Where the happening community people now hang

Eric of Proven Scaling commented on a lack of IRC action in the normal mysql channels today when he visited the #drizzle channel on irc.freenode.net.

ebergen: I'm still in #mysql-dev and #planet.mysql but they are hardly active these days [1:51pm]
rbradfor: ebergen: funny, #drizzle is where the action is. [1:51pm]

There is active movement on the Drizzle project. Why is this? Well, I think most importantly is that there is active contribution from the community, at least 5 different companies and more individuals are pushing code to Drizzle, and it’s being accepted and incorporated. Something you can not say about the MySQL Community branch.

As I write this, there are 35 active people on the #drizzle channel now, and 137 members of the Drizzle Discuss list.

My contribution is as Monty put’s it, “Your the build team”. I am managing the Build Master for Drizzle and my company 42SQL is providing the hosting and support. I’ve even managed to push my first small code changes to the project using the very simple Contributing Code instructions. No fuss, no pain, and I don’t care if it doesn’t get included, but it’s available for all to see and use.

In 2 days we now have 15 build slaves covering Ubuntu 8.04 32 & 64bit, Debian 32 & 64 bit, CentOS 5 64 bit, Gentoo 32 & 64 bit, and Mac OS/X 10.5, with definitely some color at times in the waterfall display.

Jay has a good article on Drizzle Buildbot Now Accepting BuildSlaves.

We need your help! There are plenty of Linux/Unix OS’s out there, and we want to know Drizzle can be compiled as broadly as possible. Most of the contributors of build slaves to date are not names I know well in the MySQL community, which is excellent. What I’d like to see is more names I do know.

It’s easy, just check out Instructions for setting up a BuildSlave for Drizzle.

Drizzle needs you


Use MySQL, but want to follow the new kid on the block?
Want to help contribute to Drizzle?

We are seeking help in compiling across different platforms.
Please help us by becoming a buildbot slave.

There are detailed instructions, so now is the time to take a few minutes and help out the project.

The Drizzle Buildbot is hosted and supported by 42SQL.

Building sources with BuildBot

Unless your in the desert under a rock (where rain is clearly needed), you will have heard of Drizzle – A Lightweight SQL Database for Cloud and Web. My company 42SQL is sponsoring the BuildBot for the Drizzle project. BuildBot is a system to automate the compile/test cycle required by most software projects to validate code changes.

Check out Installing Buildbot for what’s necessary to get a working installation. This is necessary for the Master and Slaves.

Configuration was a little more complicated then expected, due to lack of accurate documentation, and reading old docs at sourceforge. Be sure now to read here.

This is a step by step approach I used to successfully configure Drizzle Build Bot (Master and Slave).

1. Create OS User.

su -
useradd buildbot
su - buildbot

2. Create Master Installation

buildbot create-master /home/buildbot/master
cd /home/buildbot/master
cp master.cfg.sample master.cfg
vi master.cfg

Here is a diff of what simple changes I made to the master.cfg.sample work in my environment.

$ diff master.cfg.sample master.cfg
23c23
< c['slaves'] = [BuildSlave("bot1name", "bot1passwd")]
---
> c['slaves'] = [BuildSlave("centos5_64", "paSSw0rd")]
95c95
< cvsroot = ":pserver:anonymous@cvs.sourceforge.net:/cvsroot/buildbot"
---
> cvsroot = ":pserver:anonymous@buildbot.cvs.sourceforge.net:/cvsroot/buildbot"
105c105
< f1.addStep(Trial(testpath="."))
---
> #f1.addStep(Trial(testpath="."))
108c108
< 'slavename': "bot1name",
---
>       'slavename': "centos5_64",
148c148
< #c['debugPassword'] = "debugpassword"
---
> c['debugPassword'] = "paSSw0rd"
156c156
< #                                       "admin", "password")
---
> #                                       "admin", "paSSw0rd")
166,167c166,167
< c['projectName'] = "Buildbot"
< c['projectURL'] = "http://buildbot.sourceforge.net/"
---
> c['projectName'] = "Drizzle Buildbot"
> c['projectURL'] = "https://launchpad.net/drizzle/"

Initially I’m just going to test with a CVS checkout of buildbot to confirm operations.
NOTE: The example provided pserver URL in master.cfg.sample is invalid.

3. Start Master

buildbot start /home/buildbot/master > start.log
tail -f /home/buildbot/master/twistd.log

4. Confirm Master

lynx http://drizzlebuild.42sql.com:8010

5. Create Slave

buildbot create-slave /home/buildbot/slave drizzlebuild.42sql.com:9989 centos5_64 PaSSw0rd

I got stuck here based on docs, be sure the port number is the client port.

6. Configure Slave

cd /home/buildbot/slave/info
echo "Ronald Bradford < ronald .bradford @ google mail >" > admin
echo "Drizzle CentOS 5 64bit "`uname -a` > host
cat admin host

7. Start Slave

cd /home/buildbot/slave
buildbot start /home/buildbot/slave > start.log
tail -f /home/buildbot/slave/twistd.log

8. Confirm Slave

If everything is working by the time you look at the twistd.log you will see work happening.
You can also via the web interface and see in Lasted Builds the first build is working.

9. Stop Services

To stop the BuildBot master and slave.

buildbot stop /home/buildbot/master
buildbot stop /home/buildbot/slave

10. Change Master Configuration

Should you make any changes to master.cfg the following command will re-read the configuration file.

buildbot sighup /home/buildbot/master

11. Startup
The following was added to cron

@reboot buildbot start /home/buildbot/master
@reboot buildbot start /home/buildbot/slave

The @reboot is new sytax for me, so I can’t yet confirm it’s operation.

If you want to be a build slave for Drizzle, check out Instructions here

Installing Buildbot

BuildBot is a system to automate the compile/test cycle required by most software projects to validate code changes.

Here is my environment.

$ uname -a
Linux app.example.com 2.6.18-53.el5 #1 SMP Mon Nov 12 02:14:55 EST 2007 x86_64 x86_64 x86_64 GNU/Linux
$ python
Python 2.4.3 (#1, May 24 2008, 13:57:05)

Here is what I did to get it installed successfully.

CentOS

$ yum install python-devel
$ yum install zope

Ubuntu

$ apt-get install python-dev
$ apt-get install python-zopeinterface
$ cd /tmp
# installation of Twisted
$ wget http://tmrc.mit.edu/mirror/twisted/Twisted/8.1/Twisted-8.1.0.tar.bz2
$ bunzip2 Twisted-8.1.0.tar.bz2
$ tar xvf Twisted-8.1.0.tar
$ cd Twisted-8.1.0
$ sudo python setup.py install
# installation of BuildBot
$ cd /tmp
$ wget http://downloads.sourceforge.net/buildbot/buildbot-0.7.8.tar.gz
$ tar xvfz buildbot-0.7.8.tar.gz
$ cd buildbot-0.7.8
$ sudo python setup.py install


And a confirmation.
$ buildbot --version
Buildbot version: 0.7.8
Twisted version: 8.1.0

You will notice a few dependencies. I found these out from the following errors.

Error causing needing python-devel

$ python setup.py install
....
gcc -pthread -fno-strict-aliasing -DNDEBUG -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m64 -mtun
e=generic -D_GNU_SOURCE -fPIC -fPIC -I/usr/include/python2.4 -c conftest.c -o conftest.o
building 'twisted.runner.portmap' extension
creating build/temp.linux-x86_64-2.4
creating build/temp.linux-x86_64-2.4/twisted
creating build/temp.linux-x86_64-2.4/twisted/runner
gcc -pthread -fno-strict-aliasing -DNDEBUG -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m64 -mtun
e=generic -D_GNU_SOURCE -fPIC -fPIC -I/usr/include/python2.4 -c twisted/runner/portmap.c -o build/temp.linux-x86_64-2.4/twisted/runner/portmap.o
twisted/runner/portmap.c:10:20: error: Python.h: No such file or directory
twisted/runner/portmap.c:14: error: expected ‘=’, ‘,’, ‘;’, ‘asm’ or ‘__attribute__’ before ‘*’ token
twisted/runner/portmap.c:31: error: expected ‘=’, ‘,’, ‘;’, ‘asm’ or ‘__attribute__’ before ‘*’ token
twisted/runner/portmap.c:45: error: expected ‘=’, ‘,’, ‘;’, ‘asm’ or ‘__attribute__’ before ‘PortmapMethods’

Error causing zope to be installed

$ buildbot start /home/buildbot/master/
Traceback (most recent call last):
  File "/usr/bin/buildbot", line 4, in ?
    runner.run()
  File "/usr/lib/python2.4/site-packages/buildbot/scripts/runner.py", line 939, in run
    start(so)
  File "/usr/lib/python2.4/site-packages/buildbot/scripts/startup.py", line 85, in start
    rc = Follower().follow()
  File "/usr/lib/python2.4/site-packages/buildbot/scripts/startup.py", line 6, in follow
    from twisted.internet import reactor
  File "/usr/lib64/python2.4/site-packages/twisted/internet/reactor.py", line 11, in ?
    from twisted.internet import selectreactor
  File "/usr/lib64/python2.4/site-packages/twisted/internet/selectreactor.py", line 17, in ?
    from zope.interface import implements
ImportError: No module named zope.interface

Installation was the easy part, configuration a little more complex.

