|
||
|
||
| Mon, 04 Apr 2005 |
|
||
|
I've been doing a lot of thinking lately on what distro to use in the server room. Shortly after RedHat dropped support for it's free distro, we moved to Debian. Debian is nice because things tend to work, and upgrading is minimal. It has it's issues though, Sure, packages are old, but that isn't my big beef with Debian. How many features from new software does one normally run on their server? I know for me it's very little. But there's other issues that have arisen from my usage of Debian. Or perhaps better said, I recognized issues in Debian and other distro's through my usage of Gentoo. Here is an email I recently put together explaining my thoughts on the matter: Now -- all that being said -- and with everything I had mentioned about servers a few years back when RedHat dropped it's free products, we're considering switching the distro on our internal servers. Believe it or not, we're actually seriously considering Gentoo. Gentoo has some really interesting features that make it really nice to work with. It has absolutally nothing to do with speed. This article here gets into some of the positive aspects of Gentoo: "Gentoo for all the Unusual Reasons" http://www.linuxjournal.com/article/7438 The two biggest downsides someone would see to Gentoo on the server are: * Newer (possibly less tested) packages * Required to compile any package you want to install Of course, both of those are issues. It was enough for me to write off Gentoo as a server platform before having actually used it. Now that I have it -- nearly every issue I run into with another distro, I think "Wow, this would be easily solved in Gentoo". I actually feel that there's less issues to be run into with the above two "downsides" in Gentoo, as compared to the issues run into with other distros. Before I get too much further -- I will state that this isn't necessarily true for everyone, or always the best option. I have simply discovered that it seems to work well for me :-) I plan on getting around those two issues with Gentoo using a dedicated build box. The build box would handle compiling all the packages that the other servers might want. It'll store them as binaries (GRP packages). The other servers would then install the binaries, not compile the source. Further, to help the dedicated build box, some of the servers would probably get distcc (a distributed compiler) put on them to help out with the building process. Building packages would only occur at night, so running distcc on things like our nameservers wouldn't be a problem, and would really speed things up. I also would not just install these new packages onto the servers without first testing the packages on the dedicated box. The dedicated box would need to have all the software installed on it that the servers have. A setup like this would also allow us to make use of Hardened Gentoo. Hardened Gentoo has a lot of security tools available to it, including memory and buffer overrun protections, as well as things like SELinux. SELinux isn't for everyone. Even without that, the protections provided by the other tools is excellent. A good many of the remote vulnerabilities we're seeing fall under the categories of being a buffer overflow, or being a poorly coded web app. Hardened Gentoo, even without SELinux, can largely take care of the buffer overflows. Just take care as to what web apps you're running, and what they have rights to. Or run SELinux, and protect everything :-) But there's administrative overhead that comes with SELinux deployments. If you're willing to learn it, and you have the time, it's a useful tool. Two things are coming up -- Debian will have a new release out, and we'll need to upgrade our hardware. As opposed to installing the new Debian on this hardware, it might work out to begin a migration to what I feel is a more flexible Gentoo. |
||
| /Blog/Computers/Organizations/CPLUG | Permanent Link | Comments (4) | ||
|
|
||
| Comments | ||
| sure Posted at Mon Apr 4 23:43:02 2005 by nathan | ||
| I think that's great and will likely be fine considering you are willing to administer it all. Though my only objection is that if you ever decide to leave there, you are really going to put a lot of pressure on the next guy. SELinux and Gentoo are not something the average sysadmin wants to walk into. So, while you have no intentions of leaving now, what about the next few years...just something you may want to ask yourself. Other than that...I'd do it :) | ||
| Posted at Tue Apr 5 15:27:38 2005 by Don Spidell | ||
| Another way to look at what Nathan said... Job Security :-)
| ||
| Not "Job Security" Posted at Wed Jun 1 02:31:30 2005 by Doug | ||
| Unfortunately it's not "job security" but "you're to valuable to promote." All of a sudden you don't get ahead in the company because moving you out of your current position creates too much of a vacuum - not good. | ||
| Job Security? Posted at Wed Jun 1 03:05:36 2005 by Eric Andreychek | ||
| Yeah, I haven't done anything with that... I think I'm leaning towards not touching Gentoo unless another admin comes on board. While I think Gentoo would allow me to do some really neat stuff, and to have an excellent setup, it ultimatly may be more complex than the average admin may want to deal with. Even without the SELinux. The company is small enough that there really isn't a new position to move into, barring management, which I'm not all that interested in. But still, I'm thinking about the next guy. I can handle doing some things in what, IMO, is somewhat suboptimal, to make things easier for the rest of the world in the long run :-) | ||
|
Post a Comment
trackback TrackBack ping me at: http://www.openthought.org/blosxom.cgi/Blog/Computers/Organizations/CPLUG/enterprise_gentoo.trackback |
||
|
Also, be sure to check out the OpenThought Web Application Environment |
|
Copyright 2003 Eric Andreychek |