November 2009 (#168):

• Mailbag
• Talkback
• 2-Cent Tips
• News Bytes, by Deividson Luiz Okopnik and Howard Dyckoff
• A Short CGI Script for Passing 404s to Another Server, by Silas Brown
• Away Mission - Upcoming in November '09, by Howard Dyckoff
• Setting up a MySQL Cluster for your Linux desktop, by Mark Nielsen
• A 'revisited' guide to GNU Screen, by Anderson Silva and Steve 'Ashcrow' Milner
• HelpDex, by Shane Collinge
• Ecol, by Javier Malonda
• XKCD, by Randall Munroe
• Doomed to Obscurity, by Pete Trbovich
• The Linux Launderette

Mailbag

This month's answers created by:

Gazette Matters

Garbage mails in TAG

Dr. Parthasarathy S [drpartha at gmail.com]

There is tooooo many garbage mails/spam mails in the TAG mailing list. Can no one take some action ? Please filter out all garbage. Otherwise people will stop reading or answering TAG mails.

Thank you,

partha

-- 
---------------------------------------------------------------------------------------------
Dr. S. Parthasarathy                    |   mailto:[email protected]
Algologic Research & Solutions    |
78 Sancharpuri Colony                 |
Bowenpally  P.O                          |   Phone: + 91 - 40 - 2775 1650
Secunderabad 500 011 - INDIA     |
WWW-URL: http://algolog.tripod.com/nupartha.htm
---------------------------------------------------------------------------------------------

[ Thread continues here (18 messages/26.82kB) ]

Our Mailbag

Disabling Gnome application sounds in KDE4

Suramya Tomar [security at suramya.com]

Hey,

I recently installed Ubuntu 9.04 on my system and then installed KDE 4.2.2 on the system using apt-get.

I don't like any system sounds (beeps when opening windows/ on popups etc etc) so I went to the KDE control panel and disabled all the system sounds over there. In KDE 3.5 this used to disable all sounds in Gnome applications as well but in 4.2.2 this doesn't seem to be the case.

So I ran the gnome Control panel and disabled all system sounds over there. Now the problem is that this setting doesn't seem to be saved. Thus when I restart the system all the system sounds on gnome apps (Nautilus/Firefox/gedit/X-Chat etc) are back.

So how do I set it so that the sounds are disabled and stay that way over reboots?

Also the same issue is there with the fonts in the Gnome applications. The fonts in gnome applications (Size mainly) doesn't match the ones in KDE applications. I have to change them using the Gnome control panel and the changes don't stay saved.

Any idea's/Suggestions? I did try searching for solutions but none of the solutions suggested by google worked.

Let me know if you need any additional details.

- Suramya

-- 
-------------------------------------------------
Name : Suramya Tomar
Homepage URL: http://www.suramya.com
-------------------------------------------------

[ Thread continues here (6 messages/7.23kB) ]

Efficient chmod/chown

Mike Orr [sluggoster at gmail.com]

Is there a version of chmod/chown that recursively changes files only if they're different from the specification? The stock version changes them unconditionally, which updates their 'ctime' and thus makes 'rsync' transfer the inode properties even if there's no real change. I could write a Python version, but I was hoping there might be a C version somewhere that would be more efficient with millions of filenames.

-- 
Mike Orr <[email protected]>

[ Thread continues here (8 messages/8.34kB) ]

Developing a small OS from a preinstalled Linux

Abhishek Sharma [spyzer.abhishek0 at gmail.com]

Hi,

I am working on a project of making a small OS which will boot up on my dual core machine and will serve as a simple calculator. If its possible I would like to add GUI to it too.

My question is for this job do I need to program all the stuff as assembly code or is there a way in which the precompiled linux kernel or those "boot.img | initrd.img" can help me without having to go through assembly development?

If its possible I would be glad if you tell me how I can achieve this.

Thank You.

[ Thread continues here (7 messages/8.71kB) ]

Followup: Regarding How can Define congestion windows in Fedora

Predrag Ivanovic [predivan at nadlanu.com]

On Fri, 18 Sep 2009 23:06:39 -0500

<forwarded message from Master of Technology>

>i have some proble to define the
>
>1.Congestion  windows Size 2.algorithms. in NS-2 with Linux(Fedora)
>
>Thanking you.
>
>Mr. Brahm Deo sah
>India

http://lmgtfy.com/?q=Congestion+%20windows+Size+%20algorithms.+in+NS-2

Pedja

-- 
 improperly oriented keyboard

[ Thread continues here (3 messages/2.96kB) ]

Talkback: Discuss this article with The Answer Gang

Talkback

Talkback:167/okopnik.html

[ In reference to "A Quick-Fire chroot Environment" in LG#167 ]

Jimmy O'Regan [joregan at gmail.com]

'as the "fortune" program says, "The best book on programming for the layman is Alice in Wonderland; but that's because it's the best book on anything for the layman."'

I think I laughed harder than I usually would have at this, because I'd recently read a linguistics paper (http://www.kwartjez.amu.edu.pl/Krazynska,%20Sloboda.pdf) whose 'Further Reading' section contained, sandwiched between various works about proto-Indo-European, a pointer to a translation of this very book. It also happens that when asked by a friend if it were possible to get old books in Polish and English text, and English audio, to be the only book I could find legally in all three forms.

Curiouser and curiouser.

Talkback:128/saha.html

[ In reference to "Coding a Simple Packet Sniffer " in LG#128 ]

Aisha S. Azim [aisha.s.azim at gmail.com]

On Tue, Oct 6, 2009 at 7:54 PM, Aisha S. Azim <[email protected]>wrote:

Hi,

This tutorial is great, and so is sniffex.c. EXCEPT, it doesn't print the payloads of packets when using UDP. Why ever not? It'll print out packet number, source/destination, protocol, but never the hex/ascii payload and header of the packet itself. Help!

Thanks Aisha S. Azim

Talkback:160/okopnik.html

[ In reference to "The Unbearable Lightness of Desktops: IceWM and idesk" in LG#160 ]

mrfrank [czechfox at sbcglobal.net]

ATTN:Ben Okopnik

Just happened to run across the question in Linux Gazette Mar issue about idesk. Just so happens I finally got idesk up a couple hrs ago using your info at http://linuxgazette.net/160/okopnik.html then downloaded your idesk icon configurator. Small problem at this end.

Instead of showing icons, I get a series of 'Submit Query' in my browser. If I configure a link by clicking on one of the 'Submit Query's it makes my link just fine, even giving me the complete path to the .png file & the .png on the desktop link, even though the icon itself is no shown in browser.

This is a sample of the source code showing icon names;

<h3 style="background-color: blue; color: white; text-align: center">Click  
on an icon to create the IDesk link file</h3>
<input type="image" name="24x24/places/network-server.png"  
src="24x24/places/network-server.png" vspace="2" hspace="3" />
<input type="image" name="24x24/places/folder-saved-search.png"  
src="24x24/places/folder-saved-search.png" vspace="2" hspace="3" />
<input type="image" name="24x24/places/user-home.png"  
src="24x24/places/user-home.png" vspace="2" hspace="3" />
<input type="image" name="24x24/places/gnome-fs-trash-empty.png"  
src="24x24/places/gnome-fs-trash-empty.png" vspace="2" hspace="3" />

<input type="image" name="24x24/places/user-home.png"  
src="24x24/places/user-home.png" vspace="2" hspace="3" />

so it looks like this:

<input type="image" name="24x24/places/user-home.png"  
src="/home/xxxxxxx/.idesktop/builder/24x24/places/user-home.png"  
vspace="2" hspace="3" />

Any ideas how I can get the full path to show in the html?