MySQL Proxy lua scripts from presentation

The following Lua scripts are the examples are from my MySQL Proxy @ OSCON 08 presentation.

analyze_query.lua

MySQL Proxy Analyze Query.

Requires MySQL Proxy Logging Module.

What is released is the Version for MySQL 5.0. A generic version for all MySQL versions is not yet released.

histogram.lua

This script is part of the standard MySQL Proxy examples.

Other Scripts

Additional Lua scripts from MySQL forge are available here.

MySQL Proxy @ OSCON 08

Today I presented with Giuseppe Maxia of Sun Microsystems Inc at OSCON 08 on “MySQL Proxy: From Architecture to Implementation”. I was surprised to find that MySQL has a strong showing with a number of presentations this week.

Our talk covered the basics of MySQL Proxy, what’s coming in future features, and a number of examples of how I have used Proxy in consulting engagements to improve the information retrieval particularly for identifying performance problems.

Download Presentations Slides

The lighter side at O'Reilly OSCON 08

Between the keynotes, general sessions, BoFs that are plenty of events at OSCON 08.

Last night in the BoF time, a selected few enjoyed the relaxed music mode of the Good Company Soul and Blues Review. You can find my photos of what I’ve deemed the O’Reilly Band here.

Lacking the offers of free beer and activities of the larger events, may missed out on knowing about a great time, it was most enjoyable and relaxing. I think they should warm up for the keynotes in future, but then who would run all the AV!

The lights were rather bright, so grayscale actually helped out, so I’ve got a mixture of color and black and white.

larger versions.

The fast paced open source ecosystem

This morning at OSCON 08, Tim O’Reilly’s opening keynote Open Source on the O’Reilly Radar included a slide on Drizzle, giving this new project maximum exposure to the Open Source community.

Drizzle was only officially announced yesterday in Drizzle, Clouds, “What If?” by primary architect Brian Aker. Things move fast. There has been a number of comments from people yesterday including Mark Attwood, Monty Widenus,Monty Taylor,Ronald Bradford, Arjen Lentz, Lewis Cunningham, Jeremy Cole, 451 Group,Matt Asay, Assaf Arkin, SlashDot, Builder.au and MySQL HA.

The Drizzle Launchpad project has reached 5th on a Google Search.

Unfortunately, not all uptake and feedback was positive. The official Wikipedia page for Drizzle was marked for speedy deletion almost instantly, and within a few hours permanently deleted.

The new kid on the block – Drizzle

Before today, Drizzle was known as a light form of rain found in Seattle (among other places). Not any more. If you have not read the news already today, Drizzle, Clouds, “What If?” is the new kid on the RDBMS bock.

Faster, leaner and designed with the original goals of ease-of-use, reliability and performance, Drizzle will make an impact in those organizations that are seeking a viable database storage solution for large scalable applications. The key to Drizzle is several fold. First, the crud has been removed. The first part of Drizzle development is to remove bloat or non functioning software from the MySQL tree. In fact if you monitor the commits, it reads like, this has been removed, these files have been deleted, this code has been refactored, this new library has been introduced. Design decisions that have limited MySQL’s development for years are being simply cast aside.

The current landscape has become more complicated in 2008. You have the official MySQL releases, 5.0 is becoming ancient (being released almost 3 years ago), 5.1 is now clearly a lame duck, no release date for past few years, (the internal joke was 5.1 will be released in Q2, but the year is unspecified). 6.0 is in identity crisis with beta parts in alpha. These versions are moving so slowly, they are moving towards extinction like the dinosaurs. Monty Widenius is working solidly on Maria (and unofficial MySQL 5.1 branch), probably more stable and possibly released before 5.1. MySQL cluster has gone it’s own way, it was shackled by the 5.1 legacy and simply couldn’t wait for a GA product.

Jim Starkey (creator of Falcon, the new 6.0 storage engine) now is working in the clouds with Nimbus DB. Dorsal Source is lying dormant, Proven Scaling has it’s enterprise binaries and now Percona has it’s own patched ports. You have strong patches from Google and eBay that have zero hope of every being introduced into the official MySQL releases, probably until 7.x (5.1 and 6.0 have been frozen for a long time). Innodb from Oracle invested heavily in new features in a 5.1 plugin, announced at the MySQL Conference, broken in 1 day by MySQL releasing a new RC version making it in-compatible. Kickfire and Infobright have their own hacked versions, and Nitro DB I suspect have just given up waiting (now like 2 years).

With Sun’s acquisition now at T+6 months, cash and resources doesn’t appear to have helped with the official product. The single greatest movement in this period is that MySQL is now hosted under Launchpad, enabling anybody to access the source code, and even create branches like Jim Winstead reported. However I doubt you will see this helping code getting in the mainline product, but at least it will be more visible. This was an initiative long before the Sun acquisition, and indeed is against Sun policy of using Mercurial.

So why is Drizzle going to be any different or better?

You start with a committed list of contributors from already 6-7 different organizations. The clear goals of simplification, to make it faster and scale better on multi-core servers echo the work being done. You have developers who work in real world situations, not just coders for many years without experiencing operational use, and you have zero sales and marketing getting in the way. Removal of incomplete or stagnant functionality is key for the alpha version and includes stored procedures, triggers, prepared statements, query cache, extra data types, full-text, timezones etc is just the start.

Being small and nimble will enable Drizzle to develop and release code in much shorter iterations. You will see new developments allowing far greater plugin support via the new modularity approach and far better coding standards, making expert knowledge of how MySQL internals work a lesser requirement to contribute.

Will it fizzle, will it dazzle? Drizzle has the potential to be a stellar product. I’m a supporter and I hope to contribute in some small way.

References


About the Author

Ronald Bradford provides Consulting and Advisory Services in Data Architecture, Performance and Scalability for MySQL Solutions. An IT industry professional for two decades with extensive database experience in MySQL, Oracle and Ingres his expertise covers data architecture, software development, migration, performance analysis and production system implementations. His knowledge from 10 years of consulting across many industry sectors, technologies and countries has provided unique insight into being able to provide solutions to problems. For more information Contact Ronald.

An East Coast option

Within the present MySQL ecosystem, there are limited options for dedicated MySQL Consulting in the US. Outside of the official Sun/MySQL Consulting, Percona and Proven Scaling both based in Silicon valley are the only options generally known and accepted by the MySQL Community.

There is now an east coast option based in New York, and that is Ronald Bradford. Providing expert MySQL Consulting in Architecture, Performance, Scalability, Migration and Knowledge Transfer.

With two decades working in the IT industry, Ronald is well qualified in MySQL having previously provided consulting services for MySQL Inc combining 9 years experience with the product. His consulting experience is not limited to MySQL, having also worked extensively with Oracle, and previously with Ingres. More details of this experience is available at Linked In

This week you will find him on the west coast. If your at OSCON 2008, then please track me down. You can use my Contact Form, email [me] at [this domain], ping me on Twitter, track me on irc://irc.freenode.net (~arabxptyltd) or drop in to my OSCON session at 2:35pm Thursday.

Your data and the cloud

I will be speaking on July 29th in New York at an Entrepreneurs Forum on A Free Panel on Cloud Computing. With a number of experts including Hank Williams of KloudShare, Mike Nolet of AppNexus, and Hans Zaunere of New York PHP fame is should be a great event.

The focus of my presentation will be on “Extending existing applications to leverage the cloud” where I will be discussing both the advantages of the cloud, and the complexities and issues that you will encounter such as data management, data consistency, loss of control, security and latency for example.

Using traditional MySQL based applications I’ll be providing an approach that can lead to your application gaining greater power of cloud computing.


About the Author

Ronald Bradford provides Consulting and Advisory Services in Data Architecture, Performance and Scalability for MySQL Solutions. An IT industry professional for two decades with extensive database experience in MySQL, Oracle and Ingres his expertise covers data architecture, software development, migration, performance analysis and production system implementations. His knowledge from 10 years of consulting across many industry sectors, technologies and countries has provided unique insight into being able to provide solutions to problems. For more information Contact Ronald.

When (n) counts?

I have seen on many engagements the column data type is defined as INT(1).

People have the misconception that this numeric integer data type is of the length of one digit, or one byte. (One digit is 0-9 an one byte is 0-255)

This is incorrect.

Integer

For integer numeric data types in MySQL, that is TINYINT, SMALLINT, MEDIUMINT, INT, BIGINT the (n) has no bearing on the size of data stored within the specific data type. The (n) is simply for display formatting.

In the MySQL Manual 10.2. Numeric Types you read This optional display width is used to display integer values having a width less than the width specified for the column by left-padding them with spaces. The display width does not constrain the range of values that can be stored in the column, nor the number of digits that are displayed for values having a width exceeding that specified for the column.

The following example shows the (n) in this case 3 has no effect on the size of data stored.

DROP TABLE IF EXISTS numeric_int;
CREATE TABLE numeric_int(i INT(3) NOT NULL);
INSERT INTO numeric_int VALUES (1),(22),(333),(444),(55555);
SELECT * FROM numeric_intG
i: 1
i: 22
i: 333
i: 444
i: 55555

Floating Point

When it comes to floating point precision of FLOAT and DOUBLE, the syntax of (m,n) has a different inteperation. The manual states A precision from 0 to 23 results in a four-byte single-precision FLOAT column. A precision from 24 to 53 results in an eight-byte double-precision DOUBLE column.
I will discuss this some more in a different post with some interesting findings.

