2015-01-26 12:09pm Matt Graham wrote:

This sounds like a failure of sysadminning.

Yes, but is it Sys admin or more of a Dev/Ops issue?.

It's fairly easy to generate a kickstart (automated installation) file for RHEL/CentOS, which can install all the ordinary packages from the standard RHEL/CentOS repos. The kickstart file can also run an arbitrary shell script after all the packages are installed, which'd get additional repos and the packages in them and so forth. I did this for the last round of physical server installs; we could get a physical machine from "has no OS" to "running all the stuff" fairly quickly.

Kickstart, (RHEL-based), pre

ed, or preseed plus kickstart (Debain/Ubuntu) for base server installs are a given, especially for installs on physical servers.

These go without saying. Building a 3-tier web, data, file server system, installing and configuring the stack for the specific enterprise application, and then spinning it up all from a versioned source code, IMHO are better handled by provisioners such as Vagrant, Packer and other configuration managers.

Or was this more like "we didn't think about anything but Ubuntu until the last minute"?

It was actually the opposite. The team struggled to complete the upgrades to extensions, modules, new applications, ini/config changes etc and migrating to the existing RHEL environment. As an independent test case the same application stack and environment was built using Ubuntu 12.04 LTS Servers. It was completely rebuilt, destroyed, re-configured, and re-built 3 times in the same time frame.

Some things that can be useful (imagemagick, for one) can have dependencies on X11 libraries. Some things like LibreOffice require a running X server but can also be used to do things like convert PDFs to MSWord in an automated fashion. (The guy I was talking to who needed to do that used Xvfb to fake an X server....)

I use imagemagick on the web server to process images (in Drupal). Installation of the X11 libraries and the php extensions are done by script through the provisioner. What I meant is that they do not require that a full desktop environment be installed on the server to manually access the server and use this feature. However perhaps I was unclear, I was referring to the comment regarding bloat. I assume that the bloat referred to in previous comments was due to the "belief" that Ubuntu server had to, a) be installed graphically, and b) that it always installs "bloat" or extra processes that need to be removed or shut down.

Did you mean "human labor"? Installing junk on a server feels a lot more like labor than capital to me if you go by economic terms. (Buzzwords don't tend to respect things like dictionaries or previous word usages though.)--

No, I meant capital (I am an Economics graduate). A corporations human capital in micro economic terms is a function of the value of relevant knowledge and experience in aggregate of its human resources (employees). The cost of human capital (in one form) is inversely proportional to the value of a process that can replace it (the human). The more processes that require less expertise but produce greater returns on investment that company "A" has as compared to company "B"; the smaller the human capital cost of company "A". While human capital is considered a company's most valuable resource it is still by far the single biggest cost per unit value of any of its resources. In other words faster/easier server implementation with fewer needed human experts to build and maintain them equals potential lower human capital costs.

Not sure what you mean by "junk". I am commenting based on 2 premises: 1) The server build/examples I have given have been made production ready, and 2) Ubuntu server is equal to the task of being a production web server.

James

Linkedin