Shane Sveller » Linux

Redmine and Passenger bugfix

Shane — Mon, 11 Aug 2008 15:15:25 +0000

For those of you who are running Redmine on Phusion Passenger, if you have certain versions of the RedCloth gem installed, you are in for some trouble.

I was getting something along the lines of:

Error message:
    superclass must be a Class (Module given)
Exception class:
    TypeError
Application root:
    /var/rails/redmine-0.7

This was on Ubuntu Hardy 8.04.1, with Apache 2.2.8-1, Ruby Enterprise Edition 080704 (I think), and Passenger 2.0.2, with Redmine from SVN-stable r1729.

A simple solution that worked perfectly for me was to remove ALL version of the RedCloth gem, allowing Redmine to use its own included version of RC.

gem uninstall RedCloth

One of the responses on that bug thread also indicates that gem version 3.0.4 of RedCloth is the most recent version compatible with Redmine, so you could also try using (only) that version.

Passenger and RubyEnt updates

Shane — Mon, 11 Aug 2008 15:01:00 +0000

Updated my web server with Ruby Enterprise Edition 1.8.6-20080810 and Phusion Passenger 2.0.3 successfully. Both releases feature bugfixes for security issues and the like. My web server currently powers this blog, a Redmine instance for my own projects, and some “production” servers for a few Rails apps I’m writing right now.

I’d highly recommend Passenger and RubyEnt for low-volume or even high-traffic production Rails servers. It has greatly simplified my deployment and performance on a VPS with, say, 540MB of RAM is stellar. If you set the idle timeout to something huge, you can get almost Nginx+Mongrel/Thin level performance even on a low-traffic server; otherwise you have to wait a brief time (10-15 seconds) when the app has been idle for a while.

Building libxml-ruby gem on Ubuntu Hardy

Shane — Mon, 14 Jul 2008 17:07:39 +0000

Source: Tibs’ Blog

I had been having a lot of trouble using libxml-ruby on Ubuntu Hardy for a hobby project of mine. Any task including libxml calls via require 'xml/libxml' fails with the message:

no such file to load -- libxml_so

To solve this:

sudo apt-get install libxml2 libxml2-dev
cd /usr/lib/ruby/gems/1.8/gems/libxml-ruby-0.7.0
sudo ruby ext/libxml/extconf.rb
sudo make; sudo make install

Another successful Passenger deployment

Shane — Wed, 25 Jun 2008 18:16:38 +0000

This morning, as a result of a hardware failure on my home server, I decided to move my Redmine devtracker to my Linode host.I used:

Ubuntu Server 8.04 AMD64
Apache2-prefork 2.2.8
Passenger 2.0.1
Ruby Enterprise Edition 1.8.6-20080624
Redmine 0.7 stable (SVN r1547)

My apache vhost is as follows:


  ServerName dev.shanesveller.com

  DocumentRoot /var/rails/redmine-0.7/public


   Options FollowSymLinks
   AllowOverride None
   Order allow,deny
   Allow from all


  # Custom log file locations
  # ErrorLog  /var/rails/redmine/log/error.log
  # CustomLog /var/rails/redmine/log/access.log combined

The passage in the middle is there because my CSS styles and images were missing. This is explained in the User’s Guide.

This allows me to serve my WordPress blog and my Redmine tracker and use about 260MB of RAM. If you take off the buffers and cache memory, it’s actually more like 110MB.

To enable Passenger and RubyEE in Apache, I used the following file in /etc/apache2/conf.d/passenger:

   LoadModule passenger_module /usr/lib/ruby/gems/1.8/gems/passenger-2.0.1/ext/apache2/mod_passenger.so
   PassengerRoot /usr/lib/ruby/gems/1.8/gems/passenger-2.0.1
   RailsRuby /opt/ruby-enterprise/bin/ruby

The relevant Ubuntu packages are:

apache2 apache2-prefork-dev apache2-mpm-prefork build-essential ruby1.8-dev ruby ri rake irb rdoc libmysqlclient15-dev libopenssl-ruby

Live backups using LVM and dump

Shane — Tue, 27 May 2008 19:15:09 +0000

If, like me, you’ve built your Ubuntu machine using LVM for all partitions except /boot, you can do a live backup of your system using LVM snapshots and the dump tool.

# lvdisplay
  --- Logical volume ---
  LV Name                /dev/odin/root_1
  VG Name                odin
  ...
  LV Size                30.00 GB
  ...