And MySQL allows a non-standard syntax: FLOAT(M,D) or REAL(M,D) or DOUBLE PRECISION(M,D). Here, “(M,D)” means than values can be stored with up to M digits in total, of which D digits may be after the decimal point. For example, a column defined as FLOAT(7,4) will look like -999.9999 when displayed. MySQL performs rounding when storing values, so if you insert 999.00009 into a FLOAT(7,4) column, the approximate result is 999.0001.

So in the case of FLOAT,DOUBLE the (n) does both affect storage and presentation where it rounds the number as confirmed by the following test. Look a the last 2 rows for the rounding confirmation.

DROP TABLE IF EXISTS numeric_float;
CREATE TABLE numeric_float(f1 FLOAT(10,5)  NOT NULL);
INSERT INTO numeric_float values (1),(2.0),(3.12345),(4.123451),(5.123456);
Query OK, 5 rows affected (0.00 sec)
Records: 5  Duplicates: 0  Warnings: 0
SELECT * FROM numeric_floatG
f1: 1.00000
f1: 2.00000
f1: 3.12345
f1: 4.12345
f1: 5.12346
5 rows in set (0.01 sec)

Fixed Precision

The DECIMAL data type (NUMBER is a synonym) stores numbers to a fixed number of precision. From the manual again When declaring a DECIMAL or NUMERIC column, the precision and scale can be (and usually is) specified; for example: salary DECIMAL(5,2)
In this example, 5 is the precision and 2 is the scale. The precision represents the number of significant digits that are stored for values, and the scale represents the number of digits that can be stored following the decimal point. If the scale is 0, DECIMAL and NUMERIC values contain no decimal point or fractional part.

So in our test:

DROP TABLE IF EXISTS numeric_decimal;
CREATE TABLE numeric_decimal(f1 DECIMAL(10,5)  NOT NULL);
INSERT INTO numeric_decimal values (1),(2.0),(3.12345),(4.123451),(5.123456);
Query OK, 5 rows affected, 2 warnings (0.00 sec)
SELECT * FROM numeric_decimalG
f1: 1.00000
f1: 2.00000
f1: 3.12345
f1: 4.12345
f1: 5.12346

What is also interesting is that with a FLOAT, the rounding of a number greater then (n), produces no warnings, yet when using DECIMAL you will see warnings. These are:

INSERT INTO numeric_decimal values (1),(2.0),(3.12345),(4.123451),(5.123456);
Query OK, 5 rows affected, 2 warnings (0.00 sec)
Records: 5  Duplicates: 0  Warnings: 2

mysql> show warnings;
+-------+------+-----------------------------------------+
| Level | Code | Message                                 |
+-------+------+-----------------------------------------+
| Note  | 1265 | Data truncated for column 'f1' at row 4 |
| Note  | 1265 | Data truncated for column 'f1' at row 5 |
+-------+------+-----------------------------------------+
2 rows in set (0.00 sec)

What is also interesting is that the manual states the following When such a column is assigned a value with more digits following the decimal point than are allowed by the specified scale, the value is converted to that scale. (The precise behavior is operating system-specific, but generally the effect is truncation to the allowable number of digits.)

The number is generally truncated, buy differs per OS. In the case on Mac O/S and Linux it is rounded. The two test environments in this case where:

mysql> show variables like '%version%';
+-------------------------+------------------------------+
| Variable_name           | Value                        |
+-------------------------+------------------------------+
| protocol_version        | 10                           |
| version                 | 5.1.23-rc                    |
| version_comment         | MySQL Community Server (GPL) |
| version_compile_machine | i686                         |
| version_compile_os      | apple-darwin9.0.0b5          |
+-------------------------+------------------------------+
5 rows in set (0.01 sec)

mysql> show variables like '%version%';
+-------------------------+------------------------------+
| Variable_name           | Value                        |
+-------------------------+------------------------------+
| protocol_version        | 10                           |
| version                 | 5.1.24-rc                    |
| version_comment         | MySQL Community Server (GPL) |
| version_compile_machine | i686                         |
| version_compile_os      | redhat-linux-gnu             |
+-------------------------+------------------------------+
5 rows in set (0.41 sec)

Conclusion

So just to conclude, (n) for Integer types is for display formatting only, (m,n) for floating point will round the number at n places, while in fixed point (m,n) n will round or truncate the number.


About the Author

Ronald Bradford provides Consulting and Advisory Services in Data Architecture, Performance and Scalability for MySQL Solutions. An IT industry professional for two decades with extensive database experience in MySQL, Oracle and Ingres his expertise covers data architecture, software development, migration, performance analysis and production system implementations. His knowledge from 10 years of consulting across many industry sectors, technologies and countries has provided unique insight into being able to provide solutions to problems. For more information Contact Ronald.

References

The minimum testing for a shared disk MySQL environment

Recently I was asked to provide guidelines for testing fail over of a MySQL configuration that was provided by a hosting provider.

The first observation was the client didn’t have any technical details from the hosting provider of what the moving parts were, and also didn’t have any confirmation other then I think a verbal confirmation that it had been testing.

The first rule in using hosting, never assume. Too many times I’ve seen details from a client stating for example H/W configuration, only to audit and find out otherwise. RAID is a big one, and is generally far more complex to determine. Even for companies with internal systems I’ve seen the most simple question go unanswered. Q: How do you know your RAID is fully operational? A: Somebody will tell us? It’s really amazing to investigate on site with the client to find that RAID system is running in a degraded mode due to a disk failure and nobody knew.

It took some more digging to realize the configuration in question was with Red Hat Cluster Suite. A word of warning for any clients that use this, DO NOT USE MyISAM. I’ll leave it to the readers to ask me why.

Here is a short list I provided as the minimum requirements I’d test just to ensure the configuration was operational.

Verifying a working Red Hat Cluster Suite MySQL Environment

The MySQL Environment

The database environment consists of two MySQL database servers, configured in an active/passive mode using a shared disk storage via SAN.
For the purposes of the following procedures the active server will be known as the ‘primary’ server, and the passive server will be the ‘secondary server’.
The two physical servers for the purposes of these tests will be defined as ‘alpha’ and ‘beta’, with specific H/W that does not change during these tests.

Normal Operations

Expected Configuration under normal operations.

Primary Server

  • server is pingable
  • server accepts SSH Connection
  • MySQL service is started
  • has /data appropriately mounted
  • has assigned VIP address
  • MySQL configuration file and settings are correct

Secondary Server

  • server is pingable
  • server accepts SSH Connection
  • MySQL service IS NOT started
  • DOES NOT have /data mounted
  • DOES NOT has assigned VIP address
  • MySQL configuration file is not available

1. Reboot servers ‘alpha’ and ‘beta’.

Test Status:

  • alpha server is the designated primary server
  • alpha and beta servers are operational

Action:
1.1 Restart alpha server (init 6)
1.2 Restart beta server (init 6)

Checklist:
1.3 Alpha server matches primary server configuration
1.4 Beta server matches secondary server configuration

2. Controlled fail over from ‘alpha’ to ‘beta’

Test Status:

  • alpha server is the designated primary server
  • alpha and beta servers are operational

Action:
2.1 Alpha server – Instigate Cluster failover (clusvcadm -r mysql-svc)

Checklist:
2.2 Beta server matches primary server configuration
2.3 Alpha server matches secondary server configuration

3. Controlled failover from ‘beta’ to ‘alpha’

Test Status:

  • beta server is the designated primary server
  • alpha and beta servers are operational

Action:
3.1 beta server – Instigate Cluster failover (clusvcadm -r mysql-svc)

Checklist:
3.2 Alpha server matches primary server configuration
3.3 Beta server matches secondary server configuration

Exception Operations

4. Loss of connectivity to primary server

Test Status:

  • alpha server is the designated primary server
  • beta server is online

Action:
4.1 Stop networking services on ‘alpha’ (ifdown bond0)

Checklist:
4.2 Monitoring detects and reports connectively loss
4.3 Automated failover occurs
4.4 Beta server matches primary server configuration
4.5 Alpha server matches secondary server configuration

5. Restore connectivity to secondary server

Test Status:

  • beta server is the designated primary server
  • alpha server is online, but not accessible via private IP

Action:
4.1 Start networking services on ‘alpha’ (ifup bond0)

Checklist:
5.2 Monitoring detects and reports connectively restored
5.3 No failback occurs
5.4 Beta server matches primary server configuration
5.5 Alpha server matches secondary server configuration

6. Loss of connectivity to secondary server

Test Status:

  • beta server is the designated primary server
  • alpha server is online

Action:
6.1 Stop networking services on ‘alpha’ (ifdown bond0)

Checklist:
6.2 Monitoring detects and reports connectively lost
6.3 No failback occurs
6.4 Beta server matches primary server configuration
6.5 Alpha server matches secondary server configuration

7. Restore connectivity to secondary server

Test Status:

  • beta server is the designated primary server
  • alpha server is online, but not accessible via private IP

Action:
7.1 Start networking services on ‘alpha’ (ifup bond0)

Checklist:
7.2 Monitoring detects and reports connectively restored
7.3 No failback occurs
7.4 Beta server matches primary server configuration
7.5 Alpha server matches secondary server configuration