Thank you.

later xxxxxxx

[ Thread continues here (2 messages/3.64kB) ]

Talkback: Discuss this article with The Answer Gang

2-Cent Tips

2-cent tip - rpath

Oscar Laycock [oscar_laycock at yahoo.co.uk]

This all started when I wanted to run the latest version of Firefox. I decided to build my own from source. But my kernel, assembler, compiler and C library were too old - in fact, nine years old. So I built new ones under the /usr/local directory. I used the Linux From Scratch book as a guide.

Now when I build new programs, I set the GCC compiler's "rpath" option to point to the libraries in /usr/local rather than in the usual /lib and /usr/lib. The rpath is a list of libraries at the start of a program that can tell Linux where to look for shared libraries when Linux runs a program. A program called the "dynamic linker" does the job. On my system it is "/lib/ld-linux.so.2". You can see a program's rpath by running a command such as "readelf -a /bin/ls". Of course, normally there isn't one. Also you can watch the dynamic linker at work using the "ldd" command. I set GCC's rpath by including it in the CFLAGS environment variable when configuring programs before building them. (You typically type "configure", "make" and "make install" to build a program.) I found a small number of programs ignore CFLAGS, so I made the gcc program a shell script, which then calls the real gcc with the right rpath option.

So I can now run old commands such as "ls" and "find" alongside new programs such as the KDE suite. The now eleven-year-old commands run fine on top of the recent kernel. I also put /usr/local/bin at the start of my path. This may be a security risk but my PC is not connected to the internet or a network.

There is a bit more too it. So here is the CFLAGS setting I used only few days ago:

export CFLAGS="-O2 -I. -I.. -I/usr/local/myglibc27/include -I/usr/local/include
-L/usr/local/myglibc27/lib -L/usr/local/lib -L/usr/local/lib/gcc/i686-pc-linux-gnu/4.2.3 -Wl,-rpath=/usr/local/myglibc27/lib:/usr/local/lib:/usr/local/lib/gcc/i686-pc-linux-gnu/4.2.3,-dynamic-linker=/usr/local/myglibc27/lib/ld-linux.so.2 -specs=/home/oscar/tmp/glibc/myspecs08scr -march=pentium2"

I also similarly set these environment variables: LDFLAGS, CXXFLAGS, CPPFLAGS, LIBCFLAGS, LIBCXXFLAGS. You can see that the include file path (-I's) and libraries path (-L's) match the rpath. The "-I. -I.." is there because some programs need to look at the header files in the build directory first - a bit of a quick fix. Notice how I now have two separate dynamic linkers on my PC. I had to edit the compiler specs file a little. Here is a section to really confuse you:

*startfile:
%{!shared: %{pg|p|profile:/usr/local/myglibc27/lib/gcrt1.o%s;pie:/usr/local/myglibc27/lib/Scrt1.o%s;:/usr/local/myglibc27/lib/crt1.o%s}}��� /usr/local/myglibc27/lib/crti.o%s %{static:/usr/local/lib/gcc/i686-pc-linux-gnu/4.3.2/crtbeginT.o%s;shared|pie:/usr/local/lib/gcc/i686-pc-linux-gnu/4.3.2/crtbeginS.o%s;:/usr/local/lib/gcc/i686-pc-linux-gnu/4.3.2/crtbegin.o%s}

I think this is choosing which C runtime code to put at the start of the program.

And here is the shell script that stands in for gcc:

[ ... ]

[ Thread continues here (2 messages/4.00kB) ]

Two-cent Tip: bash script to create animated rotating mark

Mulyadi Santosa [mulyadi.santosa at gmail.com]

During my boring saturday, I was thinking to create simple animated cycling mark. Here's the script:

$ while(true); do for a in \\ \| \/ -; do echo -n $a; sleep 1 ; echo
-n -e \\r ; done; done

Notice the usage of escaped "\r" (carriage return) and "-n" option to display continous marks at the same line and at the same column

-- 
regards,

blog: the-hydra.blogspot.com training: mulyaditraining.blogspot.com

[ Thread continues here (5 messages/5.54kB) ]

2-cent Tip: Load Python modules at Startup

Amit Saha [amitsaha.in at gmail.com]

Hello:

I have been using CPython as a calculator, while I do all those number crunching in C. SO, 'import math' is a must.

This is what I did:

- Create a file: .pythonrc in my $HOME and place this line:

   import math

- Now in your BASH, .bashrc or similar: export PYTHONSTARTUP= $HOME/.pythonrc

Everytime you start Python interactively, you should have the 'math' module already imported.

$ python
Python 2.6.4rc1 (r264rc1:75270, Oct 10 2009, 02:40:56)
[GCC 4.4.1] on linux2
Type "help", "copyright", "credits" or "license" for more information.


>>> math.pi

3.1415926535897931

Hope this helps.

-- 
Journal: http://amitksaha.wordpress.com,
�-blog: http://twitter.com/amitsaha

Two-cent Tip: Detexify

Jimmy O'Regan [joregan at gmail.com]

http://detexify.kirelabs.org/classify.html

Nifty website: draw the TeX symbol you were thinking of, and it tells you which one it (probably) is. Source (MIT license) here: http://github.com/kirel/detexify

[ Thread continues here (3 messages/1.80kB) ]

2-cent Tip: Piping to GNU Plot from C

Amit Saha [amitsaha.in at gmail.com]

Hello TAG:

Can this be a possible 2-cent tip?

Couple of things first up:

* GNU plot supports piping, So, echo "plot sin(x)" | gnuplot will plot the sin(x) function.

* However, the plot disappears even before you could see it. For that echo "plot sin(x)" | gnuplot -persist , is useful. It persists the GNU plot main window

The usefulness of the second point is that, if you have a "pipe descriptor" describing a pipe to the open GNU plot instance , you can plot more plots on the first plot, without opening a new GNU plot instance. We shall be using this idea in our code.

#include <stdio.h>
#define GNUPLOT "gnuplot -persist"
 
int main(int argc, char **argv)
{
        FILE *gp;
        gp = popen(GNUPLOT,"w"); /* 'gp' is the pipe descriptor */
        if (gp==NULL)
           {
             printf("Error opening pipe to GNU plot. Check if you have it! \n");
             exit(0);
           }
 
        fprintf(gp, "set samples 2000\n");
        fprintf(gp, "plot abs(sin(x))\n");
        fprintf(gp, "rep abs(cos(x))\n");
        fclose(gp);
 
return 0;
}

The above code will produce a comparative plot of absolute value of sin(x) and cos(x) on the same plot. The popen function call is documented at http://www.opengroup.org/pubs/online/7908799/xsh/popen.html. This code/idea should work on GCC and Linux and any other language and OS that supports piping.

Utility: If you have a application which is continuously generating some data, which you will finally plot, then you can plot the data for every new set of data- that gives a nice visualization about how the data is changing with the iterations of your application. This is a perfect way to demonstrate convergence to the best solutions in Evolutionary Algorithms, such as Genetic Algorithms.

Best, Amit

-- 
Journal: http://amitksaha.wordpress.com,
�-blog: http://twitter.com/amitsaha

[ Thread continues here (4 messages/7.26kB) ]

Talkback: Discuss this article with The Answer Gang

News Bytes

By Deividson Luiz Okopnik and Howard Dyckoff

Contents: News in General Conferences and Events Distro News Software and Product News

Selected and Edited by Deividson Okopnik

Please submit your News Bytes items in plain text; other formats may be rejected without reading. [You have been warned!] A one- or two-paragraph summary plus a URL has a much higher chance of being published than an entire press release. Submit items to [email protected]. Deividson can also be reached via twitter.