Since my root partition is on LVM, I can do a snapshot to get a point-in-time view of the root partition even as the system is running and software is making changes to the filesystem. To create the snapshot:

lvcreate -L 30G -s -n root_1_snap /dev/odin/root_1

The arguments are: lvcreate -L {size} -n {snapshot name} {lvm device path}

Then, you do a dump backup:

# dump 0uf odin.0.img /dev/mapper/odin-root_1_snap
  DUMP: Date of this level 0 dump: Mon Apr 28 10:52:39 2008
  DUMP: Dumping /dev/mapper/odin-root_1_snap (an unlisted file system) to odin.0.img
  DUMP: Label: none
  DUMP: Writing 10 Kilobyte records
  DUMP: mapping (Pass I) [regular files]
  DUMP: mapping (Pass II) [directories]
  DUMP: estimated 3130770 blocks.
  DUMP: Volume 1 started with block 1 at: Mon Apr 28 10:52:46 2008
  DUMP: dumping (Pass III) [directories]
  DUMP: dumping (Pass IV) [regular files]
  DUMP: Closing odin.0.img
  DUMP: Volume 1 completed at: Mon Apr 28 10:55:49 2008
  DUMP: Volume 1 3120540 blocks (3047.40MB)
  DUMP: Volume 1 took 0:03:03
  DUMP: Volume 1 transfer rate: 17052 kB/s
  DUMP: 3120540 blocks (3047.40MB) on 1 volume(s)
  DUMP: finished in 182 seconds, throughput 17145 kBytes/sec
  DUMP: Date of this level 0 dump: Mon Apr 28 10:52:39 2008
  DUMP: Date this dump completed:  Mon Apr 28 10:55:49 2008
  DUMP: Average transfer rate: 17052 kB/s
  DUMP: DUMP IS DONE

The arguments are: dump {level}uf {output file} {source device path}

Next, remove the snapshot to keep from hogging LVM space:

  lvremove /dev/odin/root_1_snap
Do you really want to remove active logical volume "root_1_snap"? [y/n]: y
  Logical volume "root_1_snap" successfully removed

Optionally, compress the backup image for safekeeping.

gzip /backup/odin.0.img &

The ampersand (&) will make the process run in the background. By doing this, I saved 71.2% of the size, down to an image that was 878MB from 3050MB!

To restore from said backup, assuming you have a ready ext3 partition of the appropriate size: Uncompress it if need be:

gunzip /backup/odin.0.img.gz

Then:

cd / && rrestore rf /backup/odin.0.img

Quick Tip: Ubuntu default runlevel

Shane — Thu, 24 Apr 2008 18:39:16 +0000

If you, like me, use sysv-rc-conf to change the services that run a particular runlevel, i.e. so that runlevel 2 is actually only networked and not GUI as tradition holds, you can set the default runlevel to boot in Ubuntu by editing:

/etc/inittab:
id:3:initdefault

The number in the middle is the runlevel to start by default.

Personal internet acceleration, part 1

Shane — Mon, 14 Apr 2008 17:57:07 +0000

My SFF server machine has been configured to speed up my digital life.

Squid caching web proxy
apt-cacher caching apt proxy
BIND9 caching/local DNS server

Installing the squid web proxy

apt-get install squid squid-prefetch

You will then need to configure the ACL to allow access through the proxy from the local network.

nano /etc/squid/squid.conf

visible_hostname server1     #Line added
...
http_port 8080     #Line changed
...
cache_dir ufs /var/spool/squid 2000 16 256     #Line changed
...
acl local src 10.0.1.0/255.255.255.0     #Line added
http_access allow local     #Line added

This configures the squid cache to listen on port 8080 instead of 3128, allows access from all IPs in the range 10.0.1.1 – 10.0.1.255, and sets the cache size to 2000MB. Then you need to restart the squid service:
/etc/init.d/squid restart

You can then configure your end users’ machines to use http://server1:8080/ as their internet proxy. I recommend setting this for insecure (non-SSL) traffic only as I have noticed some odd behavior with SSL proxying.

In Firefox this is dead-simple:

Open the Preferences window from the application menu on OSX or the Options window from the Tools menu on Windows.
Choose the Advanced section at the top, followed by the Network tab.
Click the Settings button under the Connection header.
Choose the “Manual configuration” radio button and enter http://server1:8080/ in the HTTP Proxy box.

Safari on OSX uses the system-wide proxy settings:

Choose System Preferences from the Apple menu
Click the Network item under the Internet & Network heading
Select your primary connection, wireless or wired, from the left-hand list and click the Advanced button
Choose the Proxies tab, choose Manually in the drop-down menu, check the box next to Web Proxy
Enter server1 in the first box and 8080 in the second box

Using binary NVidia 8800GT drivers on Ubuntu Gutsy

Shane — Mon, 14 Apr 2008 14:31:36 +0000

In order for me to use the official binary drivers, version 169.12, from NVidia with my GeForce 8800GT on Ubuntu Gutsy, I had to do some research. Alberto Milone‘s envy tool makes the whole thing almost idiot-proof.