8. Power down secondary server

Test Status:

  • beta server is the designated primary server
  • alpha server is online

Action:
8.1 Power down alpha (init 0) NOTE: Need remote boot capabilities

Checklist:
8.2 Monitoring detects and reports connectively lost
8.3 Beta server matches primary server configuration
8.4 Additional paging for extended down time for ‘degraded support for failover’

9. Loss of connectivity to primary server

Test Status:

  • beta server is the designated primary server
  • alpha server is offline

Action:
9.1 Power down beta (init 0) NOTE: Need remote boot capabilities

Checklist:
9.2 Monitoring detects and reports connectively lost
9.3 Site database connectively completely unavailable
9.4 Additional paging for loss of HA solution

10. power restored to secondary server
Test Status:

  • alpha server is offline
  • beta server is offline

Action:
10.1 Power on alpha

Checklist:
10.2 Monitoring detects and reports server up
10.3 Alpha server assumes primary role (previously it was beta)
10.4 Alpha server matching primary server configuration
10.5 Addition paging for degraded HA

11. power restored to secondary server

Test Status:

  • alpha server is primary server
  • beta server is offline

Action:
11.1 Power on beta

Checklist:
11.2 Monitoring detects and reports server up
11.3 Alpha server matching primary server configuration
11.4 Beta server matching secondary server configuration

Database Operations

12. MySQL services on primary server go offline

Test Status:

  • alpha server is the designated primary server
  • beta server is online

Action:
12.1 Stop mysql services on ‘alpha’ (/etc/init.d/mysqld stop)

Checklist:
12.2 Monitoring detects and reports database loss (while connectivity is still available)
12.3 Automated failover occurs
12.4 Beta server matches primary server configuration
12.5 Alpha server matches secondary server configuration

13. MySQL services on secondary server go offline

Test Status:

  • beta server is the designated primary server
  • alpha server is online

Action:
13.1 stop mysql services on ‘beta’ (/etc/init.d/mysqld stop)

Checklist:
13.2 Monitoring detects and reports database loss (while connectivity is still available)
13.3 Automated failover occurs
13.4 Alpha server matches primary server configuration
13.5 beta server matches secondary server configuration

14. Load Testing during failure

Test Status:

  • alpha server is the designated primary server
  • beta server is online

Action:
14.1 Agressive load testing against database server
14.2 MySQL killed without prejudice (killall -9 mysqld_safe mysql)

Checklist:
14.3 Monitoring detects and reports mysql service loss
14.4 Automated failover occurs
14.5 Beta server matches primary server configuration
14.6 Alpha server matches secondary server configuration
14.7 Beta mysql logs shows a forced MySQL Recovery in logs

15. Forced Recovery

Test Status:

  • alpha server is the designated primary server
  • beta server is online

Action:
15.1 Manual full database backup is done (in case recovery does not work). Hosting Provider not told of this.
15.2 Dummy new table/schema is created (used as verification point)
15.3 Database on alpha primary server is dropped
15.4 Hosting Provider is notified stating a full database recovery including Point In time to just before drop (no time given, only command that was run)

Checklist:
15.5 Site is marked as unavailable
15.6 Hosting Provider restore data from backup and recover to point in time
15.7 Confirmation that new table/schema is restored, and full schema is available
15.8 Site is made available
15.9 Record of time for full disaster is recorded

Conclusion

This is not an exhaustive test, in fact it is just a documented approach for consideration to show a client what the minimum testing should be. As no dry run actually occurred, there may be inaccuracies and additions necessary to this document when first executed. I would need access to an appropriate configuration in order to perform a level of testing to complete this document.


About the Author

Ronald Bradford provides Consulting and Advisory Services in Data Architecture, Performance and Scalability for MySQL Solutions. An IT industry professional for two decades with extensive database experience in MySQL, Oracle and Ingres his expertise covers data architecture, software development, migration, performance analysis and production system implementations. His knowledge from 10 years of consulting across many industry sectors, technologies and countries has provided unique insight into being able to provide solutions to problems. For more information Contact Ronald.

BIGINT v INT. Is there a big deal?

The answer is yes.

In this face off we have two numeric MySQL data types, both Integer. In fact MySQL has 9 different numeric data types for integer, fixed precision and floating point numbers, however we are just going to focus on two, BIGINT and INT. This design consideration is part of my recent presentation Top 20 Design Tips for Data Architects.

What is the difference?
We turn to the MySQL Reference Manual first, in 10.1.1. Overview of Numeric Types we see the following.


INT[(M)] [UNSIGNED] [ZEROFILL]

A normal-size integer. The signed range is -2147483648 to 2147483647. The unsigned range is 0 to 4294967295.

BIGINT[(M)] [UNSIGNED] [ZEROFILL]

A large integer. The signed range is -9223372036854775808 to 9223372036854775807. The unsigned range is 0 to 18446744073709551615.

Ok, well an INT can store a value to 2.1 Billion, and an a BIGINT can store a value to some larger number to 20 digits. That MySQL search didn’t help much with details, we have to dig deeper to find 10.2. Numeric Types in which we find that INT is a 4 byte integer, and a BIGINT is an 8 byte integer.

So what’s the big deal?

Quite a lot actually. Using INT rather then BIGINT can make a significant reduction in disk space. Just this one change alone can save you 10%-20% (depends on your particular situation). More significantly, when used as a primary key, and for foreign keys and indexes, reducing your index size could be 50%, and this will improve performance when these indexes are used.

My approach is this. Let’s just focus on primary keys and foreign keys to begin with. Are you going to store more then 2.1 Billion rows in your table? The answer should be no? Should you say yes, then you do have grand plans, but you are also failing to consider the ramifications of handling larger data sets (a topic for later discussion).

There are exceptions to this rule, if you do a huge number of inserts and deletes, then while you may not have 2.1 Billion rows, you may have done 2.1 Billion inserts. Again better design practices should be considered in this case.

The Test

As with everything, we need some evidence to stake the claim. Using the Sakila sample database.

We start with a simple intersection table, that has a high number of numeric only columns. This will show the best case situation.

We will create two tables, one with all BIGINT columns, and one with all INT columns and then compare the size. These tables are only small, but they show the proportion of savings of disk space.

CREATE TABLE inventory_bigint LIKE inventory;
ALTER TABLE inventory_bigint
  MODIFY inventory_id  BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
  MODIFY film_id BIGINT UNSIGNED NOT NULL,
  MODIFY store_id BIGINT UNSIGNED NOT NULL;
INSERT INTO inventory_bigint SELECT * from inventory;
CREATE TABLE inventory_int LIKE inventory;
ALTER TABLE inventory_int
  MODIFY inventory_id  INT UNSIGNED NOT NULL AUTO_INCREMENT,
  MODIFY film_id INT UNSIGNED NOT NULL,
  MODIFY store_id INT UNSIGNED NOT NULL;
INSERT INTO inventory_int SELECT * from inventory;

select table_name,engine,row_format, table_rows, avg_row_length,
        (data_length+index_length)/1024/1024 as total_mb,
         (data_length)/1024/1024 as data_mb,
         (index_length)/1024/1024 as index_mb
from information_schema.tables
where table_schema='sakila'
and   table_name LIKE 'inventory%'
order by 6 desc;
+------------------+--------+------------+------------+----------------+-------------+-------------+-------------+
| table_name       | engine | row_format | table_rows | avg_row_length | total_mb    | data_mb     | index_mb    |
+------------------+--------+------------+------------+----------------+-------------+-------------+-------------+
| inventory_bigint | InnoDB | Compact    |     293655 |             51 | 43.60937500 | 14.51562500 | 29.09375000 |
| inventory_int    | InnoDB | Compact    |     293715 |             37 | 29.54687500 | 10.51562500 | 19.03125000 |
| inventory        | InnoDB | Compact    |     293707 |             33 | 22.54687500 |  9.51562500 | 13.03125000 |
+------------------+--------+------------+------------+----------------+-------------+-------------+-------------+
3 rows in set (0.15 sec)

In this example, the data portion decreased from 14MB to 10MB or 28%, and the index portion from 29M to 19M or 34%.

CREATE TABLE customer_bigint LIKE customer;
ALTER TABLE customer_bigint
     MODIFY customer_id BIGINT UNSIGNED NOT NULL AUTO_INCREMENT,
     MODIFY store_id BIGINT UNSIGNED NOT NULL,
     MODIFY address_id BIGINT UNSIGNED NOT NULL,
     MODIFY active BIGINT UNSIGNED NOT NULL;

CREATE TABLE customer_int LIKE customer;
ALTER TABLE customer_int
     MODIFY customer_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
     MODIFY store_id INT UNSIGNED NOT NULL,
     MODIFY address_id INT UNSIGNED NOT NULL,
     MODIFY active INT UNSIGNED NOT NULL;

select table_name,engine,row_format, table_rows, avg_row_length,
        (data_length+index_length)/1024/1024 as total_mb,
         (data_length)/1024/1024 as data_mb,
         (index_length)/1024/1024 as index_mb
from information_schema.tables
where table_schema='sakila'
and   table_name LIKE 'customer%'
order by 6 desc;

+-----------------+--------+------------+------------+----------------+-------------+-------------+-------------+
| table_name      | engine | row_format | table_rows | avg_row_length | total_mb    | data_mb     | index_mb    |
+-----------------+--------+------------+------------+----------------+-------------+-------------+-------------+
| customer_bigint | InnoDB | Compact    |     154148 |            139 | 37.09375000 | 20.54687500 | 16.54687500 |
| customer_int    | InnoDB | Compact    |     151254 |            121 | 30.06250000 | 17.51562500 | 12.54687500 |
| customer        | InnoDB | Compact    |      37684 |            125 |  7.81250000 |  4.51562500 |  3.29687500 |
| customer_list   | NULL   | NULL       |       NULL |           NULL |        NULL |        NULL |        NULL |
+-----------------+--------+------------+------------+----------------+-------------+-------------+-------------+
4 rows in set (0.22 sec)

In this example, the data portion decreased from 20MB to 17MB or 15%, and the index portion from 16M to 12M or 25%.