Talkback: Discuss this article with The Answer Gang

Howard Dyckoff is a long term IT professional with primary experience at Fortune 100 and 200 firms. Before his IT career, he worked for Aviation Week and Space Technology magazine and before that used to edit SkyCom, a newsletter for astronomers and rocketeers. He hails from the Republic of Brooklyn [and Polytechnic Institute] and now, after several trips to Himalayan mountain tops, resides in the SF Bay Area with a large book collection and several pet rocks.

A Short CGI Script for Passing 404s to Another Server

By Silas Brown

There are several reasons why you might have more than one website. For example, you might have an account on a fast server with limited quota, and another on a server that is slower but has more space, and you keep certain rarely-used files on the slower server. Or you might have an easy-to-remember dyndns alias pointing to your home server, but some pages on another server that has a better connection but a longer address. If this is the case, then it might be useful to set one of the servers to forward any requests for pages it can't find to the other server. Then you can use this first server in all the addresses you type, and if the page requested is on it, it will be returned, otherwise the browser will be automatically redirected to check the second server for the same page. (Don't set that second server to redirect back to the first or you'll have an infinite loop.)

Using Apache, this can be accomplished as follows. Firstly, put the following line into your public_html/.htaccess file, replacing /~your-user-ID as appropriate (if you have your own hostname then you can omit it):

ErrorDocument 404 /~your-user-ID/cgi-bin/redirect.cgi

This tells Apache to run the script whenever it would normally issue a 404. If you want the redirection to take effect for index.html also (that is if index.html is missing), then you should also add:

Options -Indexes
ErrorDocument 403 /~your-user-ID/cgi-bin/redirect.cgi

This prevents directory listing, and tells Apache to send the browser to the redirection script instead of showing the "Forbidden" message. This arrangement might be useful if your main site is on the other server but this server contains a few extra files that you want Apache to check for first.

Then create public_html/cgi-bin/redirect.cgi as follows, replacing other-site-address-goes-here and ~your-user-ID as appropriate:

#!/bin/bash
echo Status: 301 Moved
echo Location: "$(echo "$REQUEST_URI"|sed -e s,/,http://other-site-address-goes-here/, -e s,~your-user-ID/,,)"
echo

Then chmod +x public_html/cgi-bin/redirect.cgi and test.

The above script should work regardless of whether or not ~your-user-ID is included in the request, that is, if Apache is serving your public_html on your own hostname then this script should work regardless of whether the incoming request is for your hostname or for your user ID on the main hostname.

Some antiquated browsers might not follow the 301 redirect. If this is a concern then you can add some text in the response as well. But the script as it stands is small and easily set up, and should work without trouble in most cases.

Talkback: Discuss this article with The Answer Gang

Away Mission - Upcoming in November '09

By Howard Dyckoff

It's nearly time to retire 2009, yet many conferences for the Linux and open source community are just about to happen. A plethora will happen in the first week of November, so this preview may give you ideas for 2010.

Here's a short list of what's on offer:

LISA '09 - 23rd Large Installation System Administration Conference

Nov 1-6, 2009, Marriott Waterfront Hotel, Baltimore, MD
http://www.usenix.org/events/lisa09/

Cloud Computing & Virtualization 2009

Nov 2-4, Santa Clara Convention Center, Santa Clara, CA
https://www3.sys-con.com/cloud1109/registernew.cfm

iPhone Developer Summit

Nov 2-4, Santa Clara, CA
http://www.iphonedevsummit.com/

Enterprise 2.0 Fall 2009

Nov 2-5, Moscone Center, San Francisco, CA
http://www.e2conf.com/sanfrancisco/

VoiceCon-SF 2009

Nov 2-5, 2009, San Francisco
http://www.voicecon.com/sanfrancisco/

ApacheCon

Nov. 2-6, Oakland, CA
http://us.apachecon.com/c/acus2009/

2nd Annual Linux Foundation End User Summit

Nov 9-10, 2009, Jersey City, NJ
http://events.linuxfoundation.org/events/end-user-summit/

QCon - San Francisco

Nov 18-20, SF, CA
http://qconsf.com/sf2009/

As you may know, the End User Summit organized by the Linux Foundation is by invitation only. The event is closed to the press, as was the first summit, in 2008. The Summit is primarily an opportunity for big corporate end-users to meet with leaders from within the Linux community, including the highest-level maintainers and developers. The event was originally created at the request of the Linux Foundation's Technical Advisory Board.

The Summit will take place November 9-10, 2009, at the Hyatt Jersey City on the Hudson. Its location, just off the Exchange Place PATH Station, will give corporate Linux users quick access from East Coast hubs.

There will be sessions on open source legal fronts, Linux and the challenges of a down economy, tracing and performance management, scaling Linux to 4096 processors, and other technical sessions.

LISA '09, the annual sysadmin conference organized by USENIX, starts the month off in Baltimore. Locals in the Baltimore-DC area can get the new LISA Evening Pass. This allows access after work for conference receptions, the Vendor Exhibition on Wednesday only, the popular Birds of a Feather sessions (BoFs), and more.

There will be sessions on cfEngine, VMware vSphere 4.0, and zero-emission data centers. Cutting-edge practices and new or developing work are presented in the paper presentations and the poster sessions. Check out the list of accepted posters at http://www.usenix.org/events/lisa09/posters.html. Also, all registered LISA '09 attendees are eligible to received a Google Wave Sandbox invitation.

Here's an overview of LISA '09: http://www.usenix.org/events/lisa09/ataglance.html

Here's a link to past LISA conference proceedings: http://www.usenix.org/events/bytopic/lisa.html

ApacheCon '09 will be held in Oakland, CA; this year is also the 10th anniversary of the Apache Foundation. The venue is the Marriott City Center Hotel (and that's a very safe location). Last year, it was held in New Orleans.

The first 2 days include half-day tutorials. There are also several free events paralleling the conference tutorial days, including the Hackathon and the BarCamps.

Apache Hackathon is held over the two days prior to the main three-day conference, November 2-3, and everyone can join in this free codefest event. Coinciding with the traditional Hackathon, BarCampApache is open to the public, and the schedule will be determined by the participants. Topics at BarCampApache need not be Apache-specific.

Here's a conference overview: http://www.us.apachecon.com/c/acus2009/schedule/grid/

There will be several BoFs or MeetUps at ApacheCon '09, and the current list is here: http://www.us.apachecon.com/c/acus2009/schedule/meetups/

Video recordings from the keynotes at last year's ApacheCon are here: http://streaming.linux-magazin.de/en/archive_apachecon08us.htm

The theme for Enterprise 2.0 San Francisco is "unlocking the business value of Enterprise 2.0", with an emphasis on how real customers are using enterprise 2.0 social media to enable a more efficient, agile, and highly productive workforce.

Conference topics include, among others:

• Social Networking in Business
• Social Networks as New Media
• Social Messaging & Twitter
• Enterprise Mash-ups
• Enterprise Software
• Enterprise RSS & Syndication

Since this is a new conference, you might want to take a quick look at the keynotes from June's Enterprise 2.0 conference in Boston: http://www.e2conf.com/e2tv/

Enterprise 2.0 San Francisco is held jointly with VoiceCon, at the Moscone Center this year. Registration for one event allows access to the other.

QCon SF. This is the third annual San Francisco incarnation of the enterprise software development conference, and its focus is on developer team leads, architects and agile project management. This is primarily an open source conference, and has separate Ruby and Java tracks.

Organized jointly by the JAOO conference and the InfoQ.com Enterprise Software Development Community (original creators of TheServerSide.com and Symposium conferences), QCon is "...organized by the community, for the community."

One unique aspect of QCon is its Euro-centric origins. Many - but not most - of the speakers are leading developers and architects from Europe, and QCon is a great opportunity to meet peers from across the Atlantic without a passport and plane ticket. The global prospective is refreshing, and there is no overbearing presence of giant software vendors. All content, no spin.

