Sivarg

Services avaible on Woopets.com

2009-07-31T23:10:00.002+01:00

Please feel free to find and add veterinarians, associations, clubs and more on http://www.woopets.com/services

Services are using mostly google geokit to manage proximity.

Disable mpt-status "detected non-optimal RAID status" log messages

2009-07-30T14:32:00.002+01:00

I'm using mpt-statusd on a debian lenny server to monitor raid Status.
However the status of raid is OPTIMAL :

$ sudo mpt-status -i 4
ioc0 vol_id 4 type IM, 2 phy, 698 GB, state OPTIMAL, flags ENABLED
ioc0 phy 1 scsi_id 6 ATA Hitachi HDS72107 AB0A, 698 GB, state ONLINE, flags NONE
ioc0 phy 0 scsi_id 5 ATA Hitachi HDS72107 AB0A, 698 GB, state ONLINE, flags NONE

I keep having messages in syslog "detected non-optimal RAID status" (and sent by mail by logcheck...).

Here's a patch to fix /etc/init.d/mpt-statusd (has been submitted to Debian tracker already) :

--- mpt-statusd 2009-07-30 11:35:21.000000000 +0200
+++ /etc/init.d/mpt-statusd 2009-07-30 11:14:09.000000000 +0200
@@ -55,11 +55,12 @@
# Check ever $PERIOD seconds, send email on every status
# change and repeat ever $REMIND seconds if the raid is still
# bad.
- if (mpt-status -i $ID) |grep -q 'state OPTIMAL' ; then
+ if (/usr/sbin/mpt-status -i $ID) |grep -q 'state OPTIMAL' ; then
BADRAID=false
else
BADRAID=true
- logger -t mpt-statusd "detected non-optimal RAID status"
+ OUTPUT=$(/usr/sbin/mpt-status -i $ID)
+ logger -t mpt-statusd "detected non-optimal RAID status ($OUTPUT)"
fi
STATUSCHANGE=false
if [ true = "$BADRAID" ] ; then

Restart the service, and remember to create /etc/default/mtp-statusd to specify your SCSI ID if needed (mine has ID=4 setted)

Using Authlogic with Cucumber

2009-07-30T14:06:00.005+01:00

[UPDATE] : Unfortunately, I have really odd results with this method. I'll post the right method when I find it ! Sorry !

Recently, I decided to give a chance to Authlogic. Authlogic is a very powerful auth framework. I must admin I miss the simplicity of Clearance, but nevermind.

Authlogic has a lot of doc, including for testing. Anyway, the testing doc is only about Test::Unit, and I mostly use cucumber + shoulda currently.

If you want to test authlogic with cucumber, just add to your step file (features/steps/authentication_steps.rb in my case) :

Before do
include Authlogic::TestCase
activate_authlogic
end

And that's all !

Now, you can write a step like :

Given /^I am a signed in user$/ do
@user = Factory(:user) # or use fixtures, mocks, whatever !
assert @user_session = UserSession.create(@user)
assert_equal controller.session["user_credentials"], @user.persistence_token
end

and that's ALL. Don't require anything, config.gem is already loading everything for you.

Can't understand why this simple wasn't on the net before :)

Happy testing !

Display RAILS_ENV in rails console

2009-05-13T15:13:00.002+01:00

This post is related to this lighthouse ticket.

If you want to have some clue about the current RAILS_ENV (because you need some sleep, or like me you have several terminal instances, with severals rails consoles), just add this to your ~/.irbrc file :

# Prompts
IRB.conf[:PROMPT][:CUSTOM] = {
 :PROMPT_N => "[#{ENV["RAILS_ENV"].capitalize}]>> ",
 :PROMPT_I => "[#{ENV["RAILS_ENV"].capitalize}]>> ",
 :PROMPT_S => nil,
 :PROMPT_C => "?> ",
 :RETURN => "=> %s\n"
}  

# Set default prompt
IRB.conf[:PROMPT_MODE] = :CUSTOM

You will end with a console like mine :

[Development]>> 1+1
=> 2

Enjoy !

Shoulda 2.10.0 and should_respond_to_xml

2009-03-05T10:22:00.004Z

Yesterday, after upgrading Shoulda to 2.10.0, I had the bad surprise of having undefined methods for my tests. Especially, should_respond_to_xml has purely disappear from the macros.

If, like me, you don't want to change your tests (because they make sense, of course), simple create a file in RAILS_ROOT/test/shoulda_macros and put this gist into it : http://gist.github.com/74287

Let's color Git

2009-03-04T19:59:00.003Z

A must-have if you're using git, and especially through command line, put this in your ~/.gitconfig :

[color]
   diff = auto
   status = auto
   branch = auto

This will put some (very) useful colors in diffs, status, and branches.
This is particulary useful when using git status, git log -p and git diff.

