Thu, 14 Aug 08
Posted in MySQL at 12:27 pm by moore
Just listened to another mysql webinar. This time it was about mysql and memcached. I got in a little late, but here are my notes.
Memcached is a distributed caching mechanism. Not as robust as Coherence, for example, but free, open source, and widely used and supported.
There is a project that hooks mysql UDFs to memecached.
The main speaker worked at Grazr, a feed manipulation company. He presented some lessons and stated that some seemed obvious, but we all know how it is, you learn mistakes the hard way.
- Beware architecture momentum–early decisions influence later ones.
- Your scaling plan is wrong in some way!–grazr overbuilt.
- Testing is limited: they had issues with replication speed. Users are better at testing site than any perl script.
- Replication is fast, until it isn’t. It’s asynchronous nature can make it hard to program to.
- Memcached is your friend
- Sphinx, an open source SQL full text search engine (like Lucene), is your friend. It also has a mysql storage engine–virtual table that you can join with other data sources. This sounds cool.
- Do bulk insert and lazy writes–if you don’t need to immediately read it, do the write later.
- Emphasis on scaling hurt user experience. Fast, data correct transaction are available with two options: one server, or a sync buffer like memcached.
- Use instrumentation–he mentioned tools like nagios and cacti (cacti has mysql plugin [which I couldn’t find]) to help scaling–close the loop: when you make a change, find out what that change did.
- Try new things–for example, the MySQL blackhole storage engine for data that is only read on slaves, not on the master.
- Everyone has same problems–avoid Not Invented Here syndrome. Not just the obvious stuff (who writes a web server these days–oh wait), for example, there’s a batch job processor like gearman. Note that he also said they had to Invent a few things before they learned this lesson.
- Accept change–design for now+1, not now+100
- Listen–even folks not intimately familiar with your technology may have good ideas, and more brains are better.
End notes
Looks like a couple of interesting whitepapers are available (you’ll have to register and get pinged via email a few times by sales folks): “Designing and Implementing Scalable Applications with Memcached and MySQL”. And there are future webinars on the MySQL/memcached topic, one on SAAS applications and one on Scalable web architecture.
In a meta note, I enjoy the webinars because:
- Easy to get to
- Free
- High quality
- One can bail if you want (which I did a few weeks ago when MySQL proxy http://forge.mysql.com/wiki/MySQL_Proxy didn’t turn out to be all that interesting to me).
- The give me something to blog about
Technorati Tags: webinar, memcached, users are the best testers
Permalink
Mon, 11 Feb 08
Posted in Programming, MySQL at 11:37 am by moore
Last Tuesday, I attended a MySQL webinar. I registered on the MySQL website (with a site-specific email address, of course) and periodically am invited to these webinars. I’d tried to attend in the past, but something else (usually billable) always interfered. Not this time!
The talk was titled “Performance Coding for MySQL” and the author, Jay Pipes, did a fantastic job. (He is also the co-author of Pro MySQL.) The slides from the presentation are up, and he also answered questions sent to him during the presentation in some detail as well. His presentation, about an hour in length, covered both basics (like, normalize first (slide 4), think in sets rather than iterators (slides 20-23)–basic, but not intuitive), and under the hood intricacies (like, think about the size of your primary keys and consider record size (slide 6), avoiding deletes with MyISAM (slide 27) and vertical partitioning to take advantage of the query cache (slides 9-11) ) . He also pointed to a script that he wrote to find useless indices.
Well worth my time. Thanks MySQL and Jay, for making a resource like this available and free! There’s a whole lot more, so I’d recommend downloading the slides and giving them a run through, if you interact with MySQL as a DBA or a developer (or, as is often the case, both).
(On that note, I’d like to recommend the MySQL DBA blog for your perusal–apparently recently renamed the ‘Senior MySQL DBA’ blog, heh.)
