Waiting for a reply :D, Baron.

It has been demonstrated that Gentoo can be run perfectly on lots of servers, since 10 is enough to call it a lot. Does it make news? I guess not.

The next number to consider is 1000, which is about automating tasks like propagating binary packages from a build/testing server to all production ones. People like chort do not speak about that, so they have not reached that level yet.

It has been also perfectly demonstrated that 1) some people cannot run Gentoo as a server distribution, and 2) some people do not need to. Can anyone expect different?

Now, how come that this thread turned to be a bit heated? It often happens when a person who cannot do something gets high tempered when he or she feels that the lacking ability potentially gives some advantage to others. Ask your doctor about details.

Let us go through the chort’s list:

1. Long release cycles to reduce the frequency that a company has to do QA in order to upgrade.

Pipe dream. Vulnerabilities and bugs get known when discovered and no release cycle can be enforced on them.

2. Ultra-stable platform that can run the company’s software with a minimum of testing. This includes that platform being certified to run all the major commercial software that the company has purchased.

Pipe dream. The platform is as good as made by the application developers and no distro can influence them much.

3. Something that “runs itself” and requires little to no on-going maintenance from IT staff.

Dangerous pipe dream. Nothing can “run itself” so far. If something gets close, just consider what happens if the beast misbehaves.

4. Comprehensive and well-implemented security patch process to ease applying patches to a large number of machines (and not too frequently–there’s a reason why Microsoft adopted the “Patch Tuesday” routine).

Pipe dream, except the process. Conflicts with item 1. The spam I get every day from zombies demonstrates that Microsoft is no good with security regardless of the “Patch Tuesday”.

The process to apply patches is a valid requirement, and if there is any one working then it should be ported to Gentoo.

5. Paid technical support from the vendor itself (for accountability and contractual obligations).

Bingo! That is what a “real company” employee always wants: a way to pass responsibility to someone else.

I have to conclude that chort is under more stress from not that much computer competent business people than Josh and part of the stress has beens revealed in this thread.

It has been mentioned that in many companies IT is a liability and I cannot agree more. If it is the case, items 1 through 5 above are to be expected. My 2c are: is IT a liability at Google?

Back to the topic. One distro differs from another in exactly 4 areas: package management, the base layout, software selection, and patches.

I cannot imagine how it is possible to upgrade anything mission critical without etc-update or better. I do not know about equivalents in other distros. My experience with Windows suggests that on any non-trivial update something is either broken or new useful features I want are turned off. It looks like portage is better for servers than other package managers.

Gentoo has simple and easy to maintain base layout. The changes it introduces to the treatment of run levels and environment variables may be beneficial for servers, but just a little bit.

Servers do not need most of the software in the portage tree. Oracle and Beowulf are not there. Gentoo is not a server distribution.

Gentoo never enforces any silly patches (for vendor lock-in, Windows look and feel, etc.) and has its use flags facilitates removing unnecessary and potentially problematic code. This is good for servers.

Final thoughts. Josh has not demonstrated when and how Gentoo is much better as a server distribution than any other. I doubt that somebody will do that ever soon, since it is hard to acheive without revealing the competitive advantage of some company.

Link system not as good as advertised.

Post..take 2.

I love your rebuttal, josh!

Gentoo is also my favorite Linux distribution by far! I’m not sure if this was mentioned in a past comment, but dispatch-conf is a superior tool over etc-update.

I think you might enjoy this: How to be l33t

Pay close attention to step 10 :)

All the best,

jownz

Gentoo is also my favorite Linux distribution by far! I’m not sure if this was mentioned in a past comment, but dispatch-conf is a superior tool over etc-update.

I think you might enjoy this: Pay close attention to step 10 :)

jownz

When new changes come out, I build them on my build server. I can then selectively roll out changes to one server at a time, allowing me to do a fully audited change and test program.

Once that update has been certified, I can then roll out the changes to the rest of the servers in that environment.

By the way, my servers have never gone down because of Gentoo. I have 100% uptime. yes, not 99.999, 100. Literally the only time the servers have gone down is when our hosting provider had major power problems.

My personal server (that runs this website actually), has run RedHat and Fedora in the past. I’ve achieved perfect stability running Gentoo. I can’t say the same for those others.

We are a firm of systems integration and software development.
We run about 450 servers for customers having up to 1Million clients.

Gentoo is a no go.
Reason ? It’s Troublesome. Securitly patches and dependencies have a way to disrupt the ability to compile after a while, requiring way too much time and making the RTO (return to online) longer in case of problems.

We need a stable OS, which when it has to be updated won’t break ANY dependency… ever. The software on it will then have to be certified and supported too. We can’t afford interdependency changes.
Do you have ani idea of a 99.985 uptime means ? We provide that. (%age is monthly on a thwo months period)

The only unixes that have passed the requirements tests and the POCs (Proof of Concept - validation of each piece of software we integrate) are AIX and Solaris.

Red Hat is in the certification process but only for DB and Web servers. (and in these case we use clusters or active standby redundancy)

Gentoo is cool, nerdy, and on my home computers. Not on the ones at work. Simple as it is.

Gentoo has distinct advantages over RedHat, Debian, etc. in certain environments. For example, in an environment where you run open source application services such as LAMP, open source IMAP, etc. In those environments the packages provided by the binary distributions are actually detrimental to stability and manageability because they don’t provide exactly what you need and you must break the native package management system by building your own binaries from source.

Now, obviously, if you’re running Oracle or DB2 some other proprietary vendor supported/qualified product on Linux, then you want RedHat or SUSE because you’re not supporting and qualifying the ‘app’–they are.

The best of all would be if you could have RedHat/CentOS or SUSE with a system like portage for certain packages. So you could have RedHat but still run Cyrus or LAMP on it without major hassles.

/djs