Upgrading clearance to 0.4.X

2009-01-22T10:52:00.002Z

If you were using clearance 0.3.X and want to upgrade to 0.4.X, the generator should help a little. Nevertheless, you should use this migration instead :


class CreateOrUpdateUsersWithClearanceColumns < ActiveRecord::Migration
  def self.up
    change_table(:users) do |t|
      t.rename :crypted_password, :encrypted_password
      t.rename :confirmed, :email_confirmed
    end
    remove_index :users, :name => "index_users_on_email_and_crypted_password"
    add_index :users, [:email, :encrypted_password], :name => "index_users_on_email_and_encrypted_password"
  end
  
  def self.down
    change_table(:users) do |t|
      t.rename :encrypted_password, :crypted_password
      t.rename :email_confirmed, :confirmed
    end
    remove_index :users, :name => "index_users_on_email_and_encrypted_password"
    add_index :users, [:email, :crypted_password], :name => "index_users_on_email_and_crypted_password"

  end
end

Be sure to rename all your login* methods to the new sign* naming convention.

Variable affectations in ruby

2009-01-11T22:45:00.002Z

Something I would share today, since I've spend some time yesterday dealing with variables copies and references.

Sounds pretty logical afterward, but it might save you some time :


>> a = {}
=> {}
>> a["my_key"] = "value"
=> "value"
>> a
=> {"my_key"=>"value"}
>> b = a
=> {"my_key"=>"value"}
>> b["my_key"] = "new_value"
=> "new_value"
>> b
=> {"my_key"=>"new_value"}
>> a
=> {"my_key"=>"new_value"}
>>

This simple example demonstrates that affecting an existing Hash to a new variable will send the reference of the Hash, not a copy.
Now, let's try the same with a basic float object :


>> a = 1.0
=> 1.0
>> b = a
=> 1.0
>> b= 2.0
=> 2.0
>> a
=> 1.0

Logical, indeed. But if you're using Hashes and floats in a structure, like in the rails session object, do not forget this example !

Clearance : coming from where your were

2009-01-06T12:13:00.002Z

Clearance will save your current location if you try to access a restricted area, and will forward you to the same place after authentication. Nevertheless, if you put a "Login" link in your pages, your users will be redirected to "/", which is not always what we want. To avoid this, we will use a param to return where we want (this is usually in your app layout) :


link_to t(:login), new_session_url(:return_to => request.request_uri)

Now we have a :return_to param with the correct URI, let's handle this in our session controller :


before_filter :set_return_to, :only => :new
 
  def set_return_to
    session[:return_to] = params[:return_to] if params[:return_to]
  end

Clearance is using this session var to handle the returning URI.

Now, if you're not on the home page, click on "login", enter your email / password, you will be redirected to the previous page, being identified.

Rails Authentication : Clearance

2009-01-06T12:02:00.004Z

Some weeks ago, I was looking for a simple authentication system for rails, with minimal generators (so unlike restful-authentication). I was just experiencing some new testing techniques, and was starting using Shoulda and Factory Girl. Thoughtbot did an amazing job there.
Then, I was browsing github for some new auth system, and my eyes were attracted by "shoulda" and "factory girl" in the page. Thoughtbot has released a (very) simple authentication system, with testing based on shoulda and factory girl. After some gem issues, the new 0.3.7 version is ready to be used. With authentication based on stategies (let's say a simple email/password strategy, or facebook connect), this could be the new killer feature for rails.

I'll try to post some usage examples here, to help getting full advantage of the package.

First, let's install the necessary package :


$ sudo gem install thoughtbot-clearance -s http://gems.github.com

Then, you can simply follow instructions here :

http://github.com/thoughtbot/clearance/tree/master

Logcheck regexp for backup-manager (debian etch)

2008-09-17T09:15:00.003+01:00

If, like me, you're a big fan of logcheck, and you are using backup-manager to backup and upload your data on a distant server, you may have noticed that logcheck will warn you everyday about Info statements in the log files :

To remove info-level alerts regarding backup-manager, you can create a file

/etc/logcheck/ignore.d.server/backup-manager

contening :

^\w{3} [ :0-9]{11} [._[:alnum:]-]+ backup-manager(-purge|-upload)?\[[0-9]+\]: info[[:blank:]]*\*.*$

This will filter all info-level output. If you want to filter also the warn entries (not recommended) :

^\w{3} [ :0-9]{11} [._[:alnum:]-]+ backup-manager(-purge|-upload)?\[[0-9]+\]: (info|warn)[[:blank:]]*\*.*$

UPDATE FOR DEBIAN LENNY :

Sounds like default loglevel is set to DEBUG, and I don't know where to change that, so let's ignore the line in logcheck :

^\w{3} [ :0-9]{11} [._[:alnum:]-]+ backup-manager(-purge|-upload)?\[[0-9]+\]: (info|debug)[[:blank:]]*\*.*$

Note : I prefer to remove alerts on these lines to remove the lines from the logs. In case of problem, I can check system.log for debug information.