NOTE: The sample data set was increased for this example.

Conclusion

Even with these simple tables and small data sets it’s clear that INT is a saving of diskspace over BIGINT. In many clients I’ve seen huge savings in multi TB databases, just with a small number of schema optimizations. If this saving alone for a more optimized database design was only 10%, it is an easy 10% that will reflect a direct improvement in performance.


About the Author

Ronald Bradford provides Consulting and Advisory Services in Data Architecture, Performance and Scalability for MySQL Solutions. An IT industry professional for two decades with extensive database experience in MySQL, Oracle and Ingres his expertise covers data architecture, software development, migration, performance analysis and production system implementations. His knowledge from 10 years of consulting across many industry sectors, technologies and countries has provided unique insight into being able to provide solutions to problems. For more information Contact Ronald.

References

Off to OSCON

I will be heading to my first OSCON next week where I will be presenting MySQL Proxy: from Architecture to Implementation in conjunction with Giuseppe Maxia .

As was written by Colin Charles Our booth is yours… Sun at OSCON, Sun/MySQL would appear to also have a reasonable turnout. So it will be good to see some old colleagues and friends, and hopefully meet some new contacts.

While I am based on the East Coast, I do also provide expert MySQL consulting for clients in any location. Should you like to find out more about my offerings covering Architecture, Performance, Scaling, Migration and Knowledge Transfer for MySQL Solutions, please Contact Me and I will arrange a time to meet next week.

Why SQL_MODE is important? Part I

MySQL pre version 5.0 was very lax in it’s management of valid data. It was easy for data integrity to be abused if you knew how. The most common examples were truncations and silent conversions that if not understood could provide a serious data integrity issue.

In version 5.0, the introduction of SQL_MODE solved this problem. We will look at one example of how SQL_MODE can be enabled to provided improved data integrity.

You want to store the individual RGB (red/green/blue) decimal values of colors in a table. Each of these has a range from 0 to 255. You read that you can store 255 values in a TINYINT Integer data type, so you create a table like:

DROP TABLE IF EXISTS color_to_decimal;
CREATE TABLE color_to_decimal(
name VARCHAR(20) NOT NULL PRIMARY KEY,
red    TINYINT NOT NULL,
green TINYINT NOT NULL,
blue   TINYINT NOT NULL);

You insert some data like:

INSERT INTO color_to_decimal (name, red,green,blue) VALUES ('white',255,255,255);
INSERT INTO color_to_decimal (name, red,green,blue) VALUES ('black',0,0,0);
INSERT INTO color_to_decimal (name, red,green,blue) VALUES ('red',255,0,0);
INSERT INTO color_to_decimal (name, red,green,blue) VALUES ('green',0,255,0);
INSERT INTO color_to_decimal (name, red,green,blue) VALUES ('blue',0,0,255);
INSERT INTO color_to_decimal (name, red,green,blue) VALUES ('yellow',255,255,0);

Great, but when you look at you data you get?

SELECT     name, red, green, blue
FROM       color_to_decimal
ORDER BY name;
+--------+-----+-------+------+
| name   | red | green | blue |
+--------+-----+-------+------+
| black  |   0 |     0 |    0 |
| blue   |   0 |     0 |  127 |
| green  |   0 |   127 |    0 |
| red    | 127 |     0 |    0 |
| white  | 127 |   127 |  127 |
| yellow | 127 |   127 |    0 |
+--------+-----+-------+------+
6 rows in set (0.01 sec)

What happened, you delete the data and re-insert only to find no changes. You have been the victim of a silent conversion, via a means of truncation.

The TINYINT data type is 1 byte (8 bits). 8 bits can store the values from 0 to 255. When you use this integer data type, only 7 bits are actually available, which gives the range of 0 to 127. Why? Because MySQL reserved one bit for the sign, either positive or negative, even though you didn’t want a sign.

So knowing this, you go back and recreate your table with the following definition.

DROP TABLE IF EXISTS color_to_decimal;
CREATE TABLE color_to_decimal(
name VARCHAR(20) NOT NULL PRIMARY KEY,
red    TINYINT UNSIGNED NOT NULL,
green TINYINT UNSIGNED NOT NULL,
blue   TINYINT UNSIGNED NOT NULL);

You load your data and look at it again, and you see.

+--------+-----+-------+------+
| name   | red | green | blue |
+--------+-----+-------+------+
| black  |   0 |     0 |    0 |
| blue   |   0 |     0 |  255 |
| green  |   0 |   255 |    0 |
| red    | 255 |     0 |    0 |
| white  | 255 |   255 |  255 |
| yellow | 255 |   255 |    0 |
+--------+-----+-------+------+
6 rows in set (0.00 sec)

But, should you have been told about this, should there have been an error. Well, in MySQL this is actually a warning, and most applications never support and cater for warnings. It is only when you use the MySQL client program, as in these examples, you are given an indication, with the following line after each insert. If you look closely.

mysql> INSERT INTO color_to_decimal (name, red,green,blue) VALUES ('blue',0,0,255);
Query OK, 1 row affected, 1 warning (0.00 sec)


However there is a savior for this situation, and that is SQL_MODE.

When set to the setting TRADITIONAL, an error and not a warning is generated, and most applications support catching errors. Look at what happens in our example using the original table.

SET SQL_MODE=TRADITIONAL;
DROP TABLE IF EXISTS color_to_decimal;
CREATE TABLE color_to_decimal(
name VARCHAR(20) NOT NULL PRIMARY KEY,
red    TINYINT NOT NULL,
green TINYINT NOT NULL,
blue   TINYINT NOT NULL);
INSERT INTO color_to_decimal (name, red,green,blue) VALUES ('white',255,255,255);
ERROR 1264 (22003): Out of range value for column 'red' at row 1


As an added benefit you get this data integrity for free. We didn’t test it, because we know the data coming in is in the range of 0-255, but what if the user entered 500 for example. Let’s see.

TRUNCATE TABLE color_to_decimal;
SET SQL_MODE='';
INSERT INTO color_to_decimal (name, red, green, blue) VALUES('a bad color',500,0,0);
SELECT name, red, green, blue FROM color_to_decimal;
SET SQL_MODE=TRADITIONAL;
INSERT INTO color_to_decimal (name, red, green, blue) VALUES('a bad color',500,0,0);

Looking closely at the response in the client.

mysql> TRUNCATE TABLE color_to_decimal;
Query OK, 0 rows affected (0.02 sec)

mysql> SET SQL_MODE='';
Query OK, 0 rows affected (0.00 sec)

mysql> INSERT INTO color_to_decimal (name, red, green, blue) VALUES('a bad color',500,0,0);
Query OK, 1 row affected, 1 warning (0.00 sec)

mysql> SELECT name, red, green, blue FROM color_to_decimal;
+-------------+-----+-------+------+
| name        | red | green | blue |
+-------------+-----+-------+------+
| a bad color | 127 |     0 |    0 |
+-------------+-----+-------+------+
1 row in set (0.00 sec)

mysql> SET SQL_MODE=TRADITIONAL;
Query OK, 0 rows affected (0.00 sec)

mysql> INSERT INTO color_to_decimal (name, red, green, blue) VALUES('a bad color',500,0,0);
ERROR 1264 (22003): Out of range value for column 'red' at row 1

As discussed in my presentation Top 20 design tips for Data Architects the UNSIGNED column construct should be always defined unless there is a specific reason not to.

In some respects I would argue that the default for an Integer column should be actually UNSIGNED, and that SIGNED should be specified when you want a sign. Most integer columns generally in schema’s only contain positive numbers. There are of course plenty of examples, positional geo data, financial data, medical data for example.

One could also argue that MySQL should make the default SQL_MODE at least TRADITIONAL, and that only when you want backward compatibility should you then change the SQL_MODE.

This is Part I on SQL_MODE, there are few more interesting cases to discuss at a later time.


About the Author

Ronald Bradford provides Consulting and Advisory Services in Data Architecture, Performance and Scalability for MySQL Solutions. An IT industry professional for two decades with extensive database experience in MySQL, Oracle and Ingres his expertise covers data architecture, software development, migration, performance analysis and production system implementations. His knowledge from 10 years of consulting across many industry sectors, technologies and countries has provided unique insight into being able to provide solutions to problems. For more information Contact Ronald.

References

Sun Stock Prices

Sun Microsystem’s (NASDAQ:JAVA) hit a low this week of $8.71. There was a stronger rally and a close at $9.16 today. The financial times reports Sun Micro chief sees rays of hope, and Bloomberg Sun Rises After Fourth-Quarter Profit Tops Estimates.