Technorati Tags: mysql dba, think in sets, webinar
Permalink
Fri, 04 Jan 08
Posted in Java, Databases, MySQL at 7:59 pm by moore
I ran into a crazy error last week. One of my clients was upgrading from MySQL4 to MySQL5. The application in question was using Hibernate 3.2. Here’s the table structure, and the hibernate bean definition. See if you can spot the issue:
mysql> desc stat;
+--------------+-------------+------+-----+---------------------+-------+
| Field | Type | Null | Key | Default | Extra |
+--------------+-------------+------+-----+---------------------+-------+
| stat_date | date | | PRI | 0000-00-00 | |
| stat_type | varchar(50) | | PRI | | |
| stat_count | int(11) | YES | | NULL | |
| last_updated | datetime | | | 0000-00-00 00:00:00 | |
+--------------+-------------+------+-----+---------------------+-------+
4 rows in set (0.00 sec)
<class name="com.foo.common.data.Statistic" table="stat" lazy="false">
<cache usage="read-write"/>
<composite-id name="statisticId" class="com.foo.common.data.StatisticId">
<key-property name="date" type="java.util.Date" column="stat_date"/>
<key-property name="type" column="stat_type"/>
</composite-id>
<property name="count" column="stat_count"/>
<property name="lastUpdated" type="java.util.Date" column="last_updated" />
</class>
The exception stack trace I was seeing was something like this:
2007-12-31 14:15:09,888 ERROR [Thread-14] def.AbstractFlushingEventListener (AbstractFlushingEventListener.java:301)
- Could not synchronize database state with session
org.hibernate.exception.ConstraintViolationException: Could not execute JDBC batch update
at org.hibernate.exception.SQLStateConverter.convert(SQLStateConverter.java:71)
at org.hibernate.exception.JDBCExceptionHelper.convert(JDBCExceptionHelper.java:43)
at org.hibernate.jdbc.AbstractBatcher.executeBatch(AbstractBatcher.java:249)
at org.hibernate.engine.ActionQueue.executeActions(ActionQueue.java:235)
at org.hibernate.engine.ActionQueue.executeActions(ActionQueue.java:139)
at org.hibernate.event.def.AbstractFlushingEventListener.performExecutions(AbstractFlushingEventListener.java:298)
at org.hibernate.event.def.DefaultFlushEventListener.onFlush(DefaultFlushEventListener.java:27)
....
Caused by: java.sql.BatchUpdateException: Duplicate key or integrity constraint
violation message from server: "Duplicate entry '2007-12-31-stattype' for key 1"
at com.mysql.jdbc.PreparedStatement.executeBatch(PreparedStatement.java:1492)
at com.mchange.v2.c3p0.impl.NewProxyPreparedStatement.executeBatch(NewProxyPreparedStatement.java:1723)
at org.hibernate.jdbc.BatchingBatcher.doExecuteBatch(BatchingBatcher.java:48)
at org.hibernate.jdbc.AbstractBatcher.executeBatch(AbstractBatcher.java:242)
... 57 more
I ended up turning on the mysql logging (the log setting in the my.ini file, which logs all sql statements mysql makes) to see what was happening.
Basically, I was looking to see if an entry in the stat table existed; if it did, increment and update, if it did not, insert. And the insert was always happening, so the entry was not found–it did exist because mysql threw the ‘integrity constraint’ exception.
The cause of the issue was the date type of stat_date and the fact that I incorrectly mapped it to java.util.Date. It really should have been mapped to java.sql.Date. How this worked in mysql4 is beyond me, but it did. Changing the hibernate dialect to mysql5 had no impact.
Technorati Tags: mysql upgrade, hibernate
Permalink
Wed, 11 Apr 07
Posted in Databases, MySQL at 4:18 pm by moore
If you’re trying to tune MySQL, make sure you measure before and after. mysqlreport is a great way to measure a number of aspects of mysql (make sure you check out the guide). If you want graphing and long term tracking for mysqlreport and just about any other measure you care to track, I recommend cacti, which works with rrdtool.
Technorati Tags: mysql performance, monitoring
Permalink
Wed, 27 Sep 06
Posted in Programming, Databases, MySQL at 11:57 am by moore
I just re-read “How To Lie With Statistics”, which is so good I think it should be required reading in every middle school. In it, the author makes the point that there are three kinds of ‘averages’: arithmetic mean, median and mode (here I am, contributing to Wikipedia’s dominance, due to my laziness in looking up alternative definitions of statistical concepts). In general, the median is the most informative average, because it’s not skewed by a small number of outliers.But mysql (and other databases I’ve worked on) don’t natively supprt the medan, whereas I believe most support average (by which they mean ‘arithmetic mean’). Sure, you can use a stored procedure (as suggested here for PostgreSQL. However, I’m working with MySQL 4, which does not support stored procs. However, there is another solution: user defined functions. These seem like stored procedures, except you have to write them in C (or C++).
Now, I’m not a C programmer. Luckily, someone has written and released a set of mysql user defined functions that include median (as well as many other statistical manipulations). The bad news is that it hasn’t been updated for years. The good news is that with a bit of luck and many downloads, I was able to get the median function working on mysql, both on windows as a dll, and on linux as a shared library. To repeat, I am not a C programmer, so if you see any head thumping errors below, please let me know and I’ll update this document.