Jiwa, good surprise

2008-03-26T09:13:00.002Z

This morning I have discovered Jiwa :

"Jiwa.fm is an online music social network offering free unlimited and high quality music streaming. Users can listen to the music they like, create and share playlists, make recommendations and discuss between them." (Jiwa website)

They propose a very large range of bands and artists (from Mika to Bob Marley, and from Crystal Method to Rage Against the Machine). The sound is very clear and all songs are encoded in high quality (since songs are NOT provided by users). Especially, I appreciate the Jazz "Radio".

What a good competitor to Deezer !

What's missing in ruby on rails

2008-03-21T12:32:00.003Z

I was thinking of this post since a few days, and I discover something today I'd like to share.

I've started to use Ruby on Rails since a few months, and really love the language / framework. I can set up a working website a very few lines, without spending my time configuring and mapping . Nevertheless, when it comes to production websites, rails is very deceptive. Let me explain a standard installation (there are a lot of ways to do that) :

you put your files on the server, and Capistrano is wonderful for that, and not only.
Install and configure some mongrel_cluster instances (yes, for good performances, you'll have to do that)
Install and configure a webserver such as nginx or apache2
Proxify all your mongrel instances using your webserver.

Sounds pretty simple ! It is, actually, once you already have done that before.

My problem is more for hosting several sites on one server. Mongrel is eating a lot of memory, and you'll need to create at least as many instances as the number of sites. That means, if you have a single page site, just for info, you'll need to setup mongrel, and have an instance running and eating your memory. This is mainly because ruby / rails is not thread safe, and one Mongrel instance will handle one request at a time. What a shame.
That also means you need to reserve memory for your site, even if you don't have visitor !

Today I have discovered a new hope in my rails quest : http://modrails.com

Modrails is a new module for apache2, acting like mod_php or mod_perl. That means => EXIT MONGREL ! Now you can host a lot a mini rails websites, just by running a single apache2.
If this module goes out in stable version soon, this will allow (at last !) hosters to provide some performant and scalable mutualized environments for customers.

As far as I know, the module is slightly more performant than mongrel, but I more care about memory usage. I'll try to post a performance comparison post soon.

Mail.app is driving me crazy !

2008-02-19T08:40:00.003Z

I've been using Mail.app for more than 3 years now. I've always considered the product as "correct", and since it's the only decent mail client to use Addressbook (and therefore .mac), I don't have much choice.

Nevertheless, since my upgrade to leopard, this damn client has become more than instable, crashing several times per day (in a professional usage). So sad, Microsoft outlook is crapy, but don't crash that often.

I have no plugin or bundle installed in mail.app, I'm just using the vanilla version. I must also mention I have 4 mail accounts, one is more than 3GB (I like to keep my last 2 years mails).

Can't wait for thunderbird 3, which is supposed to handle the addressbook.app integration.

Mighty Mouse dead

2007-05-25T03:33:00.001+01:00

Here it comes again, my mighty mouse is dying slowly. Again and again, the scrollwheel refuses to scroll down or up. I've found a simple solution, using some kind of "Mr Clean" , and a dry tissue.
Unfortunately, I have not wait enough, and the scrollwheel won't work at all now. The mouse is doing a stange noice, and sometimes even some buttons are not working.

This mouse is great, and especially the idea of the ball scrollwheel. Anyway, Apple should have designed a bigger ball, and more important, a cleaning system. At least something to open easily the mouse is necessary, since it's almost impossible with the current without breaking something.

Idea of the day : Easy opening system to be able to clean the ball.

Hinted 3GP files with Quicktime

2007-05-17T22:49:00.000+01:00

A lot of tools exist over the Internet to transcode video files. A few of them are usable using a CLI (Command Line Interface), and fewer are available for free. I've tested the most common ones, like ffmpeg, VLC (using again ffmpeg for encoding), or mencoder.
Those (free) (and opensource) tools, are a little hard to use for beginners, and they don't provide any hinting feature, necessary for streaming using QTSS or DSS.

Install qt_tools :

qt_tools is a set of perl scripts available for free. Those tools let access to quicktime functionalities, even those available only with QT pro, without having the licence.

qt_tools can be found here. I've installed the tools into the default location : /usr/local

Transcode video files :

This is an example of the usage of qt_tools. We are using the tool qt_export to transcode the file into a 3gp format:

$ /usr/local/bin/qt_export --video=mp4v,12.5,80,16 --audio=samr,8000,8,1 --datarate=18 --exporter=3gpp --replacefile original-movie.mov new-movie.amr.mpg4.3gp

Please refer to the manual of qt_export for more information.