I cashed out in March at $16.32, so that’s like a 50% drop in share price. I was lucky having been at MySQL long enough to have options to vest. Newer employees are not that lucky. I certainly hope MySQL Sun Employees get the Q4 weighted bonuses. (A structure I didn’t believe compensated with the old bonus structure).

I have been following more closely since Matt Asay’s comments in Who is buying Sun?



Image courtesy of Google Financial’s.

A Bill Gates bio

In the recent Wired magazine (yes, the paper one), there was an interesting time line of Bill Gates. It was rather an odd format, but I found the two page spread an enjoyable read. Some things of note in his early childhood, tips perhaps for us wanting to be successful.

  • 1968 Gates and Allen learn basic and blow entire school budget of computing time in a few weeks.
  • 1968 Employed to report PDP-10 software bugs.
  • 1971 Writes class scheduling program that places him in classes with the “right” girls.
  • 1973 Photographic memory, lucky him.
  • 1975 Writes Basic for MTIS.
  • 1976 Registers the trade name Microsoft.
  • 1980 Buys QDOS for $50k, later renaming and reselling as DOS.
  • 1984 Microsoft is one of the first software developers on Macintosh.
  • 1986 Company goes public.
  • 1996 Daily income is $30million, that’s per day.
  • 1998 The famous pie incident.

So it seems, finding bugs, using technology to meet the right women, buying and reselling somebody else’s work worked out, but will it work now.

I read elsewhere that the companies of today, such as Amazon, eBay, Google, FaceBook etc didn’t even exist 15 years ago.

There is still the opportunity for people out there, like you and I to be billionaires. Like one of my fridge magnets states. “Life isn’t about finding yourself. Life is about creating yourself.” Time to go create the next great thing we all must have.

Auditing your MySQL Data – Part 2

Continuing from my earlier post Auditing your MySQL Data, Roland has accurately highlighted that my initial post leaves out some important information for auditing. As the original charter was only to keep a history, for the purpose of comparing certain columns, a history was all that was needed.

Providing a history of changes forms the basis of auditing, and in keeping with my post title and intended follow-up, this is the all important second part. However in order to provide true auditing additional information is necessary. This includes:

  1. When was an operation performed
  2. What operation was performed, i.e. INSERT, UPDATE and DELETE
  3. Who performed the operation

Date and operation can be determined via the database, but in order to gather all this information, interaction with the application is necessary to obtain the true user information (This can’t be determined via a trigger)

The issue becomes a greater need for design understanding. What is the purpose of the audit data? How will it be accessed? How complex in maintaining the data do you wish to consider?

One alternative is keep a separate log of audit history. The benefits are a clear and easy way to provide a history of a users’ actions, and can preserve the structure of database table between the base and audit table can remain the same, triggers can remain relatively simply. However if you want to look at the data with audit history, it is better to embed these columns within each table, and triggers have to be customized and maintained in more detail the my original post.

When considering the progression of these points, the design process normally returns to the following conclusion. The following columns are added to the base table.

  • A create_timestamp column is added
  • A last_update_timestamp column is added
  • A last_update_user_id column is added

The create_timestamp is optional from an auditing perspective, because the last_update_timestamp of the first audit row will contain the same value, however experience has shown this column is valuable for other design considerations.

The only remaining issue is the type of operation, INSERT,UPDATE & DELETE. Both INSERT and UPDATE can be inferred, DELETE can not. To maintain the simplicity model, a common approach is to use a BEFORE DELETE trigger to insert an audit record with all the same values of the previous row, with the last_update_timestamp manually set. DELETE can then be determined via a no difference in any updated values.

It ultimate conclusion comes down your application design and needs. For example, your design for example may include a flag or row status for example to indicate deletes which are later cleaned up via a batch process so you don’t really care about the date/time of the actual purging of data. This then negates the need for any DELETE trigger.

Again, thanks to Roland for providing a link to Putting the MySQL information_schema to Use which provides a number of SQL statements that help in the generation of Triggers to support full auditing.

You should be aware that CURRENT_USER normally serves zero purpose if all changes are made via an application user.

At this time, you also have another design consideration. Do you introduce a procedure to re-create the triggers via an automated means for each schema change, or do you manually maintain triggers with schema changes. With each approach, additional checking and verification is necessary to ensure your triggers are correctly configured.

Auditing your MySQL Data

I was asked recently by a client to help with providing a history of data in certain tables. Like most problems, there is no one single solution, and in this case there are several possible solutions. I was able to provide a database specific only solution, with just minimal impact to the existing schema.

Here is my approach, your feedback and alternative input as always a welcome.

The problem

Client: I want to keep a history of all changes to two tables, and have a means of viewing this history.

For the purposes of this solution, we will use one table, called ‘customer’ from the Sakila Sample Database.

Solution

For tables to be audited, we will introduce a new column called ‘audit_id’ which is NULLABLE, and hopefully will not affect any existing INSERT statements providing column naming (a Best Practice) is used.
We do this to ensure that the Audit Table has both the same structure (number and ordering of columns), and can have a Primary Key defined.

Schema Preparation

mysql> USE sakila;
mysql> ALTER TABLE customer ADD audit_id INT UNSIGNED NULL;

We can then create an Audit Schema to store the Audit Table. This helps to ensure a clean schema and support for appropriate backup and recovery. Using a standard of suffixing existing schemas with ‘_audit';

mysql> CREATE DATABASE IF NOT EXISTS sakila_audit;
mysql> USE sakila_audit;
mysql> SET FOREIGN_KEY_CHECKS=0;   # (1)
mysql> CREATE TABLE customer LIKE sakila.customer;
mysql> ALTER TABLE customer DROP KEY `email`;   # (2)
mysql> # Foreign Keys (3)
mysql> ALTER TABLE customer
           DROP PRIMARY KEY,
           MODIFY customer_id SMALLINT UNSIGNED NOT NULL,  # (4)
           MODIFY audit_id INT UNSIGNED NOT NULL AUTO_INCREMENT PRIMARY KEY;

NOTES:
(1) Due to Foreign Key Constraints,
(2) We need to remove all existing UNIQUE Keys, I use an I_S SQL as shown below to find these programmatically, you can also view via SHOW CREATE TABLE customer;
(3) It appears that Foreign Keys are not created with the LIKE syntax. This may change in the future, so this note to check appropriately.
(4) We need to drop the primary key, but as this involves an AUTO_INCREMENT column we need to alter this as well, and this step involves naming the primary key column as well, which will vary per table.

We now have an Audit table in a separate schema that has the same columns. Part of the process of any new Schema Releases is to ensure these tables are kept in sync. An appropriate I_S statement could be used for verification. In this case, the support for a TRIGGER on Instance Startup to run, and throw an error to the error log or system error log (possible via UDF) would enable a balance check here.

Access to Audit Information

We provide a VIEW to the Audit table for history purposes. In addition, we use this for simplification of trigger management.

mysql> CREATE VIEW sakila.customer_history AS SELECT * FROM sakila_audit.customer;

NOTE: No index optimization has been performed on the Audit Table. It would be anticipated that existing Indexes could indeed be dropped and replaced with new indexes appropriate for data access.

Trigger Creation

In order to keep a copy of all data, we introduce two database triggers to manage a copy of all data in the history table. It is possible to say that History consists of the Current Version (in the base table) and all previous versions in the Audit Table. This requires 3 triggers. An alternative is to keep a full copy of all versions including current in the Audit Table. This requires 2 triggers, and takes more diskspace, however is a simpler and cleaner implementation.

USE sakila;
DELIMITER $$
DROP TRIGGER IF EXISTS customer_ari$$
CREATE TRIGGER customer_ari
AFTER INSERT ON customer
FOR EACH ROW
BEGIN
  INSERT INTO customer_history
  SELECT * FROM customer
  WHERE  customer_id = NEW.customer_id;
END;
$$
DROP TRIGGER IF EXISTS customer_aru$$
CREATE TRIGGER customer_aru
AFTER UPDATE ON customer
FOR EACH ROW
BEGIN
  INSERT INTO customer_history
  SELECT * FROM customer
  WHERE  customer_id = NEW.customer_id;
END;
$$
DELIMITER ;

NOTE: I do not generally like to use ‘SELECT *’ however in this situation, the trigger is significantly simplified. This is of benefit if you are maintaining audit triggers on many tables. The disadvantage is you must ensure your schema tables (e.g. sakila and sakila_audit) are always kept in sync with the same number and order of columns. Failing to add a column to the audit database will result in an error, which is a good confirmation. Failing to add a column in the right order, may corrupt your data. Exercise caution when modifying the schema in this situation.

Testing

As with any proper coding, we need to test this. The following sample SQL was run to test on a sample database.

