiny gets released.

Re:Really?

[timmyf2371]

I use Ubuntu Server on my home file/media server and having only used Linux here and there back in the day, I selected Ubuntu because many of the easy to follow tutorials online were written with Ubuntu server in mind.

Re:Really?

[greg1104]

Because there are potentially large performance gains to be had in VMs running postgres by running a 3.x kernel, which CentOS doesn't use yet. Fedora or Ubuntu server are what i'm going to look at to see if it's of use to me.

This is an excellent example of why that sort of thing is avoided by risk-adverse deployments. Early adopters of PostgreSQL on various 3.X kernels are still seeing a variety of nasty kernel issues, and many of them are rolling back to the stable RHEL or Debian kernels based on 2.6.32 to avoid them. A good example is High CPU usage / load average after upgrading to Ubuntu 12.04 [postgresql.org]. I'm tracking about 5 such PostgreSQL issues that only show up in 3.X kernels we're trying to get sorted out still. (I'm a PostgreSQL contributor) Yes, the 3.X kernels are faster in general, but they're still not very stable compared to the boring old ones in CentOS.

Re:Really?

[gagol]

XFS, well balanced, fault tolerant, well tested, and very mature under many loads. Please do not take my words for granted, test it for yourself.

Re:Ubuntu Core

[higuita]

No Linux distro on the planet uses the stock kernel.

Slackware uses stock kernels

All of them have different locations for many different files.

Slackware puts the files where the app developers want to, they dont move files around, breaking stuff (are you listen redhat/fedora!)

All of them have major patchs to all sorts of 'standard' apps.

Slackware tried to used just the upstream code. Only when there are problem reported and there is a fix in the upstream cvs/svn/git, its is ported to the latest release (or the git version is used)

So yes, there are standard, plain and simple distros... slackware is one of the most stable distros there is by not messing all over

Linux's lack of standardization is repeatedly brought up as one of its largest problems in becoming a more common desktop since software vendors don't want to target a bunch of slightly different distro's to pick up a statistically insignificant portion of the population.

Strange, there are things like static binaries, that work EVERYWHERE... you can also ship the libraries, for a pseudo static binary.
But solving that isnt that hard, just have several VMs with the main distros and recompile... yes, its harder than having the source code open and let users/distros developers compile it for you, but that is the price for having closed source.