First off, I was working with these versions of mysql: c:\Program Files\MySQL\MySQL Server 4.1\bin\mysql.exe Ver 14.7 Distrib 4.1.10a, for Win95/Win98 (i32) and mysql Ver 14.7 Distrib 4.1.7, for pc-linux (i686)
To get median working on windows, you need to:
- Download the mysql-udf tarball.
- patch the files if you’re running a version of mysql greater than 4.1.1. patch available here, or the patched tarball is here.
- Download and install Visual C++ Express. (If you have a C compiler on Windows, you can skip this step and the next. Oh, and the ones following that will probably be different. (Here’s a blog post about creating a UDF using Visual Studio C++ 2003.)
- Download and install the platform SDK; I only followed through step 3. If you don’t, you’ll get ‘windows.h’ errors when you try to compile the UDF.
- Untar the myslq-udf tarball. Patch if needed.
- Install the mysql header files. I was able to do this via the Windows Installer, which let me modify my existing mysql installation; I had to add the ‘C Include Files / Lib Files’ feature.
- Create a new directory. Copy
udf_median.cc from the untarred directory to this new directory.
- Create a new file in that directory called
udf_median.def. This file contains all the methods the UDF is exporting. Or you can just download the file I used here.
- Open Visual C++ Express
- Choose File / New / Project From Existing Code. Hit Next. Browse to the directory you just created. Create a name for the project. Hit Finish
- Edit the
udf_median.cc file and comment out the #ifdef HAVE_DLOPEN line as well as the corresponding #endif. If I didn’t do this, I kept getting link errors, as I guess everything between those preprocessor directives was not being compiled.
- Add the mysql include files: right click on the project and choose properties. Expand ‘Configuration Properties’ then ‘C/C++’ and click ‘General’. On the right, add an include directory. Navigate to the Mysql include directory and add that.
- Add the module definition file: right click on the project and choose properties. Expand ‘Configuration Properties’ then ‘Linker’ and click ‘Input’. Add ‘udf_median.def’ to the key ‘Module Definition File’.
- Make sure VC knows this is a DLL: right click on the project and choose properties. Expand ‘Configuration Properties’ and click ‘General’. Choose ‘Dynamic Library (.dll)’ for Configuration Type. If you don’t do this, you’ll get errors like:
error LNK2019: unresolved external symbol _WinMain because the compiler thinks you’re trying to build an application.
- Right click on the project and choose ‘Build’. This gives you a DLL in the Debug directory.
- Copy the DLL to the bin directory of your mysql installation.
- Create the function by logging in to mysql and running this command:
CREATE AGGREGATE FUNCTION median RETURNS REAL SONAME 'udf_median.dll';. (The user you log in as will need to have the ability to insert rows into the mysql tables.)
- Test and enjoy.
Deploying the UDF to linux is much simpler, mostly because you don’t have to install a compiler, linker, etc. I used ‘gcc (GCC) 3.3.4′.
- Download the mysql-udf tarball.
- patch the files if you’re running a version of mysql greater than 4.1.1. patch available here, or the patched tarball is here.
- Untar the myslq-udf tarball. Patch if needed.
- Edit the
udf_median.cc file and comment out the #ifdef HAVE_DLOPEN line as well as the corresponding #endif.
- Compile and link the code. Do not use the instructions on the mysql-udf homepage. If you compile with those flags, you’ll get this error when you try to add the function:
mysql> CREATE AGGREGATE FUNCTION median RETURNS REAL SONAME 'udf_median.so';
ERROR 1126 (HY000): Can't open shared library 'udf_median.so' (errno: 22 /usr/lib/udf_median.so: undefined symbol: _Znwj). Rather, use the instructions in this bug report: gcc -shared -lstdc++ -I /usr/include -I /usr/local/include -I /usr/local/mysql/include/ -o udf_median.so udf_median.cc'
- Copy the shared library to a directory where mysql will see it. I put it in
/usr/lib.
- Create the function by logging in to mysql and running this command:
CREATE AGGREGATE FUNCTION median RETURNS REAL SONAME 'udf_median.so';. (The user you log in as will need to have the ability to insert rows into the mysql tables.)
- Test and enjoy.
We have pushed this UDF to production with replicated servers and haven’t seen any issues with it yet.
I want to extend my thanks to:
Technorati Tags: median, mysql, user defined functions
Permalink