QCon speakers include Martin Fowler, Amazon.com CTO Werner Vogels, Spring creator Rod Johnson, Hibernate creator Gavin King, Scrum co-founder Jeff Sutherland, LINQ inventor Erik Meijer, JRuby core developer Ola Bini, and others.

Here's a list of the conference tracks by day: http://qconsf.com/sf2009/tracks/

And here's a link to the archives from prior QCons, so you can see for yourself: http://qcon.infoq.com/archives/

Happy Thanksgiving, and a Happy New Year to all!

Talkback: Discuss this article with The Answer Gang

Howard Dyckoff is a long term IT professional with primary experience at Fortune 100 and 200 firms. Before his IT career, he worked for Aviation Week and Space Technology magazine and before that used to edit SkyCom, a newsletter for astronomers and rocketeers. He hails from the Republic of Brooklyn [and Polytechnic Institute] and now, after several trips to Himalayan mountain tops, resides in the SF Bay Area with a large book collection and several pet rocks.

Setting up a MySQL Cluster for your Linux desktop

By Mark Nielsen

1. Introduction
2. Row Level Replication and why it frustrates me
3. Misc Setup
4. General Approach
5. Configuration Files
6. Installing MySQL Cluster on Ubuntu
7. Setting up and Testing Replication to a Slave and Cluster Slave
8. Let's test MySQL
9. Installing Stored procedure on Slaves and Masters
10. Conclusion and Future Topics
11. Some SSD preliminary testing

Introduction

MySQL Cluster has come a long way in the 4 years since I experimented with it. Compared to when I first got the cluster working on my home computer, I didn't have to change much get the latest version up and running.

So, what is MySQL Cluster? It is a database solution which tries to solve high availability issues by using multiple synchronous masters (for the lack of a better phrase) in a shared-nothing environment. In order to solve a lot of the latency issues associated with multiple masters, it keeps all the indexed data in memory to enable fast processing of the data.

So then what is the purpose of this article? To let you setup a MySQL Cluster for fun, on a single box, so you can test out its features¹. The main things I wanted to test out are:

• Setting up a cluster on one or two boxes. People tell me this is impossible. For personal and business reasons, this article attempts to squash those rumors so I can just point people to this article instead of explaining the details to them. There are advantages to setting up a cluster on 3 boxes as compared to 2 boxes, however. There is a difference between a network outage and a single server becoming unavailable and 3 servers help with this.
• Using other storage engines and setup dangerous replication between the frontend MySQL servers.
• Replicating data to a cluster slave as well as a normal slave.

There is also one important item to keep in mind about MySQL Cluster. I have no idea why they make "MySQL Cluster" as a separate product from the main MySQL server. In the past, the "MySQL Cluster" was just the NDB engine running in MySQL. It seems to be the same way even now and I suspect that it might just be due to politics and marketing that they have something called a "MySQL Cluster" when it is really just MySQL with the NDB engine. The mysqld server running in "MySQL Cluster" can still store data in InnoDB and MyISAM formats.

For more info about MySQL Cluster, refer to:

• http://dev.mysql.com/doc/refman/4.1/en/mysql-cluster.html
• http://www.mysql.com/products/database/cluster/
• For non-cluster MySQL, I highly recommend the "percona" version. I'd like to see a MySQL Cluster supported version of MySQL. Download MySQL from http://www.percona.com/docs/wiki/release:start

1. Data Node: Where all the data is stored.
2. MySQL frontend server: The MySQL frontend doesn't store any data itself (unless you really want it to, like I do). It handles the processing of the queries and gets the data from the data nodes.
3. Management Node: Manages all the nodes.

• Read the Limitations.
• Specifically: http://dev.mysql.com/doc/refman/5.1/en/mysql-cluster-limitations-unsupported-missing.html
• Statement level replication: http://dev.mysql.com/doc/refman/5.1/en/mysql-cluster-replication-general.html
• MySQL Cluster replication issues: http://mirror.facebook.net/mysql/doc/refman/5.1/en/mysql-cluster-replication-issues.html

Row Level Replication and why it frustrates me

MySQL copies data from one computer to another computer through MySQL Replication. The old way of replicating data was to copy over and execute all the SQL commands from one MySQL server to the other MySQL servers. This type of replication was called "Statement Level Replication.". This copying and executing of queries was NOT real-time and had some issues. Basically, it was like this:

• Query is executed on the master. Let's say the query took 30 seconds to finish.
• Query is saved to a log and then each slave server downloads the query. There is a time delay in downloading the query. Usually it isn't very much.
• Query gets executed on the slave servers. From the point the slaves start to execute the query, they will take an equal amount of time to execute the query as the master did. If the master took 30 seconds to execute an insert, update, or delete command, it will take 30 seconds for each slave to do the same (usually). Thus, right before a slave finishes the 30 second query, it is behind in replication by 30 seconds, or rather its data won't be in sync until 30 seconds later (after the query is finished).

For most practical purposes, most small queries finish very fast and replication is also very fast.

So what is Row level Replication? Basically, the SQL query which executes the changes on the master is forgotten and we only record the actual row differences and copy them over. For example, let us say that a query updates a table and sets the country code to 'US' for all entries that are NULL. In a table with 1 million rows, let's say only 10 rows were NULL and thus 10 rows get updated. The difference between Statement and Row Level Replication is the following:

• Statement Replication will copy the query and execute it.
• Row Replication will copy only the changes for those 10 rows and ignore what the original query is.

Row Level Replication doesn't care what the SQL query was. It only looks at the data that has changed and copies those changes to the slaves.

Why does this frustrate me? For a few reasons:

1. Everything is moving towards Row Level Replication and Statement Level Replication is being made incompatible under some conditions. I think both are very powerful and you should let the end user decide what they want to use. I understand that Statement Level Replication can cause problems (under some circumstances), so my attitude is that they should work around the problems. The power retained by keeping both as an option at all times is the ability to customize how you do things. I have been in many situations where a slight improvement in performance meant a big deal. Thus, you want all your options open.
2. Also, Row Level Replication is MUCH HARDER to debug. Statement Level Replication is very easy to debug. I have seen weird replication issues with MySQL in the past and I don't like to be in a situation where I can't debug things easily. Everything about being a DBA is debugging database problems and finding out where the problem is. All the advantages of Row Level Replication are almost irrelevant if you can't debug issues in replication.

I am glad we have Row Level Replication as an option and it can be nice to use depending on your needs. I just wish Statement Level Replication could be used if handled correctly under all all configurations. To give you a better perspective of why Statement Level Replication is nice to debug, we had a weird case one time where 1 byte in a replication stream would change for no reason over a WAN. If it changed in the data, nobody would notice (that's really bad). But if it changed in one of the SQL words, replication would break. We were easily able to compare the master's binlogs to the slave's binlogs and find out what the problem was. I remain sceptical how easy it will be to verify why things break with row level replication. Maybe I am just paranoid, but you have to be when managing databases.

Misc Setup

• First computer: Laptop with 3 GB of RAM running Ubuntu.
• Second computer: Desktop with 2 GB of RAM running Ubuntu 9 and also using a Vortex SSD hard drive. I have an external drive enclosure so I could test the SSD hard drive on other systems.
• Third system: 2 GB server with CentOS.