Text-based install

After completing a fresh install of Gutsy via alternate CD, boot into the “recovery mode” option from GRUB. You will be dumped into a BASH shell.

apt-get update && wget http://albertomilone.com/ubuntu/nvidia/scripts/legacy/envy_0.9.10-0ubuntu10_all.deb gdebi -n envy_0.9.10-0ubuntu10_all.deb envy -t

At this point it will probably install a few required packages via apt-get and then give you a menu. Here we want option 1, “Install the NVidia driver”. It theoretically will install all required development packages, download the driver from nvidia.com, and build and install it for you. It may fail saying it was missing a bunch of required packages, but if you install them by hand it will complete successfully.

apt-get install

GUI-based write-up to follow.

Ubuntu Gutsy AMD64, LVM, and Xen from repositories

Shane — Thu, 10 Apr 2008 16:08:57 +0000

My desktop just grew by two 500GB Western Digital Caviar drives. Having installed those, I loaded Gutsy-AMD64 on one using LVM. The partition schema is as follows:

Drive: 500GB #1
Volume Group: server1
Volumes:
	swap_1 - 6GB - swap
	root_1 - 30GB - ext3 - mounted as /
	home_1 - 20GB - ext3 - mounted as /home/
Free space: ~398GB

500GB number two will probably wind up being a RAID mirror via my motherboard. This LVM layout allows me to dynamically allocate space for the many potential Xen DomUs I create.

To install Xen on Gutsy-AMD64 desktop version:
sudo apt-get install ubuntu-xen-desktop-amd64

My /etc/xen-tools/xen-tools.conf:

lvm = server1               # LVM volume group to use
install-method = debootstrap
size   = 4Gb                # Disk image size.
memory = 256Mb              # Memory size
swap   = 512Mb              # Swap size
fs     = ext3               # use the EXT3 filesystem for the disk image.
dist   = gutsy              # Default distribution to install.
gateway = 10.0.1.1
netmask = 255.255.255.0
cache = yes                 # Cache downloaded debs
passwd = 1                  # Set up root password during creation
kernel = /boot/vmlinuz-`uname -r`
initrd = /boot/initrd.img-`uname -r`
arch = amd64                # 64-bit domU
mirror = http://us.archive.ubuntu.com/ubuntu
ext3_options = noatime,nodiratime,errors=remount-ro
serial_device = tty1

My /etc/xen/xend-config.sxp:

(xend-tcp-xmlrpc-server yes)
(xend-relocation-server yes)
(xend-relocation-server yes)
(xend-tcp-xmlrpc-server-address '')
(xend-tcp-xmlrpc-server-port 8006)
(xend-relocation-port 8002)
(xend-relocation-address '')
(xend-relocation-hosts-allow '')
(network-script network-bridge)
(vif-script vif-bridge)
(dom0-min-mem 196)
(dom0-cpus 0)
(vncpasswd '')

Edit the file /etc/xen-tools/xm.tmpl as follows, around line 19:

memory      = '{$memory}'
extra='xencons=tty'

To generate one Gutsy-based DomU:
xen-create-image --hostname=gutsy1 --ip=10.0.1.3 --verbose --force

To boot created DomU:
xm create gutsy1.cfg -c



God is watching over my Rails
Shane — Mon, 07 Apr 2008 19:48:58 +0000
On my home server I’m toying with some Rails clustering and such, using nginx and thin, and now God to watch over things.

With this setup, I can add apps and port-ranges to monitor any time I like, and I can split off a particularly specific configuration group if an app has unusual trends or requirements, such as using loads of memory.


RAILS_ROOT = '/var/www/rails'

APPS = {
  'redmine' => %w{5000 5001 5002 5003 5004},
  'typo5' => %w{4000 4001 4002}
  }

for app in APPS.keys
  for port in APPS[app]
    God.watch do |w|
      w.name = '#{app}-thin-#{port}'
      w.interval = 30.seconds # default
      w.start = 'thin start -C #{RAILS_ROOT}/#{app}/config/thin.yml -o #{port}'
      w.stop = 'thin stop -C #{RAILS_ROOT}/#{app}/config/thin.yml -o #{port}'
      w.restart = 'thin restart -C #{RAILS_ROOT}/#{app}/config/thin.yml -o #{port}'
      w.start_grace = 10.seconds
      w.restart_grace = 10.seconds
      w.pid_file = File.join(RAILS_ROOT, '/#{app}/tmp/pids/thin.#{port}.pid')

      w.behavior(:clean_pid_file)

      w.start_if do |start|
        start.condition(:process_running) do |c|
          c.interval = 5.seconds
          c.running = false
        end
      end

      w.restart_if do |restart|
        restart.condition(:memory_usage) do |c|
          c.above = 80.megabytes
          c.times = [3, 5] # 3 out of 5 intervals
        end

        restart.condition(:cpu_usage) do |c|
          c.above = 50.percent
          c.times = 3
        end
      end

      # lifecycle
      w.lifecycle do |on|
        on.condition(:flapping) do |c|
          c.to_state = [:start, :restart]
          c.times = 5
          c.within = 5.minute
          c.transition = :unmonitored
          c.retry_in = 10.minutes
          c.retry_times = 5
          c.retry_within = 2.hours
        end
      end
    end
  end
end