SET FOREIGN_KEY_CHECKS=0;
USE sakila;
TRUNCATE TABLE sakila.customer;
TRUNCATE TABLE sakila_audit.customer;
SELECT 'no customer data', IF (count(*)=0,'OK','ERROR'),COUNT(*) AS VALUE from customer;
SELECT 'no customer history data', IF (count(*)=0,'OK','ERROR'),COUNT(*) AS VALUE from customer_history;
INSERT INTO customer (customer_id,store_id,first_name,last_name,email,address_id,active,create_date)
              VALUES(NULL,1,'mickey','mouse',',mickey@disney.com',1,TRUE,NOW());
SELECT 'customer data = 1 row', IF (count(*)=1,'OK','ERROR'),COUNT(*) AS VALUE from customer;
SELECT 'customer history data = 1 row', IF (count(*)=1,'OK','ERROR'),COUNT(*) AS VALUE from customer_history;
INSERT INTO customer(customer_id,store_id,first_name,last_name,email,address_id,active,create_date)
              VALUES(NULL,1,'donald','duck',',d.duck@warnerbros.com',1,TRUE,NOW());
SELECT 'customer data = 2 rows', IF (count(*)=2,'OK','ERROR'),COUNT(*) AS VALUE from customer;
SELECT 'customer history data = 2 rows', IF (count(*)=2,'OK','ERROR'),COUNT(*) AS VALUE from customer_history;
UPDATE customer SET email='donaldduck@warnerbros.com' where email='dduck@warnerbros.com';
SELECT 'customer data = 2 rows', IF (count(*)=2,'OK','ERROR'),COUNT(*) AS VALUE from customer;
SELECT 'customer history data = 3 rows', IF (count(*)=3,'OK','ERROR'),COUNT(*) AS VALUE from customer_history;
DELETE FROM customer where email='donaldduck@warnerbros.com';
SELECT 'customer data = 1 rows', IF (count(*)=1,'OK','ERROR'),COUNT(*) AS VALUE from customer;
SELECT 'customer history data = 3 rows', IF (count(*)=3,'OK','ERROR'),COUNT(*) AS VALUE from customer_history;
DELETE FROM customer;
SELECT 'customer data = 0 rows', IF (count(*)=0,'OK','ERROR'),COUNT(*) AS VALUE from customer;
SELECT 'customer history data = 3 rows', IF (count(*)=3,'OK','ERROR'),COUNT(*) AS VALUE from customer_history;

Improvements

It would be great if MySQL’s Procedural Language was a little more flexible and robust. Some improvements I’d love to see that would enable a more programmatic solution as the above contains a number of dependencies in schema_name and column_name.

  • Raise Error Handling to throw errors appropriately
  • Anonymous code block support, e.g. BEGIN …code… END; and an automatic execution, not the need to create a Procedure then execute it.
  • Ability to execute dynamic SQL more easily, for example CREATE DATABASE IF NOT EXISTS @variable; or CREATE VIEW @schema.@table_name_history FROM …
  • Support for multiple type (BEFORE|AFTER INSERT|UPDATE|DELETE) triggers per table.

INFORMATION_SCHEMA Query

mysql> SET @schema='sakila';
mysql> SET @table='customer';
mysql> SELECT DISTINCT CONCAT('ALTER TABLE ', table_name, ' DROP KEY `',constraint_name,'`;') AS cmd
           FROM INFORMATION_SCHEMA.table_constraints
           WHERE constraint_schema=@schema
           AND table_name=@table
           AND constraint_type='UNIQUE';

Focus on what you do best

When you have a great idea for a web application, it can be hard to consider with all the moving parts to focus just on what’s your uniqueness or differentiator from everybody else.

You may want to have control over your forums, comments, chat, photo management etc, i.e. user data, but how much does that help you. Is allocating resources to these features when plenty of completed applications exist distracting you and lengthening your time to market.

I always like to refer to Guy Kawasaki’s quote “Don’t worry, be crappy”. While I don’t necessarily agree with just throwing functionality out to the www, I believe in quality over quantity, you want to ensure that more time is spent in reviewing the input for new or improved features rather then bugs, bugs, bugs.


Ping.fm and Plurk are two new community driven sites that have leveraged the functionality of other sites, these being Get Satisfaction – People-Powered Customer Service for Everything! and Disq Us – Turn your blog comments into a webwide discussion.

There are advantages and disadvantages to this approach. As an smaller web site with a growing community, exposing what you do to a wider audience when using a third party to manage something can greatly help in exposure and associated marketing at no cost. On the down side, you are losing traffic to another site.

You need to ensure you can always get access to your data, and your community contributions. Ensuring adequate API’s for integration and data extraction are key. From a technology perspective, BitKeeper and LaunchPad come to mind. BitKeeper is a closed source, version control system that MySQL used. This was a killer for community contributions, where individual users simply could not contribute, and if they wanted even access to getting source code via the repository had to pay for an appropriate client. SourceForge and Apache are two examples of huge communities where they leverage the power of the community. LaunchPad is the latest kid on the block, but suffers from the fact that while access to applications hosted there are free, the actually LaunchPad code itself is closed. This has caused some issues.

It’s a fine line, and in the genre of software development, the Internet can create copies of anything just about overnight. More and more I hear about companies working in stealth mode rather then open community input and interaction, but that’s a topic for another discussion.

Getting Started with Simple DB

With my continued investigation of evaluating alternative data management with cloud computing options, I’m now evaluating Amazon Simple DB. Still in restricted beta, it helps to have a friend on the inside.

Working through the Getting Started Guide (API Version 2007-11-07) was ok, annoying in parts. Here are some issues I found. I was working with Java as the programming language.

  • The Docs enable you to view the language syntax in Java, C#, Perl, PHP, VB.NET, ScratchPad. You can also restrict the view to a specific language. A rather cool feature. One observation is there is no Python, which is rather ironic as my first investigation was Google App Engine (GAE), and the only language here is Python. Something I had to learn first.
  • Preparing the Samples asks you download the Amazon SimpleDB Sample code, but this is not actually a link to the sample code but an index to Community Code. I used Java Library for Amazon SimpleDB which wasn’t even on the first page of results.
  • The supplied docs for specifying the Classpath was rather wrong, helps to simply find all .jar files and included these. Mine looks like:
    • #!/bin/sh
      #  http://docs.amazonwebservices.com/AmazonSimpleDB/2007-11-07/GettingStartedGuide/?ref=get-started
      
      SDB_HOME="/put/directory/to/unzip/here"
      export CLASSPATH=$CLASSPATH:
      $SDB_HOME/src/com/amazonaws/sdb/samples/
      $SDB_HOME/lib/amazon-simpledb-2007-11-07-java-library.jar:
      $SDB_HOME/third-party/log4j-1.2.14/log4j-1.2.14.jar:
      $SDB_HOME/third-party/commons-codec-1.3/commons-codec-1.3.jar:
      $SDB_HOME/third-party/commons-logging-1.1/commons-logging-1.1.jar:
      $SDB_HOME/third-party/jaxb-ri-2.1/jaxb-xjc.jar:
      $SDB_HOME/third-party/jaxb-ri-2.1/activation.jar:
      $SDB_HOME/third-party/jaxb-ri-2.1/jaxb-impl.jar:
      $SDB_HOME/third-party/jaxb-ri-2.1/jaxb-api.jar:
      $SDB_HOME/third-party/jaxb-ri-2.1/jsr173_1.0_api.jar:
      $SDB_HOME/third-party/commons-httpclient-3.0.1/commons-httpclient-3.0.1.jar
      
  • All examples in the docs then refer to making changes such as “invokeCreateDomain(service, action); line and add the following lines after // @TODO: set action parameters here:”, problem is all the samples don’t have the action variable, but rather a variable called request. The comment in the code ” // @TODO: set request parameters here” is at least accurate.
  • The docs contain a lot of Java syntax that is would not for example compile correctly. Plenty of occurances of a missing semicolon ‘;’
  • Each example defines
    String accessKeyId = ““;
    String secretAccessKey = “
    “;
    Ok for the first example, but as soon as I moved to the second, I re-factored these into a Interface called Constants.
  • In all the examples, they never provide any sample output, this would help just to confirm stuff. In Java Library for Amazon SimpleDB download link, there is an example output, but it’s outdated, with new data attribute called BoxUsage. My output is:
    • CreateDomain Action Response
      =============================================================================
      
          CreateDomainResponse
              ResponseMetadata
                  RequestId
                      f04df8eb-71fa-4d4e-9bd5-cc98e853a2e4
                  BoxUsage
                      0.0055590278
      
  • And now some specifics. In a Relational database such as MySQL, you have Instance/Schema/Table/Column. Within SimpleDB it would indicate that you need a separate AWS account for Instance management. That’s probably a good thing as it will enable tracking of costs. There appears to be no concept of a Schema. Data is stored in Domains, this is the equivalent to a Table. Within each Domain, you specific Attributes, a correlation with Columns. One key difference is the ability to define a set of Attributes with the same identity (much like a list that is supported via Python/GAE). For any row of data, you must specify an itemName, this being equivalent to a Primary Key.These names table me back to old days (20 years ago) of Logical Data Models that used entities,attributes and relationships.
  • The term Replace is used when updating data for a given row.
  • When retrieving data, you first return a list of itemNames, then you can via Attributes for that given item.
  • You can perform a simple where qualification using a Query Expression, including against multiple Attributes via intersection syntax
  • A observation that is of significant concern is the lack of security against any type of operation. The Getting Started guide ends with Deleting the Domain. Is there no means to define permissions against type of users, such as an Application User, and a Database Administrator for managing the objects.

Well it took me longer to write this post, then to run through the example, but at least on a lazy Sunday afternoon, a first look at SimpleDB was quite simple.

I did also run into an error initially. I first started just via CLI under Linux (CentOS 5), but switched back to installing Eclipse on Mac OS/X for better error management, and of course this error didn’t occur.