• It's interesting to note I only get a max of 30 MB/sec transfer rate from my laptop hard drive to the Vortex drive. I think it's because of the speed of USB and also the internal hard drive. I've read that USB 2.0 maxes out at 40 MB/sec. In theory, if I attach the Vortex drive directly to the motherboard on my desktop computer, it should go a lot faster. In case you wonder why I am mentioning this, at the time of this article, the Intel X25-e SSD drive looks very promising in regard to database use. I expect that 90% of the databases using MySQL will be using SSD drives in the next few years, because most MySQL servers have less than 100 GB of data. Also, the SSD hard drives will get faster and bigger. I bought the Vortex because it was cheaper and I just wanted to test SSD out in general and later use it for my laptop. If it works out well, I'll buy the Intel X25-e and put it in my desktop and beat the daylights out of it running database simulations.
• Also, when I connected the Vortex SSD hard drive to the motherboard on my desktop, I could only get about 35 MB/sec written to the SSD hard drive when reading files from cache. It seems to go faster than the new 1.5 TB SATA hard drive I bought though, which was about 23 MB/sec if I remember right.
• Have a Gb free of diskspace. You don't need all of that, maybe 300 MB at most.
• Of course, you should have Python installed. You don't really need Python installed for anything in this article, but you should have it installed anyways. (353, 413, 409, 521, 482)

/sbin/ifconfig lo:1 127.0.0.101 netmask 255.0.0.0
/sbin/ifconfig lo:2 127.0.0.102 netmask 255.0.0.0
/sbin/ifconfig lo:3 127.0.0.103 netmask 255.0.0.0
/sbin/ifconfig lo:4 127.0.0.104 netmask 255.0.0.0

As a side note, it is interesting to see how Ubuntu is taking off on the desktop market and how it is starting to creep into the server market. For me, I just like to use the same operating system for the desktops and servers because then I know that the operating system is flexible and can work under multiple environments. After all, isn't a desktop just a server running Xwindows with OpenOffice and FireFox? To me, the words "desktop" and "server" are just marketing terms.

General Approach

Here's the setup for our article:

1. Add 4 IP addresses to your box: 127.0.0.101, 127.0.0.102, 127.0.0.103, 127.0.0.104. This can be accomplished through standard linux networking commands.
2. For each IP address, bind a data node, mysqld node, and management node to its default ports. Each IP address will have its own set of nodes.
3. One management node will be a backup management node.
4. Define NDB databases with a prefix to make it easy to separate NDB databases from InnoDB and MyISAM databases.
5. Setup replication between the MySQL servers on 127.0.0.101 and 127.0.0.102 in dual master mode excluding replication of the NDB databases (the cluster takes care of the NDB databases).
6. Setup replication to a cluster slave and a normal slave using Row Level Replication. You could do Statement Level Replication, but then all writes would have to go to exactly one of the mysqld servers in the Cluster.

• Special Rep Config: The MySQLd nodes need to replicate non-NDB data.
• Hacks Necessary: The stored procedure I use is needed to start replication if a MySQLd node gets restarted.

Problems:

1. Inconsistent data on non-NDB tables.
2. Replication configurations are needed to avoid duplicate entries in the binlogs. Non-NDB data replicated on the MySQLd servers will need to be setup in a Replication Circle.

Weird Purposes:

1. Keep log data out of cluster to keep the data small in the cluster.
2. Use Read-Repeatable mode in InnoDB on the normal slave. This can be useful for *snapshot* reports.

Configuration Files

• There is one "config.ini" file which is needed by all the servers. It is here: config.ini.txt.
• MySQL config file for Master1 mysql node1 server. It is here: my.cnf_node1.txt.
• MySQL config file for Master2 mysql node2 server. It is here: my.cnf_node2.txt.
• MySQL config file for Slave1 mysql server. It is here: my.cnf_slave1.txt.
• Cluster Slave1 config.ini file. It is here: config.ini_cslave1.txt.
• MySQL config file for Cluster Slave1 mysql server. It is here: my.cnf_cslave1.txt.

These files I use to stop and start MySQL. I put it in /usr/local/mysql_cluster.

• Commands I used to setup MysQL. It is here: commands.txt.

Installing MySQL Cluster on Ubuntu

There are 3 ways to install MySQL: Do-It-Yourself, get it from a 3rd party, or from MySQL. It is unlikely that the official website for MySQL will have consistent downloads for Ubuntu versions. We'll see if things change in the future. You have two real options: Download and compile it yourself or download from a third party. I'll explain both here:

Download and Install MySQL

tar -zxvf cfiles.tgz

Note: The download steps below only work for a specific version of MySQL. Please refer to http://dev.mysql.com/downloads/cluster/ for newer versions.

Use my install script to install MySQL. You can read the various comments in the script. Please read the comments before you execute each section in the script. In general, it will attempt to setup three things:

1. A MySQL Cluster
2. A regular slave
3. A MySQL cluster slave

• You must have sudo access. If not, just log in as the root user. For example: su -l root
• Download the script.
• mv install.txt install.sh
• bash install.sh # However, I recommend only doing sections at a time. I got the script to work all the way through, but this script doesn't allow for errors, so it is safer to just do sections at a time. Copy and paste sections into a terminal session.

Setting up and Testing Replication to a Slave and Cluster Slave

NOTE: We only use Row Level Replication. It is possible to do this with Statement Level Replication, but it gets more complicated.

Look at my slave setup script.

Let's test MySQL

./bin/mysql -u root -S instances/mysqlnode1/tmp/mysql.sock -e "create database if not exists ndb"

./bin/mysql -u root -S instances/mysqlslave1/tmp/mysql.sock -e "show databases"
./bin/mysql -u root -S instances/mysqlslave1/tmp/mysql.sock -e "create database if not exists ndb"

./bin/mysql -u root -S instances/mysqlslave1/tmp/mysql.sock -e "create table if not exists ndb.ndb1 (i int) engine=InnoDB;"
./bin/mysql -u root -S instances/mysqlnode1/tmp/mysql.sock -e "create table if not exists ndb.ndb1 (i int) engine=ndb;" 
./bin/mysql -u root -S instances/mysqlnode1/tmp/mysql.sock -e "create table if not exists rep.rep1 (i int) engine=InnoDB;" 

./bin/mysql -u root -S instances/mysqlnode1/tmp/mysql.sock -e "insert into ndb.ndb1 values (1);" 
./bin/mysql -u root -S instances/mysqlnode2/tmp/mysql.sock -e "insert into rep.rep1 values (1);"

# These tables should contain one row each. 
./bin/mysql -N -u root -S instances/mysqlnode1/tmp/mysql.sock -e "select * from ndb.ndb1"
./bin/mysql -N -u root -S instances/mysqlnode1/tmp/mysql.sock -e "select * from rep.rep1"
./bin/mysql -N -u root -S instances/mysqlnode2/tmp/mysql.sock -e "select * from ndb.ndb1"
./bin/mysql -N -u root -S instances/mysqlnode2/tmp/mysql.sock -e "select * from rep.rep1"
./bin/mysql -N -u root -S instances/mysqlslave1/tmp/mysql.sock -e "select * from ndb.ndb1"
./bin/mysql -N -u root -S instances/mysqlslave1/tmp/mysql.sock -e "select * from rep.rep1"
./bin/mysql -N -u root -S instances/mysqlcslave1/tmp/mysql.sock -e "select * from ndb.ndb1"
./bin/mysql -N -u root -S instances/mysqlcslave1/tmp/mysql.sock -e "select * from rep.rep1"

# Test replication for the non-replicating database. 
./bin/mysql -u root -S instances/mysqlnode1/tmp/mysql.sock -e "create table if not exists norep.norep1 (i int) engine=InnoDB;"
./bin/mysql -u root -S instances/mysqlnode2/tmp/mysql.sock -e "create table if not exists norep.norep1 (i int) engine=InnoDB;"
./bin/mysql -u root -S instances/mysqlslave1/tmp/mysql.sock -e "create table if not exists norep.norep1 (i int) engine=InnoDB;"
./bin/mysql -u root -S instances/mysqlcslave1/tmp/mysql.sock -e "create table if not exists norep.norep1 (i int) engine=InnoDB;"

