Dylan Lloyd

2011/03/25/migrating domains

I set up dylanstestserver.com as a test domain. It's a pain to type out, so I planned on switching to dylansserver.com. By the time I got around to this though, the first domain has already been indexed by Google. My deployment process was also engrained with the original domain. This is how I made the migration without losing standing in search engines.

First I added a new virtualhost to Apache's httpd.conf.

I then checked out a new copy of the live branch into a new folder in the webroot.

Next I added a new A record to my DNS, pointing from the new domain to my elastic IP.

Then I restarted Apache and checked that the new address worked. Then I replaced the .htaccess in the original DocumentRoot with a new directive.

v .htaccess
ggcG
RewriteEngine On
RewriteCond %{HTTP_HOST} !^www\.dylansserver\.com$
RewriteRule (.*) http://www.dylansserver.com/$1 [R=301,L]

I moved the repository to it's new folder and changed it's post-recieve hook to reflect the new working directory. I also needed to edit my local gitosis-admin/gitosis.conf (and push the changes) to grant myself access to the new repository. Finally, I updated my local repository to point to the new destination.

git remote set-url origin git@dylansserver.com:dylansserver

2011/03/24/automysqlbackup

I needed a restore solution for my database before launch. I found automysqlbackup which does everything I need. Installation took a bit of trial and error, so here's what works:

curl http://tiny.cc/ey3ap -L > automysqlbackup
sudo mv automysqlbackup /opt
sudo ln /opt/automysqlbackup /usr/bin/automysqlbackup -s
sudo chmod +x /usr/bin/automysqlbackup

Now the script is installed, but is not configured. The settings are read from /etc/automysqlbackup/automysqlbackup.conf, and a sample is inside the script itself, so:

sudo vim automysqlbackup
/START CFG
ma
/END CFG
:'a,. y
:e /etc/automysqlbackup/automysqlbackup.conf
p
ggVG:norm 0dw
:bn :bd :x

All that's left to do here is fill out the information requested by the script and save the file. I wanted the backups written to /var/backups/mysql, so I took a moment to set up the permissions for that.

sudo mkdir /var/backups/mysql
sudo chgrp webdev mysql
sudo chmod 770 mysql

I also noticed that automysqlbackup takes a password variable, but uses mysqldump internally. The man page of mysqldump recommends against passing passwords through the shell. Since this is due to a security issue with ps, I wanted to use the recommended method, a configuration file. I found the dbdump() function in the source of automysqlbackup (line 506 for me), and added a --defaults-file option to the mysqldump call. Due to a bug? this must be the first option given.

# Database dump function
dbdump () {
${MYSQLDUMP} --defaults-file=/home/dylan/.autobackup.cnf --user=${USERNAME}     --host=${DBHOST} ${OPT} ${1} > ${2}
return $?
}
 

Then finish up by creating your .cnf file wherever you chose.

[client]
password=your_password

At this point, the automysqlbackup script is all set to go, but will still get rejected by MySQL. There is no reason to that automysqlbackup needs to do anything but read the database to back it up, so create a new user constrained to the loopback address with these rights alone. Make sure this user is set in automysqlbackup.cnf.

CREATE USER 'autobackup'@'localhost' IDENTIFIED BY 'password';
GRANT select, lock tables ON *.* TO 'autobackup'@'localhost';

All that's left is to schedule it, so toss it into cron.

crontab -e
Go@daily automysqlbackup<Esc>

2011/02/16/amazon ec2 ptr/reverse dns record

Here is the form to request a custom PTR record for an Amazon EC2 Elastic IP. It's included under Request to Remove Email Sending Limitations, which is not helpful.

This is good when setting up a mail server in Amazon's cloud. Without this request, the reverse DNS lookup will the default PTR record - something like ec2-50-16-219-8.compute-1.amazonaws.com. When the reverse DNS record doesn't match the origin domain, mail providers like GMail are likely to mark your mail as spam.

2011/02/14/git post-receive hook for live branch

It's easiest to use a git branch to manage the rollout of updates to my website. I wanted to add a message to the `git push` output when the commit is pushed live. The documentation notes that the post-recieve hook has access to the the ref-name. This script is placed in dylanstestserver.git/hooks/post-receive to do this. It took me some time to realize there are no arguments, the information is available on stdin.

if ! [ -t 0 ]; then
  read -a ref
fi
IFS='/' read -ra REF <<< "${ref[2]}"
branch="${REF[2]}"
if [ "live" == "$branch" ]; then 
  git --work-tree=/var/www/dylanstestserver.com/ --git-dir=.. checkout -f
  echo 'Changes pushed live.'
fi