Hendrik Volkmer

Deploying a Rails application on OpenSolaris with passenger aka modrails

2009-01-29T12:09:47+01:00

Two technologies I wanted to play around with for some time are OpenSolaris and passenger. OpenSolaris because of ZFS, Zones, SMF and the latest addition IPS and passenger because it seems to be the new default Rails deployment model and is really easy to configure.

So I deceided to combine those to and get started. Here’s what I did. Maybe it’s useful for someone who’s trying to do the same.

I grabbed the OpenSolaris 2008.11 Virtualbox image from virtualboximages.com and booted it in VirtualBox.

I wanted to make sure to have the latest updates so I updated the image (all commands are executed as root):

# pfexec pkg image-update

A reboot later I installed the necessary parts for a Rails environment:

# pfexec pkg install SUNWruby18
# pfexec pkg install gcc-dev

SUNWruby18 installes ruby 1.8.6p287 which is exactly what I needed.

The mysql installation was a bit harder:

# pfexec pkg install SUNWmysql5
# /usr/mysql/5.0/bin/mysql_db_install
# chown -R mysql:mysql /var/mysql/5.0/data

I needed to manually chown the data dir. Otherwise mysql wouldn’t start.

You can confirm that your mysql is running via:

# svcs -l mysql

The logfile gives further information if something goes wrong.

To access mysql try:

# /usr/mysql/5.0/bin/mysql

The “5.0” in the path is important because there’s also a mysql version 4 installed. This will be of interest when installing the mysql ruby gem. The command to correctly install the gem is:

# pfexec gem install mysql -- --with-mysql-lib=/usr/mysql/5.0/lib/mysql --with-mysql-include=/usr/mysql/5.0/include

Thanks to Amanda Waite for this hint.

The next installation step is Apache:

# pfexec pkg install SUNWapach22

Make sure it starts via svcadm:

# svcadm start apache22

Check that it runs:

# svcs -l apache22

(or just type http://localhost in your browser)

I needed memcached, so I installed it:

# pfexec pkg install SUNWmemcached

It didn’t start right away, though. There are some changes you have to apply to run memcached in solaris

# svccfg
svc:> select memcached
svc:/application/database/memcached> setprop memcached/options=("-u" "nobody" "-m" "2048")
svc:/application/database/memcached> quit
# svcadm refresh memcached

Of course we need rails:

pfexec gem install rails

The current stable version of passenger doesn’t support Solaris but the egde version does. So I downloaded the passenger edge Version off github.

It didn’t compile right away. There are some compiler flags that aren’t supported in Solaris so I used these changes to make passenger compile.

Commandwise this looked like this: I unpacked the tarball into /opt/passenger and then did a

# cd /opt/passenger
# vi lib/passenger/platform_info.rb # see above
# bin/passenger-install-apache2-module
# chown -R webservd /opt/passenger

Add the mentioned config snippet to /etc/apache2/httpd.conf and restart apache

# svcadmin restart apache2

Now were all set to create a new rails app. I created a seperate ZFS partition to serve the apps. Just because it’s so easy to do with ZFS and it gives you more control on how much space is used by your apps etc.

# zfs create rpool/apps

Now it’s time for the demo app:

# cd /rpool/apps
# rails demo
# chown -R webservd demo

I added the DocumentRoot of the app to the apacheconfig (DocumentRoot /rpool/apps/demo/public) and restarted apache again (svcadm restart apache2 – you know the drill by now).

And that’s it. A http://localhost gave me the demo welcome page. Nice.

Later I wanted do deploy a rails app via capistrano so I needed git. There isn’t a IPS package for git yet, so I decided to take the standard compile route. I downloaded git and compiled it like this:

# ./configure --prefix=/opt/git
# make install
# ln -s /opt/git/bin/git /usr/bin/git

I used the symlink because capistrano wasn’t able to find git otherwise – setting scm_command didn’t work.

If anyone tries this recipe and it works or it doesn’t work… just let me know. Have fun with OpenSolaris. It’s worth checking out.

Really Adobe? Really?

2008-10-09T13:09:47+02:00

I just encountered this problem and I can totally understand this blog post: If Adobe were a man, i would punch him

Let me remind you: This is a product that costs about 1K EUR and it’s the year 2008.