./bin/mysql -u root -S instances/mysqlnode1/tmp/mysql.sock -e "insert into norep.norep1 values (1);"
./bin/mysql -u root -S instances/mysqlnode2/tmp/mysql.sock -e "insert into norep.norep1 values (2);"
./bin/mysql -u root -S instances/mysqlslave1/tmp/mysql.sock -e "insert into norep.norep1 values (3);"
./bin/mysql -u root -S instances/mysqlcslave1/tmp/mysql.sock -e "insert into norep.norep1 values (4);"

# These tables should contain one unique value.
./bin/mysql -N -u root -S instances/mysqlnode1/tmp/mysql.sock -e "select * from norep.norep1"
./bin/mysql -N -u root -S instances/mysqlnode2/tmp/mysql.sock -e "select * from norep.norep1"
./bin/mysql -N -u root -S instances/mysqlslave1/tmp/mysql.sock -e "select * from norep.norep1"
./bin/mysql -N -u root -S instances/mysqlcslave1/tmp/mysql.sock -e "select * from norep.norep1"

Installing Stored procedure on Slave

When a MySQLd server using NDB restarts, it sends out an error message. It is not usually a good idea to restart replication without knowing what is going on and checking the logs. However, there may be times when you don't care. For that case, I have created a stored procedure using an event to kick off replication when a known error occurs. This stored procedure only works for the first 5 days. If you wish, you can modify it for a longer period of time.

Let's kill replication, install the stored procedure, and then see if it fixes replication.

1. Download this file: files/stored_procedure_rep.sp

2. if [ -e "instances/mysqlnode1/tmp/mysqld.pid" ]; then
    k=`cat instances/mysqlnode1/tmp/mysqld.pid`
    echo "Killing $k"
    kill $k
   fi
   if [ -e "instances/mysqlnode2/tmp/mysqld.pid" ]; then
    k=`cat instances/mysqlnode2/tmp/mysqld.pid`
    echo "Killing $k"
    kill $k
   fi
   
   sleep 2
   if [ -e "instances/mysqlnode1/tmp/mysqld.pid" ]; then
    echo "Mysql node 1 still running."
   else
    echo "starting node 1 mysql"
    sudo -u mysql nohup ./libexec/mysqld --defaults-file=etc/my.cnf_node1 &
   fi
   if [ -e "instances/mysqlnode2/tmp/mysqld.pid" ]; then
    echo "Mysql node 2 still running."
   else
    echo "starting node 2 mysql"
    sudo -u mysql nohup ./libexec/mysqld --defaults-file=etc/my.cnf_node2 &
   fi
   
   

3. Add the following to the config files if it is not already done: "event_scheduler=ON". The config files in this article already have that setting. If you had to add this to your config, restart the MySQLd servers.
4. Check replication. One of the two fields should be 'No'.

   ./bin/mysql -u root -S instances/mysqlnode1/tmp/mysql.sock -e "show slave status\G" | grep -i running
   ./bin/mysql -u root -S instances/mysqlnode2/tmp/mysql.sock -e "show slave status\G" | grep -i running
   ./bin/mysql -u root -S instances/mysqlslave1/tmp/mysql.sock -e "show slave status\G" | grep -i running
   ./bin/mysql -u root -S instances/mysqlcslave1/tmp/mysql.sock -e "show slave status\G" | grep -i running
   

5. Install the stored procedure below on the slave.

   ./bin/mysql -u root -S instances/mysqlnode1/tmp/mysql.sock -e "source stored_procedure_rep.sp" mysql
   ./bin/mysql -u root -S instances/mysqlnode2/tmp/mysql.sock -e "source stored_procedure_rep.sp" mysql
   ./bin/mysql -u root -S instances/mysqlslave1/tmp/mysql.sock -e "source stored_procedure_rep.sp" mysql
   ./bin/mysql -u root -S instances/mysqlcslave1/tmp/mysql.sock -e "source stored_procedure_rep.sp" mysql
   

6. Wait 1 minute. Slave_IO_Running and Slave_SQL_Running should be 'Yes'.

   ./bin/mysql -u root -S instances/mysqlnode1/tmp/mysql.sock -e "show slave status\G"
   ./bin/mysql -u root -S instances/mysqlnode2/tmp/mysql.sock -e "show slave status\G"
   ./bin/mysql -u root -S instances/mysqlslave1/tmp/mysql.sock -e "show slave status\G"
   ./bin/mysql -u root -S instances/mysqlcslave1/tmp/mysql.sock -e "show slave status\G"
   
   # or just
   
   ./bin/mysql -u root -S instances/mysqlnode1/tmp/mysql.sock -e "show slave status\G" | grep -i running
   ./bin/mysql -u root -S instances/mysqlnode2/tmp/mysql.sock -e "show slave status\G" | grep -i running
   ./bin/mysql -u root -S instances/mysqlslave1/tmp/mysql.sock -e "show slave status\G" | grep -i running
   ./bin/mysql -u root -S instances/mysqlcslave1/tmp/mysql.sock -e "show slave status\G" | grep -i running
   
   

7. "Show slave status" should have both the io_thread and sql_thread as 'Yes' after the stored procedure kicks in.

Conclusion and Future Topics

Here are some future topics:

1. Stress test MySQL Cluster on a single box and get it to break. Hint: The config files in this article were made so that I could load real production data on a server at work without having the cluster kick out the data nodes.
2. Break MySQL Cluster by pounding it to death or filling up diskspace or filling up the index in memory and see what happens.
3. Split the configurations out to two servers.
4. Setup Statement Level Replication.
5. Perform backups via cluster or slave.
6. Managing the cluster and Nagios alarms for the cluster to detect when things go bad.
7. Adding and dropping nodes manually.
8. Creating Ubuntu and CentOS packages.
9. Testing my SSD drive.

Some SSD Preliminary Testing

I loaded data onto both the SSD and the SATA drives I have. I have separate MySQL Cluster installations on each hard drive. My computer is from late 2004, so it probably only a SATA I bus. So this is a really bad test for testing real performance, but I just wanted to get a rough idea of what I am dealing with. Both the SATA drive and the SSD drive were brand new. I was using an SSD Vertex 60 gig drive. I loaded a 200 MB file onto the system, but, there were 4 MySQL instances running. The two data nodes for the cluster, a slave, and a cluster slave. In addition, the binlogs would have a copy of all the data for the MySQLd instances. Thus, there are 8 copies of the data (4 in the databases, 4 in logs). I expect on a system with just one MySQL instance it would run a lot faster.

For cdrom copying times of 700 megs, look at this table. The SSD drives appears to be about 66% faster. However, when you compare this to the next table, SSD shows to be even better (as expected).

We can see that the MySQL Cluster is running 2 to 5 times faster on the SSD drive. This is not surprising if the latency on the SSD drive is better. The inserts below were at 1000 rows at a time. The python script has the detailed schema information. I compared 3 different points in the log files for the inserts. I compared the recent time for 1000 rows being inserted and the total time up to that point. The loading stopped at 1 million rows.

