Shortly after I posted my last summary of MySQL releases, our son Mats was born and I went on a 2.5-week vacation. Our developers did not rest in the meanwhile and I'd like to give you a quick update of what's new since then:
If you have enjoyed my blog postings about MySQL Cluster please come to the MySQL Cluster tutorial at the 2010 O'Reilly MySQL Conference & Expo. I shall be giving this tutorial along with my colleagues Geert Vanderkelen and Andrew Morgan.
The outline of this tutorial is as follows:
"This tutorial aims to guide normal MySQL users and DBAs into the world of MySQL Cluster. From installing and configuring to creating your first clustered table and finally node failure handling. At the end of the session you’ll will be the proud owner of a full blown, tiny MySQL Cluster which you can show off at work. The data nodes will gently keep you lap warm when the …
[Read more]
MySQL Cluster is well known as an in-memory database which can be
restrictive (a server typically has a RAM limit) so it may
surprise some to learn that you can also store MySQL Cluster data
on disk. Disk Data Tables have been in MySQL Cluster for
quite a while, the first GA release with them included was
6.2. They do have caveats which I will describe here, many
of which already have plans to be improved in future
versions.
How Disk Data Tables work
A set of UNDO logs and data files can be created which are on every data node in the cluster. The UNDO logs bring the disk data back to a consistent point so that during node restart the REDO logs can be replayed when a node is started.
Not all data in a disk table is actually stored on disk, the main two things that are not are indexed columns (the entire column is in main memory, not just the index) and the first 256 bytes of a …
[Read more]
When I saw Shlomi's post on why not to use apt-get or yum for
MySQL, I thought immediately that his conclusions are quite
reasonable. What you get from the Linux distributions is not the
same thing that you find in the official
MySQL downloads page. Now, whether you value more the
completeness of the server or the ease of administration through
the distribution installation tools, it's up to you and your
business goals. We at the MySQL team have organized a meeting
with the Linux distributions with the intent of finding out which
differences and problems we may have with each other, and to
solve them by improving communication. What follows is a summary
of what happened in Brussels during the meeting.
…
A preliminary schedule is now available for the MySQL Conference & Expo 2010 (in Santa Clara, California, USA). I have two talks and a tutorial, currently scheduled as follows:
Schedule might change a bit, so I …
[Read more]A preliminary schedule is now available for the MySQL Conference & Expo 2010 (in Santa Clara, California, USA). I have two talks and a tutorial, currently scheduled as follows:
Schedule might change a bit, so I …
[Read more]MySQL Cluster can be used as a general purpose transactional storage engine, but if you convert all your InnoDB tables to it and connect your application straight to it you may not see the performance you were hoping for. This is because MySQL Cluster was originally designed for real-time telecommunications applications (such as RADIUS servers). It has slowly been modified to become more general purpose and improvements are being made every day but there are still some performance considerations which go with this. In some cases tweaking your schema and/or queries can help performance dramatically, so I shall try and outline some of things to watch for here.
Indexes
The fastest type of lookup you can do in Cluster is a primary key equality lookup (ie. SELECT * FROM table WHERE pkey = 2). This is because the primary key is stored as a hash index as well as an optional ordered index. This hash index is …
[Read more]Bernhard blogged over at http://ocklin.blogspot.com/2010/02/mysql-cluster-711-is-there.html that MySQL Cluster 7.1.1 Beta has been released. The big feature (from my point of view) is the SQL interface on top of NDB$INFO. This means there is now full infrastructure from the NDB data nodes right out to SQL in the MySQL Server for adding monitoring to any bit of the internals of the data nodes.
When MySQL Cluster data nodes start they need to go through a process of determining roles in the cluster, copying the data back into RAM and synchronising everything up. This can take longer than expected and the process is not always very verbose. So in this post I will outline a simplified version of what MySQL Cluster is doing in each start phase so that you can see why it can take time.
You can see the last completed start phase by following the management logs or you can see the current start phase by doing 'x status' in ndb_mgm where 'x' is the node ID of the data node you want the information for.
Start Phase 0
This is part of the basic initialisation, and the configuration is received. If this is an initial start then the node's file system is created.
Start Phase 1
All the blocks that are not already running are initialised in this phase. Connections to other data …
[Read more]If there is one thing that confuses people about tables in MySQL Cluster (including me at times) it is BLOB/TEXT columns. When NDB was originally created it was not designed to handle BLOB data, so the handling of BLOB data was difficult to implement and is sometimes not exactly what users expect.
How MySQL Cluster BLOBs work
When you create a table in MySQL Cluster which has a BLOB column the first 256 bytes of the BLOB is stored in the main table (and in memory when using disk data tables), subsequent data is then stored in a hidden table (typically split into 2KB rows). This means there is an extra table for every BLOB or TEXT column in your main table (and extra resource usage).
BLOB locking in MySQL Cluster
These extra tables can cause some problems, firstly with performance (retrieving BLOB data is not very fast) and more importantly with locking. MySQL Cluster works in …
[Read more]