So Dear Adobe : Please fix your products!

Russian technology

2007-05-28T13:09:47+02:00

I’ve been using russian technology for about a week now and it works really well with Rails and is easy to set up in general.

I’d like to thank err for his nginx config generator and Ezra for making me aware of nginx and his config file. This combination got me up and runing with ngnix in about 5 minutes.

However, I needed to modify the generated config file a bit to fit the standard debian file paths and users. So I modified the ERB template which is used to generate the config file. It should be fully backwards compatible, so if you don’t need all of these modifications you could use it and just use the modifications you need. Here’s the template with my modifications

My YAML file looks like this:

user: www-data
group: www-data

pid_path: /var/run/
config_path: /etc/nginx/
log_path: /var/log/nginx_

This defines the default roots and alternate roots.
The name (or key) of your site will be sprintf’d with
the below root, right.
root:
blog: /srv/hendrikvolkmer.de/blog/public
www: /var/www
All your vhosts.
sites:
blog:
upstream:
– 127.0.0.1:3000
– 127.0.0.1:3001

server_name: blog.hendrikvolkmer.de

I added a special feeds config which generates the
if ($http_user_agent !~ FeedBurner) line before the actual rewrite
line. Works fine with Mephisto and, of course Feedburner, YMMV
feeds:
– ^/feed/atom.xml http://feeds.feedburner.com/hendrikvolkmer
– ^/feed/all_comments.xml http://feeds.feedburner.com/hendrikvolkmer_comments

root: blog

www: root: www server_name: hendrikvolkmer.de

Also, it’s really easy to compile nginx 0.5.20 for Debian: Just download the source, make the necessary modifications to the current deb file specifications and compile. :)

Usefull error messages - Part III

2007-04-27T13:09:47+02:00

Today I found another one:




postfix/qmgr[10675]: warning: problem talking to service rewrite: Success

Installing Trac on Windows

2007-02-21T12:09:47+01:00

This is not a tutorial. There is a good one on the TracWiki. I just want to tell everyone, that if you ever consider installing Trac on Windows, then just don’t do it and install it on a *nix machine. This will save you a lot of time and nerves. There are even pre-build environments available with SVN, Trac and Cruisecontrol.

I spent almost an entire day to configure Trac on Windows. It’s not funny. If you have to do it, be sure to use the mod_python setup. It’s far easier to install than trying to set up CGI on a Windows box (which I think would be the first and easiest setup on *nix machines). I may not be the smartest person on earth but the Apache error messages on Windows are just plain useless. I don’t care if it’s Apache’s fault, Windows’ fault or my own. It’s just broken.

So if you have no other choice than using Windows, make sure that you use a plain Apache from apache.org – No xampp, Typo 3 package or anything like that. Some of these packages come with their own version of Python which really fucks things up, if it’s not the version you need). This will save you a lot of trouble.

The main problem we had here was the “We already have an Apache on this server, so why not install on this one?”-thing. The Apache installation turned out to be one of the packages mentioned above which made things not really easier.

Useful error messages...

2007-01-21T12:09:47+01:00

… are great. Here’s one I came across while trying to recover data from a hard disk image of a broken hard disk:

core backup # xfs_repair -f data.image
Phase 1 - find and verify superblock...
superblock read failed, offset 249108103168, size 2048, ag 0, rval 58

fatal error -- Success
core backup #

fatal error — Success. Now that’s a useful message, isn’t it? Btw. the image wasn’t recoverable…

Windows Task manger fun

2006-12-16T12:09:47+01:00

I never liked Windows, but in the last years I learned to live with it. But sometimes it reminds me why I will never ever even consider liking it. Check this out:

Some process got wild and I wanted to kill it using the Task Manager – which apparently caused the task manager to go nuts as well. I’ve seen processes which I couldn’t kill in Linux but I never saw “kill” itself go crazy.

Remote system management using MS Windows

2006-12-07T12:09:47+01:00

or: Remote access the painful way

If you are lucky enough to use the RDC protocol it’s not that painful. But if you are forced to use something like PC Anywhere via a dialup line and then VNC to another machine. Oh boy, that’s just pure pain. I had to do something like this yesterday and I really wished I had a simple SSH connection.