Query OK, 1000 rows affected (0.13 sec)
Records: 1000  Duplicates: 0  Warnings: 0

+-------------------------+--------+
| seconds_diff            | 122000 |
+-------------------------+--------+
|                      31 | 122000 |
+-------------------------+--------+
1 row in set (0.00 sec)

Query OK, 1000 rows affected (0.08 sec)
Records: 1000  Duplicates: 0  Warnings: 0

+-------------------------+--------+
| seconds_diff            | 122000 |
+-------------------------+--------+
|                      13 | 122000 |
+-------------------------+--------+
1 row in set (0.00 sec)

+-------------------------+--------+
| seconds_total           | 715000 |
+-------------------------+--------+
|                    1819 | 715000 |
+-------------------------+--------+
1 row in set (0.00 sec)

Query OK, 1000 rows affected (4.96 sec)
Records: 1000  Duplicates: 0  Warnings: 0

+-------------------------+--------+
| seconds_total           | 715000 |
+-------------------------+--------+
|                     353 | 715000 |
+-------------------------+--------+
1 row in set (0.00 sec)

Query OK, 1000 rows affected (1.18 sec)
Records: 1000  Duplicates: 0  Warnings: 0

Query OK, 1000 rows affected (4.46 sec)
Records: 1000  Duplicates: 0  Warnings: 0

+-------------------------+---------+
| seconds_diff            | 1000000 |
+-------------------------+---------+
|                    3188 | 1000000 |
+-------------------------+---------+
1 row in set (0.00 sec)

Query OK, 1000 rows affected (1.40 sec)
Records: 1000  Duplicates: 0  Warnings: 0

+-------------------------+---------+
| seconds_diff            | 1000000 |
+-------------------------+---------+
|                     743 | 1000000 |
+-------------------------+---------+
1 row in set (0.00 sec)

Execute the following in the MySQL cluster for the SSD drive and then later for the normal drive and then compare the times. Also, download my insert python script. I wrote it up very quick. It isn't that elegant.

Insert_200_megs_random.py > insert1.sql
bin/mysql -u root  -S  instances/mysqlnode1/tmp/mysql.sock

tee insert1.log
source insert1.log

[    1.338264] sata_promise 0000:00:08.0: version 2.12
[    1.338294] sata_promise 0000:00:08.0: PCI INT A -> GSI 18 (level, low) -> IRQ 18
[    1.352185] scsi0 : sata_promise
[    1.352328] scsi1 : sata_promise
[    1.352388] scsi2 : sata_promise
[    1.352430] ata1: SATA max UDMA/133 mmio m4096@0xf7a00000 ata 0xf7a00200 irq 18
[    1.352434] ata2: SATA max UDMA/133 mmio m4096@0xf7a00000 ata 0xf7a00280 irq 18
[    1.352436] ata3: PATA max UDMA/133 mmio m4096@0xf7a00000 ata 0xf7a00300 irq 18
[    1.672036] ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
[    1.707733] ata1.00: ATA-8: ST31500341AS, CC1H, max UDMA/133
[    1.707736] ata1.00: 2930277168 sectors, multi 0: LBA48 NCQ (depth 0/32)
[    1.763738] ata1.00: configured for UDMA/133
[    2.080030] ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
[    2.088384] ata2.00: ATA-7: OCZ-VERTEX 1275, 00.P97, max UDMA/133
[    2.088387] ata2.00: 125045424 sectors, multi 16: LBA48 NCQ (depth 0/32)
[    2.096398] ata2.00: configured for UDMA/133
[    2.252082] isa bounce pool size: 16 pages
[    2.252207] scsi 0:0:0:0: Direct-Access     ATA      ST31500341AS     CC1H PQ: 0 ANSI: 5
[    2.252323] sd 0:0:0:0: [sda] 2930277168 512-byte hardware sectors: (1.50 TB/1.36 TiB)
[    2.252342] sd 0:0:0:0: [sda] Write Protect is off
[    2.252345] sd 0:0:0:0: [sda] Mode Sense: 00 3a 00 00
[    2.252370] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[    2.252449] sd 0:0:0:0: [sda] 2930277168 512-byte hardware sectors: (1.50 TB/1.36 TiB)
[    2.252462] sd 0:0:0:0: [sda] Write Protect is off
[    2.252464] sd 0:0:0:0: [sda] Mode Sense: 00 3a 00 00
[    2.252487] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[    2.252492]  sda: sda1 sda2 sda4
[    2.282713] sd 0:0:0:0: [sda] Attached SCSI disk
[    2.282781] sd 0:0:0:0: Attached scsi generic sg0 type 0
[    2.282842] scsi 1:0:0:0: Direct-Access     ATA      OCZ-VERTEX 1275  00.P PQ: 0 ANSI: 5
[    2.282924] sd 1:0:0:0: [sdb] 125045424 512-byte hardware sectors: (64.0 GB/59.6 GiB)
[    2.282938] sd 1:0:0:0: [sdb] Write Protect is off
[    2.282941] sd 1:0:0:0: [sdb] Mode Sense: 00 3a 00 00
[    2.282963] sd 1:0:0:0: [sdb] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[    2.283011] sd 1:0:0:0: [sdb] 125045424 512-byte hardware sectors: (64.0 GB/59.6 GiB)
[    2.283024] sd 1:0:0:0: [sdb] Write Protect is off
[    2.283026] sd 1:0:0:0: [sdb] Mode Sense: 00 3a 00 00
[    2.283049] sd 1:0:0:0: [sdb] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA

gt;more /proc/scsi/sg/device_strs
ATA             ST31500341AS            CC1H
ATA             OCZ-VERTEX 1275         00.P
DVDRW           IDE1008                 0055

• http://www.anandtech.com/storage/showdoc.aspx?i=3531&p=25
• http://www.anandtech.com/storage/showdoc.aspx?i=3531&p=1
• http://www.anandtech.com/storage/showdoc.aspx?i=3535
• http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3403&p=11
• http://www.anandtech.com/storage/showdoc.aspx?i=3531&p=25

As a side note, SLC SSD drives last 10 times longer for writes comapared to the MLC SSD drives, from what I understand. Thus, the Intel x25-e is far more important than the Intel x25-m for database servers. You want the drives to last as long as possible and have the best random write performance. On one of my contracting jobs, because I showed them how the Vortex behaved on my home computer, I will get to test an Intel x25-e at their place. That's the second time when I invested in my computer at home it helped me in my career. The first time was getting an AMD 64 bit system before most companies had it and using it at home with Linux and MySQL. The ex complained it was a waste of money, but it was a hell of an impressive thing to say at the time during interviews that I had 64 bit Linux and MySQL running at home, and it proved to be true. If you get a good raise because you show initiative when investing money into your home equipment to prove something, it pays for itself.

Talkback: Discuss this article with The Answer Gang

Mark leads the Ubuntu California Team. You can reach him at [email protected] .

A 'revisited' guide to GNU Screen

By Anderson Silva and Steve 'Ashcrow' Milner

Remember how tabbed browsing revolutionized the web experience? GNU Screen can do the same for your experience in the command line. Screen allows you to manage several interactive shell instances within the same "window". By using different keyboard shortcuts, you can shuffle through the shell instances, access any of them directly, create new ones, kill old ones, and attach and detach existing ones.

Instead of opening up several terminal instances on your desktop or using those ugly GNOME/KDE-based tabs, Screen can do it better and simpler.

Not only that, with Screen you can share sessions with others and detach/attach terminal sessions. It is a great tool for people who have to share working environments between work and home.

By adding a status bar to your screen environment, you are able to name your shell instances on the fly or via a configuration file called .screenrc that can be created in the user's home directory.

Installation

Installing screen on a Fedora system is quite easy with yum, assuming you have root access.

1. Login as root:

   su - # enter root password

2. Use yum to install it:

   yum install screen

On Debian based distributions like Ubuntu:

1. As root:

   apt-get install screen

Enter your password. After a few minutes (depending on your network connection), Screen will be installed. But before you start playing around with it, let's look at how to do some basic configuration.

Customizing the configuration file

Screen keeps its configuration file in the same vein that many applications do: in a dot file in your user's home directory. This file is aptly named .screenrc. In my experience, most people use ~/.screenrc to do two things:

• Make a hardstatus line. This is basically a line at the bottom of the screen that lists your current terminal and all opened ones. It can also display the system clock and the hostname.
• Default screens on startup. It's quite nice to have your IRC connection, mail client, and default SSH connections auto-start for you!

The lines below are numbered for reference. Your config file should not have numbered lines.