Hinting :

The hinting part can be done using several (opensource again) tools, like MP4Box (from the GPAC project), or mp4creator (from the MPEG4IP project). Again those tools must be compiled most of the time by hand, meaning hard maintenance and upgrades.

Let's use some tools available with Mac OS X server (only !) : qtref and qtmedia :

$ qtref -t new-movie new-movie.amr.mpg4.3gp

I have now a .qtl file :

$ cat new-movie.qtl

Now let's use qtmedia to hint the file :

$/usr/bin/qtmedia -h new-movie.qtl
kCGErrorRangeCheck : Window Server communications from outside of session allowed for root and console user only

-File- "/Library/QuickTimeStreaming/Movies/new-movie.qtl"
-Processing-
-Hinting movie...
+Hinting completed.
-Info-
Hinted: Yes
Ref Movie: No
Size: 176.00 x 144.00
Duration: 119.00 seconds
DataRate: 22893.74 bytes/secs
BitRate: 171916 bits/secs
No user data.

Don't worry about the warning, it's just a warning, and will go if you run the command as root...

Now you can test the movie using quicktime or your mobile by putting the new file into QTSS folder :

$ mv new-movie.qtl /Library/QuicktimeStreaming/Movies/new-movie.3gp

(note that the qtl file is now the hinted movie !)

Of course, the purpose of all of this is to be able to automate the process using a shell script, or Automator.

xserve HDD

2007-05-16T19:59:00.000+01:00

I think it's time for Apple to renew its server product range.
I've installed some HP server (Proliant DL 365), and was quite impressed by the quality of the servers. They are using 2,5" HDD instead of the current 3,5" on the xserve. The difference is important, because the HP server can embark 6 HDD versus 3 for the xserve, and also in 1U.

[update] : I just forgot to mention that the harddrives in the HP server are SCSI versus SATA in the xserve.

Racking the xserve

2007-05-16T19:29:00.000+01:00

The rack we are using is generally xserve allergic. Our Xserve G5 are just dropped on rack shelves, since the rails don't fit our rack. Hopefully, Apple has corrected this issue today, and I was able to rack the server in less than 30 minutes (power + network installed). Moreover the protection on the power supply to avoid unwanted unplug of the power cable has been corrected. It took us 15 minutes on the G5 to install, and just a few seconds on the new one.

I just had a few network issues (forgot to remove a proxy configuration). I hope apple will do someday a configuration utility based on ncurse or such. There are a few tools to configure everything including serversetup, networksetup, changeip, ...

I didn't manage to make to Lights Out Management (LOM) work (Remote monitoring), I have to work on this.

Just received our new Apple xserve

2007-05-16T08:10:00.000+01:00

Yesterday was a day of excitement : we just received our new mail server. We've bought a new Apple Xserve Intel, with 1GB RAM and one 80GB HDD. In order to have a robust configuration (the server might be used for other purposes than mail), I added 2x512MB Memory slots and 2 80 HDD.

Memory and disk upgrades

Upgrading the xserve was extremly simple. Once the server opened (2 screws at the back, removable by hand), there's a large access to the memory slots. Apple had the good idea of letting some space between the slots. More information on upgrading memory : http://manuals.info.apple.com/en/Xserve_Intel_DIY_FB-DIMM_Memory.pdf
The HDD were also easy to plug, I just had to push on the panel, remove the empty rack, and put the new one. More info :
http://manuals.info.apple.com/en/Xserve_Intel_DIY_HardDrive.pdf

Installation

Lucky me, I had a keyboard/mouse and TFT screen to install the server. Way simpler than having to do it remotely (ssh or VNC). After booting on the installation DVD using the "C" key at boot, it was time to create a raid Array for our server.
The Disk Utility was there, and I discovered that it was impossible to create a software RAID 5 array. I decided to go with a RAID 1 with one HDD spare. Using this configuration, I just benefit of 75 GB of available space, but if a disk fails, I will have some time to change it.

Once the raid array created, I followed the installation procedure, and every thing went OK. A bad point : you can't install Mac OS X without a mouse ! Using tab won't put the focus on "Next" buttons. Too bad.

Comparison to G5

Compared to our G5 servers, the new xserve seems to be a lot faster. The back of the server changed a little, but the modifications are very basic. It's a good idea to propose by default the video card. Configuring a headless server is sometimes painful.
There's also a little plastic card on the back of the server, on which you can retrieve the serial number (that's the default root password before you install !), and the NICs mac addresses.

First Post Test

2007-05-07T17:39:00.000+01:00

Hi,

This is my first post. More to come soon !

Enjoy !