Do you store credit cards in your MySQL Database?

The Payment Card Industry Data Security Standard (PCI DSS) has been developed to help organizations that process card payments to prevent credit card fraud, cracking and various other security vulnerabilities and threats.

This has been developed by the major credit card companies such as MasterCard and Visa. If one of the companies that created the standard, Mastercard International uses PCI General for MySQL then you would be confident that the software is of the highest quality to satisfy all requirements.

A few questions to consider.

Q: Why is PCI compliance important?
A: Credit Card companies will start to demand organizations that store credit card numbers have adequate security of their (as in the credit card company) data.

Q: How can I support PCI compliance with minimal impact?
A: Any solution that requires coding changes and then the necessary testing and verification will incur a large cost for a successful deployment. A turn key solution that can be implemented in a near seamless manner without code changes is ideal for any company.

PCI General for MySQL achieves this. With the security and encryption managed at the Operating System kernel level, MySQL data, and application communications is totally secure. Introducing PCI General via MySQL replication and with a controlled fail-over to a MySQL slave running under PCI General is the simplest and easiest method of introducing PCI Compliance into your production environment

For more information, please visit PCI General for MySQL or Contact Me and I’ll direct you to additional information.

Facebook performance woes today

It seems that of late a number of successful community web sites have been experiencing problems in scalability and performance. Today it’s Facebook.

Initially I got a “Problem loading page” browser message. No big deal, you get those. A few refreshes and I’m in. I was trying to send a message, I got the following popup error message.

No Network
Transport error (#1001) while retrieving data from endpoint `/inbox/ajax/ajax.php': A network error occurred. Check that you are connected to the internet.

Well yes, I’m connected to the Internet, every other site is fine. A more technical look gives the lovely connection aborted message.

lynx http://facebook.com

Looking up facebook.com
Making HTTP connection to facebook.com
Sending HTTP request.
HTTP request sent; waiting for response.
HTTP/1.1 302 Found
Data transfer complete
HTTP/1.1 302 Found
Using http://www.facebook.com/common/browser.php
Looking up www.facebook.com
Making HTTP connection to www.facebook.com
Sending HTTP request.
HTTP request sent; waiting for response.
Retrying as HTTP0 request.
Looking up www.facebook.com
Making HTTP connection to www.facebook.com
Sending HTTP request.
HTTP request sent; waiting for response.
Alert!: Unexpected network read error; connection aborted.
Can't Access `http://www.facebook.com/common/browser.php'
Alert!: Unable to access document.

lynx: Can't access startfile

Deleting from ARCHIVE tables

I can’t say I’ve used the ARCHIVE storage engine before, but at the NY MySQL Meetup last night there was discussion of the improvements to ARCHIVE in 5.1 and the fact that you could not DELETE from archive. A simple test confirmed this indeed throws an error.

DROP TABLE IF EXISTS url_log;
CREATE TABLE url_log(
log_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
log_date TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
user_id INT UNSIGNED NULL,
url VARCHAR(100) NOT NULL,
PRIMARY KEY (log_id))
ENGINE=ARCHIVE;

DELETE FROM url_log;
ERROR 1031 (HY000): Table storage engine for 'url_log' doesn't have this option

However, part of MySQL 5.1 which is RC status, there is partitioning. Thinking that one could probably partition say a log table by DAY OF MONTH, and then you could do what you want with the data in a partition and delete the partition, I tried the following test.
NOTE: for the purposes of testing, I used SECOND() rather then DAY() and smaller ranges for simplicity.

DROP TABLE IF EXISTS url_log;
CREATE TABLE url_log(
log_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
log_date TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
user_id INT UNSIGNED NULL,
url VARCHAR(100) NOT NULL,
PRIMARY KEY (log_id))
ENGINE=ARCHIVE
PARTITION BY RANGE ( SECOND(log_date) ) (
    PARTITION p0 VALUES LESS THAN (10),
    PARTITION p1 VALUES LESS THAN (20),
    PARTITION p2 VALUES LESS THAN (30),
    PARTITION p3 VALUES LESS THAN MAXVALUE
);

However this throws an error.

ERROR 1503 (HY000): A PRIMARY KEY must include all columns in the table's partitioning function

Primary keys, and AUTO_INCREMENT do not play well with partitioning, so for the purpose of this proof of concept, I’ll drop these.

DROP TABLE IF EXISTS url_log;
CREATE TABLE url_log(
log_date TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
user_id INT UNSIGNED NULL,
url VARCHAR(100) NOT NULL)
ENGINE=ARCHIVE
PARTITION BY RANGE ( SECOND(log_date) ) (
    PARTITION p0 VALUES LESS THAN (10),
    PARTITION p1 VALUES LESS THAN (20),
    PARTITION p2 VALUES LESS THAN (30),
    PARTITION p3 VALUES LESS THAN MAXVALUE
);

Create a simple Stored Procedure to randomly generate some data. The function is not efficient using RAND() and SLEEP() but it does provide the generation of some data.

DELIMITER $$
DROP PROCEDURE IF EXISTS load_url_log;
CREATE PROCEDURE load_url_log (insert_count INT)
BEGIN
  DECLARE i INT DEFAULT 1;

  WHILE i < insert_count
  DO
     INSERT INTO url_log(user_id, url)
     VALUES (FLOOR(RAND()*100), CONCAT(REPEAT('x',FLOOR(RAND()*99)),SLEEP(RAND())));
  END WHILE;
END $$

DELIMITER ;

CALL load_url_log(500);

I quick check shows a distribution of data.

mysql> select distinct(second(log_date)) from url_log;
mysql> select distinct(second(log_date)) from url_log where second(log_date) < 10;
+--------------------+
| (second(log_date)) |
+--------------------+
|                  0 |
|                  1 |
|                  2 |
|                  3 |
|                  4 |
|                  5 |
|                  6 |
|                  7 |
|                  8 |
|                  9 |
+--------------------+
10 rows in set (0.00 sec)

And now the purpose of the test. Deleting data via deleting a partition.

mysql> ALTER TABLE url_log DROP PARTITION p0;
Query OK, 0 rows affected (0.01 sec)
Records: 0  Duplicates: 0  Warnings: 0

mysql> select distinct(second(log_date)) from url_log where second(log_date) < 10;
Empty set (0.00 sec)

And it works. Re-creating however did not.

ALTER TABLE url_log ADD PARTITION (PARTITION p0 VALUES LESS THAN (10));
ERROR 1481 (HY000): MAXVALUE can only be used in last partition definition

Ok, so in my example I was lazy, I didn’t create specific partitions as I would in real world here, e.g. 31 partitions for DAYS. Simulate a little better.

DROP TABLE IF EXISTS url_log;
CREATE TABLE url_log(
log_date TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
user_id INT UNSIGNED NULL,
url VARCHAR(100) NOT NULL)
ENGINE=ARCHIVE
PARTITION BY RANGE ( SECOND(log_date) ) (
    PARTITION p0 VALUES LESS THAN (10),
    PARTITION p1 VALUES LESS THAN (20),
    PARTITION p2 VALUES LESS THAN (30),
    PARTITION p3 VALUES LESS THAN (60)
);
ALTER TABLE url_log DROP PARTITION p0;
ALTER TABLE url_log ADD PARTITION (PARTITION p0 VALUES LESS THAN (10));
ERROR 1493 (HY000): VALUES LESS THAN value must be strictly increasing for each partition


Still doesn’t work. RTFM indicates this.

DROP TABLE IF EXISTS url_log;
CREATE TABLE url_log(
log_date TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
user_id INT UNSIGNED NULL,
url VARCHAR(100) NOT NULL)
ENGINE=ARCHIVE
PARTITION BY LIST ( SECOND(log_date) ) (
    PARTITION p0 VALUES IN (0,1,2,3,4,5,6,7,8,9),
    PARTITION p1 VALUES IN (10,11,12,13,14,15,16,17,18,19),
    PARTITION p2 VALUES IN (20,21,22,23,24,25,26,27,28,29),
    PARTITION p3 VALUES IN (30,31,32,33,34,35,36,37,38,39),
    PARTITION p4 VALUES IN (40,41,42,43,44,45,46,47,48,49),
    PARTITION p5 VALUES IN (50,51,52,53,54,55,56,57,58,59)
);

ALTER TABLE url_log DROP PARTITION p0;
ALTER TABLE url_log ADD PARTITION (PARTITION p0 VALUES IN (0,1,2,3,4,5,6,7,8,9));

And that works.

So by creating an ARCHIVE table with 31 LIST partitions, one for each day of the month, you could use ARCHIVE to log data into DAY partitions, analyze, summarize,log,copy the data from the previous day, and purge it within 28 days.

Check your spelling

I’ve been Plurking more lately rather then Twittering. I’d like to offer to help out at Twitter if I could find the right person to talk to.

I’m no English major, but I do like to ensure my spelling is correct (at least for the bulk of the audience). You see grammar problems on sites, due to the nature of English not being the first language of many people, but one should always check your spelling as per this popup message I got today.

Time Warner Cable Speed

I had my Time Warner cable installed yesterday, a rather painless process. Reported as having Internet speeds of 10MB down and 1/2 MB up, these were confirmed with speedtest.net

Installation – July 5

A day later – Jul 6