1 hardstatus alwayslastline
2 hardstatus string '%{= kG}[ %{G}%H %{g}][%= %{=kw}%?%-Lw%?%{r}(%{W}%n*%f%t%?(%u)%?%{r})%{w}%?%+Lw%?%?%= %{g}][%{B}%Y-%m-%d %{W}%c %{g}]'
3
4 # Default screens
5 screen -t shell1 0
6 screen -t shell2 1
7 screen -t server 2 ssh me@myserver

On lines 1 and 2, you are setting the hardstatus. Line 1 makes the hardstatus always show up as the last line. Line 2 is what will be shown in the hardstatus line. In this case you will see something like this at the bottom:

As you change screens, you will see the parentheses move around the active screen.

Line 4 is a comment, as it starts with #. Lines 5-7 are all screen statements in the following format:

screen -t NameOfScreen ScreenNumber ShellCommand

Shortcuts

The following are some of the most commonly used shortcuts that let you navigate through your screen environment. Note that unless modified by your .screenrc, by default every screen shortcut is preceded by Ctrl+a. Note also that these shortcuts are case-sensitive.

• 0 through 9 – Switches between windows
• Ctrl+n – Switches to the next available window
• Backspace – Switches to the previous available window
• Ctrl+a – Switches back to the last window you were on
• A – Changes window session name
• K – Kills a window session
• c – Creates a new window
• [ - Then use arrows to scroll up and down terminal
• " - Displays a lists of all opened tabs allowing navigation with your arrow keys.

You can learn more about shortcuts in Screen's man pages. In your terminal, run:

man screen

Sharing a session with others

Another great application of Screen is to allow other people to login to your station and to watch the work you are doing. It is a great way to teach someone how to do things in the shell.

Note: Screen has to be SUID if you want to share a terminal between two users. SUID allows an executable to be run with the privileges of the owner of that file, instead of with the user's own privileges. There are some security concerns when doing this, so use this tip at your own discretion.

Setup to allow screen to be shared

First, as root:

chmod u+s /usr/bin/screen
chmod 755 /var/run/screen

Log out of root, and run Screen as the user who is going to share the session:

screen

Under your new screen session:

1. Press Ctrl+a, then type ':multiuser on' and press Enter.

2. Press Ctrl+a, then type ':acladd $username'

Where $username is the username of the person who will connect to your screen session.

Connecting to the shared screen

Now that a screen session is being shared by following the previous steps, let's attach ourselves to the session and watch it by connecting to the machine via ssh and entering the following command:

screen -x $username/

Where $username is the username of the person who is sharing the screen session, and yes, you do need slash (/) at the end of the command.

And now both the users (from the host and guest) will be sharing a screen session and can run commands on the terminal.

Working from multiple locations

Let's say you have a screen session open at work with X number of windows on it. Within those screens you may be running an IRC client, an SSH connection to the web server, and your favorite text-based email client. It's 5 p.m. and you have to go home, but you still have work left to do.

Without Screen you would probably go home, VPN into your company's network, and fire up all the shells you need to keep working from home. With Screen, life gets a little easier.

You can simply SSH into your workstation at work and list your available screen sessions with the command:

screen -ls

And connect to the sessions you were running at work with the command:

screen -x screen_session_name

This way screen will let you pick things up exactly from where you left off.

Once you get used to the shortcuts in GNU Screen, not only will your desktop become more organized (due to the lower number of open windows), but your efficiency as a developer or system administrator will increase not only at work but at your home office as well.

Talkback: Discuss this article with The Answer Gang

HelpDex

By Shane Collinge

These images are scaled down to minimize horizontal scrolling.

All HelpDex cartoons are at Shane's web site, www.shanecollinge.com.

Talkback: Discuss this article with The Answer Gang

Part computer programmer, part cartoonist, part Mars Bar. At night, he runs around in his brightly-coloured underwear fighting criminals. During the day... well, he just runs around in his brightly-coloured underwear. He eats when he's hungry and sleeps when he's sleepy.

Ecol

By Javier Malonda

The Ecol comic strip is written for escomposlinux.org (ECOL), the web site that supports es.comp.os.linux, the Spanish USENET newsgroup for Linux. The strips are drawn in Spanish and then translated to English by the author.

These images are scaled down to minimize horizontal scrolling.

Click here to see the full-sized image

These cartoons are copyright Javier Malonda. They may be copied, linked or distributed by any means. However, you may not distribute modifications. If you link to a cartoon, please notify Javier, who would appreciate hearing from you.

Talkback: Discuss this article with The Answer Gang

XKCD

By Randall Munroe

Click here to see the full-sized image

Talkback: Discuss this article with The Answer Gang

Doomed to Obscurity

By Pete Trbovich

These images are scaled down to minimize horizontal scrolling.

All "Doomed to Obscurity" cartoons are at Pete Trbovich's site, http://penguinpetes.com/Doomed_to_Obscurity/.

Talkback: Discuss this article with The Answer Gang

The Linux Launderette

From the Department of Pedantry

René Pfeiffer [lynx at luchs.at]

On Sep 25, 2009 at 1317 -0700, Rick Moen appeared and said:

> I offer the following anecdote in belated celebration of National
> Punctuation Day (http://www.nationalpunctuationday.com/) -- which was
> September 24th, 2009, but we can hope that it'll prove to be... um...
> periodic.

You might be interested in this little book:

A panda bear walks into a bar and orders a sandwich. The waiter brings him the sandwich. The panda bear eats it, pulls out a pistol, kills the waiter, and gets up and starts to walk out.

The bartender yells for him to stop. The panda bear asks, "What do you want?" The bartender replies, "First you come in here, order food, kill my waiter, then try to go without paying for your food."

The panda bear turns around and says, "Hey! I=E2=80=99m a Panda. Look it up= !" The bartender goes into the back room and looks up panda bear in the encyclopedia, which read: "Panda: a bear-like marsupial originating in Asian regions. Known largely for it=E2=80=99s stark black and white colorin= g. Eats, shoots and leaves."

http://en.wikipedia.org/wiki/Eats,_Shoots_&_Leaves

, René.

[ Thread continues here (2 messages/2.18kB) ]

[OT] Meetup?

Jimmy O'Regan [joregan at gmail.com]

2009/9/15 Rick Moen <[email protected]>:

> Quoting Jimmy O'Regan ([email protected]):
>
>> I'll be in San Francisco for a couple of days next month (Google
>> Mentor Summit). Will anyone else be around?
>
> Only all the time.
>
> Chez Moen is down about 60 km south of the city of San Francisco,
> at the base of the San Francisco Peninsula, close to Stanford
> University.  I don't very often get to San Francisco, but certainly
> can/will in order to rendezvous and spend time.
>

With less than a week left, I just now took the time to look at a map and see where I'll /really/ be; turns out it's Sunnyvale, which is much closer to Stanford

-- 
<Leftmost> jimregan, that's because deep inside you, you are evil.
<Leftmost> Also not-so-deep inside you.

[ Thread continues here (6 messages/5.67kB) ]

Power management by sonar

Jimmy O'Regan [joregan at gmail.com]

From this Slashdot story: http://hardware.slashdot.org/story/09/10[...]re-Detects-Laptop-User-Presence?from=rss

Software here: http://stevetarzia.com/sonar/

I'm not really interested in the power management side of things so much as the scope it provides for a Halloween prank... Trick or treat!

-- 
<Leftmost> jimregan, that's because deep inside you, you are evil.
<Leftmost> Also not-so-deep inside you.

Meetup

Jimmy O'Regan [joregan at gmail.com]

Hi. I'm in Sunnyvale now. I think I should be around after 6, if anyone can suggest a good meeting place.

[ Thread continues here (4 messages/7.46kB) ]

Talkback: Discuss this article with The Answer Gang