Who came up with this stupid “a server OS needs a GUI, too”-idea in the first place? Being forced to use a GUI on a server is just plain stupid and being forced to use it for Remote Access is even worse. I know there’s SSH for Windows and stuff… but it’s not in the standard distribution and probably never will be. I don’t want to imagine the Vista based Server OS from Microsoft and how the GUI will look then.

Fedora upgrade vs. Kubuntu installation fun

2006-10-30T12:09:47+01:00

So Fedora Core 6 and (K)Ubuntu 6.10 were almost released at the same time last week. First, I tried to upgrade Fedora which failed because of some package dependencies yum couldn’t resolve. I have never done a Fedora Update without any hassle. It’s quite annoying. I wanted to switch back to a Debian based system anyway so I decided to give Kubuntu a try (I’m more a KDE user than a Gnome user, hence the K).

I downloaded the Live-CD image, burned it, booted off it and tried to install… realizing the main reason why I’m using Fedora right now: The ability to seamlessly install on a Software RAID / LVM setup. Ubuntu doesn’t support that out of the box. It may be possible with some teawking and stuff… but I’m tired of all that. I did enough RAID / LVM installations manually back in the day using Debian, Gentoo and whatnot. It was fun, it worked, but I’m lazy now and I want to have some nice installer do all that for me. Fedora does exactly that (but yum just isn’t as good as apt). Example (And I don’t have any fancy external repos configured… at least not many ;) – anyway yum should be able to handle that):

Error: Unable to satisfy dependencies

Error: Package elfutils-libs conflicts with elfutils < 0.123-1.fc6.

Error: Package elfutils needs elfutils-libelf = 0.119-1.2.1, this is not available.

Error: Package elfutils-libelf conflicts with elfutils < 0.123-1.fc6.

Error: Package cyrus-sasl-plain needs cyrus-sasl-lib = 2.1.21-10, this is not available.

Error: Package hal conflicts with kernel < 2.6.17.

Error: Package autofs conflicts with kernel < 2.6.17.

Error: Package openoffice.org-core needs libicui18n.so.34, this is not available.

Error: Package openoffice.org-core needs libicule.so.34, this is not available.

Error: Package openoffice.org-core needs libgcjawt.so.7, this is not available.

Error: Package openoffice.org-core needs libicuuc.so.34, this is not available.

Error: Package elfutils needs libdw.so.1(ELFUTILS_0.119), this is not available.

Error: Package openoffice.org-core needs libgcj.so.7, this is not available.

Error: Package elfutils needs elfutils-libs = 0.119-1.2.1, this is not available.

And yum is just f-in’ slow compared to apt in terms of resolving dependencies and installation/upgrade of new packages in general. So basically it’s broken.
So I did what I did when I updated Fedora the last time: I downloaded the DVD Image, burned it, booted off the DVD and said “Upgrade”. That worked quite well and after another “yum update” everything was fine. To bad that a simple “apt-get update && apt-get dist-upgrade” would have done the same thing on a Debian based system without problems… I’m quite sure.

So being lazy on one side means having to work more on the other one… Whatever. I’m running Fedora Core 6 now and a nice thing I noticed are these anti-aliased fonts. Somehow I didn’t have anti-aliasing activated… Now I have and it looks quite nice. The Monospace font in “Konsole” is nice to look at, too.

(Home?)-NAS Hardware-Plattform

2006-08-15T13:09:47+02:00

Ich bin schon länger auf der Suche nach einer stromsparenden Plattform für ein kleines Home-NAS-Serverchen. Das hört sich ideal an: 16 MIPS CPUs, 50 W Stromverbrauch, 8x GbE onboard, SATA/SAS-Controller onboard, normales ATX-Board, Unterstützung für Verschlüsselung und läuft unter Linux.. Genau richtig. Nur den Preis findet man nirgends. Wenn das noch für einen 3-stelligen Betrag zu haben sein sollte, steht einem Home-NAS nix mehr im Weg. Allerdings wird das vermutlich nur an OEMs verkauft werden und somit für den Privatmann nicht zur Verfügung stehen – egal zu welchem Preis :( – Schade eigentlich.
[via heise]