blog.erik.rainey.name

I have to say, that was the easiest Typo upgrade I've ever had. Just unziped, relinked, copied the database.yml and migrated. Done!

I just upgraded this blog to Typo-5.1.3. Here's a short description of the two night escapade that was. I install typo (copying the database.yml and other configs) and point the symbolic link I have for the domain to it, everything stops working. Fine, I know they probably updated some tables so I wasn't expecting that to work. So I run:

$ rake db:migrate

Thinking that will fix it. Nope. That doesn't even run. Why? Well there were lots of errors and what-not but after some research and having to read the code, I found out that Typo-5.1.3 thought I had to have a frozen version of Rails in my vendor/rails. However, I just installed it, so of course I don't. So I get rid of vendor/rails:

$ rm -rf vendor/rails

Now I run:

$ rake db:migrate

And failure. After some more reading I learn that Dreamhost decided to update itself to Rails 2.1.0 which is, of course, incompatible with this version of Typo. Great. Now I do have to have a frozen version of Rails. Now I have my own gems on Dreamhost, so I think, no problem, I'll just move it over to that. 

$ rails -v

Rails 2.1.0

Great, it's not even in my gems. So I wget all the Rails 2.0.2 parts (remember the Ruby-Killer on Dreamhost will stop any remote fetches gems!). I install Rails 2.0.2 into my local gems. 

$ rake rails:freeze:gems

Still fails. Great. Turns out Dreamhost installed Passenger for Rais, which is great since it'll speed up loading. However it's also not working with Rails 2.1.0, which is fine I think since I froze it, but it turns out that Passenger ignores your Gems. Awesome. So I try locking the RAILS_GEM_VERISON and the ENV['GEM_PATH'] in environment.rb and then it works. Now, I try to access the new /admin page. That works. Now we migrate. 

$ rake rb:migrate

Failed in Version 71. F**k!  I go to the MySQL Admin page with a righteous fury and open the migrate script up to see what the 071 script does. Looks like it just makes the tag names safe by converting '.' to '-'. Awesome. Now why would that fail? Who knows. I look through tags and I don't even have any with '.' in them. Great. I do have some with ',' in them, so I fix those and move the schema_info table to 71. I run the migrate again and it seems happy since 71 is the latest. Now I load the front page of the blog.

Fail. I'm fairly peeved now. 

I decide to heed previous experience here and go through the Admin pages one by one and which ones fail. Sidebar. That old f**ker. I go into the MySQL Admin and save the old sidebars in a text file and rip them all out. Reload Admin page. Works. Reload blog. Works. Great! Finally. Easy install huh? You just have to hack you database, edit a bunch of config files, and burrow through some code to get it working. At least I can use the /admin page to edit posts again. 

 The blog has been updated to Typo 5.0.3. There were just a handful of issues which had to be addressed from my old 4.1.1 install. 

• cp ../typo-4.1.1/config/database.yml ./config/
• rake db:migrate
• ./script/plugin install http://svn.typosphere.org/typo/plugins/delicious_sidebar
• ./script/plugin install http://svn.typosphere.org/typo/plugin/xbox_sidebar
• rake tmp:clear
• rake tmp:create
• rake sweep_cache
• rake db:sessions:clear
• rake log:clear

 Then I updated the theme to Scribbish and I was done. 

I've been chronicling my mid-adventures in trying to edit my Sidebar (the column you see to the right of this text) over at the Typo Forums for posterity's sake. It would seem that any "stable" release would not have this issue. I'm beginning to think that there's some sort of hidden gem version number dependency that no one mentioned on Typo's site. In the end, it is perfectly possible to hand edit the MySQL data to add back my Sidebar items.

Update: I've found that I can get to a semi-stable point by doing the following:
cd $YOUR_TYPO_DIR
rake -f Rakefile sweep_cache
rake -f Rakefile tmp:clear
killall -USR1 /usr/bin/ruby1.8
killall -USR1 /usr/bin/ruby
Then goto your admin/sidebar interface and add/delete/modify your entry. Hit "Publish" and it should work. However, if you change something and hit it again, it will fail. You'll have to reload the page so that the "Changes Published" (i.e. the flash) is empty.

About a week ago my blog died, as many of you might have noticed. Part of this was due to weird 500 errors that were occurring due my host's policies about renegade processes. Partly they were due to the mysql database not being the proper version for the version of typo I was using, which was also due to the process killer because every time I'd try to run:

rake db:migrate

I'd wait and wait and wait while the process grew past 100MB and then got the axe. I futzed around with re-running the rake task, trying to find the task manually and finally I started to get frustrated with the whole thing and I had decided to move to Mephisto, another Rails app, but it's more of a CMS than just a blog. I might still move, since they have a blog post converter, but it's only somewhat maintained by it's developers at the moment, and I certainly don't have the time to develop it further. Eventually I deleted some of my older typo installs, installed the typo-4.1.1 version, restored the MySQL database (I managed to completely screw up that thing by manually deleting the session table), ran the rake migration command with the exact rakefile I meant to use

rake -f Rakefile db:migrate

(perhaps the others were spidering all the rakefiles it could find in parent and sibling directories), and letting it "sit" for a while (originally after I had done all of that, it was failing to get to the admin console because it couldn't find the blog with "blog_id==4", which I don't have), then I could log in and everything starting working normally. I think the lesson here is to stay on the stable release branch. Having a host with a backup of your databases is really nice too.

Tired of seeing 500 errors in place of this blog? I know I am! I was considering updating to the lastest trunk version until I saw that the Dreamhost Install page at TypoSphere had been updated. It seems Alex Young found a way to get rid of this error by installing some custom signal processing in the Fast CGI rails code, which was originally suggested by Gary Lin. I installed these changes today and so far I've not seen any 500's. Here's hoping that it's a long term solution.