Living the startup life in Atlanta, GA.

The default Ubuntu install in Slicehost has the minimum number of packages. You have to install most things on your own. The documentation provided by Slicehost is thorough for the most common tasks.  However, I ran into a hiccup that took me a while to figure out.  That hiccup, no surprise, was trying to get the infamous and universally required rmagick gem installed.

My first mistake was probably typing:

$ sudo aptitude install imagemagick

And assuming that it would give me the development headers. It didn’t. Then I tried to find the libmagick10-dev package, which doesn’t exist. Eventually, after much trial, error, and increasing familiarity with aptitude, I found this combination:

$ sudo aptitude install libmagick9-dev
$ sudo gem install rmagick

Success.

In a week that included running 10 kilometers for the first time in over 12 months, getting jabbed with a needle for a blood test, and setting a personal best for the cycle ride to the office, by far the most painful was trying to get RMagick installed on a Centos5 box.

For starters, the RMagick 2.x series requires ImageMagick 6.3.0 (which came out in early 2006). Of course, CentOS5 ships with ImageMagick 6.2.28. I understand the need for stability. But what’s a developer to do if COBOL isn’t a viable solution?

David Bock’s website suggests to try the older version. Which worked, kinda. However, the rmagick build threw up while running sanity checks. The issue is that I didn’t have the Microsoft True Type Fonts installed. Now to find an easy way to install them.

More googleing lead me to Corefonts, which has the installation instructions. I followed these directions and tried again.

Same error.

The finally step is to link the directory with the fonts to where ImageMagick wants them to be.

ln -s /usr/share/fonts/msttcorefonts /usr/share/fonts/default/TrueType

One more gem install and everything is happy.

As rock-solid as Centos is, I’m getting very reluctant to recommend it anymore because of issues like this. Unless you’re running software that’s older or from a commercial vendor, it’s just not worth it. Eventually you’re going to want a library, a major upgrade or a snazzy new language that’s going to make you jump through all sorts of crazy hoops.

One Rails thing that has always felt just a bit un-rails like is deployment. There are a ton of solutions out there. I’ve used Lighttpd, fast-cgi, and mongrel proxied by Apache. Almost all of these have been tedious, a pain, and failure prone.

The Dutch have come to our rescue with Passenger (aka mod_rails). Passenger allows you to run your Rails apps natively within Apache. No need for proxies, mongrel clusters or anything else. They even claim that it recovers from server errors.

The best part is that it’s brain-dead simple. If you can execute a few lines on a command line and add a virtual host to Apache, you can host your Rails app in Passenger.

I upgraded to WordPress 2.5 a while ago and have started drawing some conclusions.  Stuff I like:

• It didn’t break my existing theme
• The overall feel of the admin interface is cleaner
• The visual editor sucks less
•  Built in tag support
• The comment moderation page shows which post the comment is on

Stuff I don’t like as much:

• Aksimet doesn’t have a ‘recheck queue for spam’ button on the comment moderation page
• The save/publish buttons are un-intuitively to the left of the post. (years of web use have taught me that these belong on the bottom)
• Tags and categories are under the post.  While it gives them more space, they become somewhat hidden.
• The checkbox for categories is at the bottom, but it still a scrolling list one-column wide.  If you’re going to move it, at least use the extra horizontal space.

Overall, I like it a lot better, but I think most of that is the colors are softer and the interface is brighter.  For my standard tasks for writing something and marking comments as spam, the experience is only moderately improved.  

I’ve got a few complaints about the image gallery so I spent some time tonight cleaning things up a bit. The the most visible change is replacing the action icons from the Ice Gallery theme with the newly released Ty-cons. They’re a little bigger, a lot cleaner, and make a whole lot more sense than what was there before. They do a great job sprucing the place up. Thanks Tyrus.

Along with the icons, I added some contextual text, made the images look right in IE, and fixed that weird off by 1 px bug in safari. The slideshow should work better now as well. First there’s an icon that reduces the level of mystery meat navigation. I also changed the background from the dark gray to white, and added sensible defaults for image size and time. The moral of that story is changing other people’s Javascript gets a lot easier when that’s what you do for a living.

Last week, I found that I had run out of space on the partitions I use for our virtual machines. This was entirely my fault. When allocated space, it never occurred to me that there might be large log files and I shaved the margin too close. I’ve gotten into the habit of setting up all machines with LVM after the intractable situation I got into with vijedi.net. Which meant that all I had to do to give my VM’s more space was:

1. shut down the VM’s (might have been optional)
2. open the OpenSUSE lvm tool
3. allocate more space
4. hit finish
5. marvel at the fact that I don’t have a space problem

I’ve always heard about the power of LVM. I just never realized how powerful it was.

Companies have been attaching the word “beta” without regard for the meaning of the word. Google is quite possibly the largest offender, affixing the “Beta” tag on software that works, is feature-full, and more production ready than many of their competitors final release offerings.

The VMWare Server team wants to fix the injustice to the word “beta” and bring it back to it’s roots. Therefore, the VMWare Server 2.0 is not very good. They’ve replaced the quite usable and responsive thick client that only works on Windows and Linux, with a buggy, unresponsive, slow-to update, web-client that only works in Windows. Huh? I thought the point of using the web is platform-independence. Apparently, VMWare wants to buck that trend as well.

Then you’ve got the hosting component. One of our VM’s kept crashing. The guest was Centos5. Thinking that it was just the guest os, we downgraded to Centos4, since the VM that is basically used for running queries was pretty stable. Nope, same problem. Apparently, it just has issues with a VM that needs to both read and write from a disk.

In an effort to be a good beta-testing citizen, I filled out a bug report, included some log files and sent it on it’s way. I got an email a few weeks later asking for more information. I dutifully ran the scripts they asked me to run. One of the commands told me to contact VMware support for an FTP account, since the generated information was too large for their webform. I emailed back, and I’m still waiting.

We’ve since downgraded to Server 1.04, in what might be a vain hope that since VMWare knows the traditional definition of “Beta,” they may know the traditional definition of “Production Ready.”

It’s been a while since I installed Zimbra on my mail server and took the opportunity of some free time to perform the update. At first, I was a little apprehensive, since there wasn’t much documentation available. However, I forged ahead, downloaded the new version and ran the install script. When it asked me to update, I said ‘Y’ and walked away from the computer. When I came back, I had an upgraded Zimbra. Apparently, there wasn’t much documentation because there just isn’t much to document.

All in all, the server was down for maybe 5 minutes and my logged in session didn’t even hiccup. It’s always nice when software just works.