Frequently Asked Questions

Are there Known Bugs? When is the Next Release?
Is this Database Engine Open Source?
My Query is Slow
How to Create a New Database?
How to Connect to a Database?
Where are the Database Files Stored?
What is the Size Limit (Maximum Size) of a Database?
Is it Reliable?
Why is Opening my Database Slow?
Column Names are Incorrect?
Is the GCJ Version Stable? Faster?
How to Translate this Project?

Are there Known Bugs? When is the Next Release?

Usually, bugs get fixes as they are found. There is a release every few weeks. Here is the list of known and confirmed issues:

Tomcat and Glassfish 3 set most static fields (final or non-final) to null when unloading a web application. This can cause a NullPointerException in H2 versions 1.1.107 and older, and may still not work in newer versions. Please report it if you run into this issue. In Tomcat >= 6.0 this behavior can be disabled by setting the system property org.apache.catalina.loader.WebappClassLoader.ENABLE_CLEAR_REFERENCES=false, however Tomcat may then run out of memory. A known workaround is to put the h2*.jar file in a shared lib directory (common/lib).
Some problems have been found with right outer join. Internally, it is converted to left outer join, which does not always produce the same results as other databases when used in combination with other joins.
When using Install4j before 4.1.4 on Linux and enabling pack200, the h2*.jar becomes corrupted by the install process, causing application failure. A workaround is to add an empty file h2*.jar.nopack next to the h2*.jar file. This problem is solved in Install4j 4.1.4.

Is this Database Engine Open Source?

Yes. It is free to use and distribute, and the source code is included. See also under license.

My Query is Slow

Slow SELECT (or DELETE, UPDATE, MERGE) statement can have multiple reasons. Follow this checklist:

Run ANALYZE (see documentation for details).
Run the query with EXPLAIN and check if indexes are used (see documentation for details).
If required, create additional indexes and try again using ANALYZE and EXPLAIN.
If it doesn't help please report the problem.

How to Create a New Database?

By default, a new database is automatically created if it does not yet exist. See Creating New Databases.

How to Connect to a Database?

The database driver is org.h2.Driver, and the database URL starts with jdbc:h2:. To connect to a database using JDBC, use the following code:

Class.forName("org.h2.Driver");
Connection conn = DriverManager.getConnection("jdbc:h2:~/test", "sa", "");

Where are the Database Files Stored?

When using database URLs like jdbc:h2:~/test, the database is stored in the user directory. For Windows, this is usually C:\Documents and Settings\<userName>. If the base directory is not set (as in jdbc:h2:test), the database files are stored in the directory where the application is started (the current working directory). When using the H2 Console application from the start menu, this is <Installation Directory>/bin. The base directory can be set in the database URL. A fixed or relative path can be used. When using the URL jdbc:h2:file:data/sample, the database is stored in the directory data (relative to the current working directory). The directory is created automatically if it does not yet exist. It is also possible to use the fully qualified directory name (and for Windows, drive name). Example: jdbc:h2:file:C:/data/test

What is the Size Limit (Maximum Size) of a Database?

See Limits and Limitations.

Is it Reliable?

That is not easy to say. It is still a quite new product. A lot of tests have been written, and the code coverage of these tests is very high. Randomized stress tests are run regularly. But there are probably still bugs that have not yet been found (as with most software). Some features are known to be dangerous, they are only supported for situations where performance is more important than reliability. Those dangerous features are:

Disabling the transaction log mechanism using SET LOG 0.
Using the transaction isolation level READ_UNCOMMITTED (LOCK_MODE 0) while at the same time using multiple connections.
Disabling database file protection using FILE_LOCK=NO in the database URL.
Disabling referential integrity using SET REFERENTIAL_INTEGRITY FALSE.

In addition to that, running out of memory should be avoided. In older versions, OutOfMemory errors while using the database could corrupt a databases.

Areas that are not fully tested:

Platforms other than Windows XP, Linux, Mac OS X, or JVMs other than Sun 1.5 or 1.6
The features AUTO_SERVER and AUTO_RECONNECT
The MVCC (multi version concurrency) mode
Cluster mode, 2-phase commit, savepoints
24/7 operation
Some operations on databases larger than 500 MB may be slower than expected
The optimizer may not always select the best plan
Fulltext search
Operations on LOBs over 2 GB

Areas considered experimental are:

The PostgreSQL server
Multi-threading within the engine using SET MULTI_THREADED=1
Compatibility modes for other databases (only some features are implemented)

Some users have reported that after a power failure, the database can sometimes not be opened because the index file is corrupt. In that case, the index file can be deleted (it is automatically re-created). To avoid this, append ;LOG=2 to the database URL. See also: SET LOG. This problem will be solved using the new 'page store' mechanism (currently beta).

Column Names are Incorrect?

For the query SELECT ID AS X FROM TEST the method ResultSetMetaData.getColumnName() returns ID, I expect it to return X. What's wrong?

This is not a bug. According the the JDBC specification, the method ResultSetMetaData.getColumnName() should return the name of the column and not the alias name. If you need the alias name, use ResultSetMetaData.getColumnLabel(). Other database don't work like this (they don't follow the JDBC specification). If you need compatibility with those databases, use the Compatibility Mode, or set the system property h2.aliasColumnName.

Why is Opening my Database Slow?

If it takes a long time to open a database, in most cases it was not closed the last time. This is specially a problem for larger databases. To close a database, close all connections to it before the application ends, or execute the command SHUTDOWN. The database is also closed when the virtual machine exits normally by using a shutdown hook. However killing a Java process or calling Runtime.halt will prevent this. The reason why opening is slow in this situations is that indexes are re-created. If you can not guarantee the database is closed, consider using SET LOG 2.

To find out what the problem is, open the database in embedded mode using the H2 Console. This will print progress information. If you have many lines with 'Creating index' it is an indication that the database was not closed the last time.

Other possible reasons are: the database is very big (many GB), or contains linked tables that are slow to open.

Is the GCJ Version Stable? Faster?

The GCJ version is not as stable as the Java version. When running the regression test with the GCJ version, sometimes the application just stops at what seems to be a random point without error message. Currently, the GCJ version is also slower than when using the Sun VM. However, the startup of the GCJ version is faster than when using a VM.

How to Translate this Project?

For more information, see Build/Translating.