openHAB

# JDBC Persistence

This service writes and reads item states to and from a number of relational database systems that support Java Database Connectivity (JDBC) (opens new window). This service allows you to persist state updates using one of several different underlying database services. It is designed for a maximum of scalability, to store very large amounts of data and still over the years not lose its speed.

The generic design makes it relatively easy for developers to integrate other databases that have JDBC drivers. The following databases are currently supported and tested:

Database Tested Driver / Version
Apache Derby (opens new window) derby-10.14.2.0.jar (opens new window)
H2 (opens new window) h2-2.2.224.jar (opens new window)
HSQLDB (opens new window) hsqldb-2.3.3.jar (opens new window)
MariaDB (opens new window) mariadb-java-client-3.0.8.jar (opens new window)
MySQL (opens new window) mysql-connector-j-8.2.0.jar (opens new window)
PostgreSQL (opens new window) postgresql-42.4.4.jar (opens new window)
SQLite (opens new window) sqlite-jdbc-3.42.0.0.jar (opens new window)
TimescaleDB (opens new window) postgresql-42.4.4.jar (opens new window)

# Table of Contents

# Configuration

This service can be configured in the file services/jdbc.cfg.

Property Default Required Description
url Yes JDBC URL to establish a connection to your database. Examples:

jdbc:derby:./testDerby;create=true
jdbc:h2:./testH2
jdbc:hsqldb:./testHsqlDb
jdbc:mariadb://192.168.0.1:3306/testMariadb
jdbc:mysql://192.168.0.1:3306/testMysql?serverTimezone=UTC
jdbc:postgresql://192.168.0.1:5432/testPostgresql
jdbc:timescaledb://192.168.0.1:5432/testPostgresql
jdbc:sqlite:./testSqlite.db.

If no database is available it will be created; for example the url jdbc:h2:./testH2 creates a new H2 database in openHAB folder. Example to create your own MySQL database directly:

CREATE DATABASE 'yourDB' CHARACTER SET utf8 COLLATE utf8_general_ci;
user if needed database user name
password if needed database user password
errReconnectThreshold 0 No when the service is deactivated (0 means ignore)
sqltype.CALL VARCHAR(200) No All sqlType options allow you to change the SQL data type used to store values for different openHAB item states. See the following links for further information: mybatis (opens new window) H2 (opens new window) PostgresSQL (opens new window)
sqltype.COLOR VARCHAR(70) No see above
sqltype.CONTACT VARCHAR(6) No see above
sqltype.DATETIME DATETIME No see above
sqltype.DIMMER TINYINT No see above
sqltype.IMAGE VARCHAR(65500) No see above
sqltype.LOCATION VARCHAR(50) No see above
sqltype.NUMBER DOUBLE No see above
sqltype.PLAYER VARCHAR(20) No see above
sqltype.ROLLERSHUTTER TINYINT No see above
sqltype.STRING VARCHAR(65500) No see above
sqltype.SWITCH VARCHAR(6) No see above
sqltype.tablePrimaryKey TIMESTAMP No type of time column for newly created item tables
sqltype.tablePrimaryValue NOW() No value of time column for newly inserted rows
numberDecimalcount 3 No for Itemtype "Number" default decimal digit count
itemsManageTable items No items manage table. For Migration from MySQL Persistence, set to Items.
tableNamePrefix item No table name prefix. For Migration from MySQL Persistence, set to Item.
tableUseRealItemNames false No table name prefix generation. When set to true, real item names are used for table names and tableNamePrefix is ignored. When set to false, the tableNamePrefix is used to generate table names with sequential numbers.
tableCaseSensitiveItemNames false No table name case. This setting is only applicable when tableUseRealItemNames is true. When set to true, item name case is preserved in table names and no prefix or suffix is added. When set to false, table names are lower cased and a numeric suffix is added. Please read this before enabling.
tableIdDigitCount 4 No when tableUseRealItemNames is false and thus table names are generated sequentially, this controls how many zero-padded digits are used in the table name. With the default of 4, the first table name will end with 0001. For migration from the MySQL persistence service, set this to 0.
rebuildTableNames false No rename existing tables using tableUseRealItemNames and tableIdDigitCount. USE WITH CARE! Deactivate after Renaming is done!
jdbc.maximumPoolSize configured per database in package org.openhab.persistence.jdbc.db.* No Some embedded databases can handle only one connection. See this link (opens new window) for more information
jdbc.minimumIdle see above No see above
enableLogTime false No timekeeping

All item- and event-related configuration is done in the file persistence/jdbc.persist.

To configure this service as the default persistence service for openHAB, add or change the line

org.openhab.core.persistence:default=jdbc

in the file services/runtime.cfg.

# Minimal Configuration

services/jdbc.cfg

url=jdbc:postgresql://192.168.0.1:5432/testPostgresql

# Case Sensitive Item Names

To avoid numbered suffixes entirely, tableUseRealItemNames and tableCaseSensitiveItemNames must both be enabled. With this configuration, tables are named exactly like their corresponding items. In order for this to work correctly, the underlying operating system, database server and configuration must support case sensitive table names. For MySQL, see MySQL: Identifier Case Sensitivity (opens new window) for more information.

Please make sure to have a dedicated schema when using this option, since otherwise table name collisions are more likely to happen.

# Migration from MySQL to JDBC Persistence Services

The JDBC Persistence service can act as a replacement for the MySQL Persistence service. Here is an example of a configuration for a MySQL database named testMysql with user test and password test:

services/jdbc.cfg

url=jdbc:mysql://192.168.0.1:3306/testMysql
user=test
password=test
itemsManageTable=Items
tableNamePrefix=Item
tableUseRealItemNames=false
tableIdDigitCount=0

Remember to install and uninstall the services you want, and rename persistence/mysql.persist to persistence/jdbc.persist.

# Technical Notes

# Database Table Schema

The table name schema can be reconfigured after creation, if needed.

The service will create a mapping table to link each item to a table, and a separate table is generated for each item. The item data tables include time and data values. The SQL data type used depends on the openHAB item type, and allows the item state to be recovered back into openHAB in the same way it was stored.

With this per-item layout, the scalability and easy maintenance of the database is ensured, even if large amounts of data must be managed. To rename existing tables, use the parameters tableNamePrefix, tableUseRealItemNames, tableIdDigitCount and tableCaseSensitiveItemNames in the configuration.

Please be aware that changing the name of itemsManageTable is not supported by the migration. If this is changed, the table must be renamed manually according to new configured name.

# Number Precision

Default openHAB number items are persisted with SQL datatype double. Internally openHAB uses BigDecimal. If better numerical precision is needed, for example set sqltype.NUMBER = DECIMAL(max digits, max decimals), then on the Java side, the service works with BigDecimal without type conversion. If more come decimals as max decimals provides, this persisted value is rounded mathematically correctly. The SQL types DECIMAL or NUMERIC are precise, but to work with DOUBLE is faster.

# Rounding results

The results of database queries of number items are rounded to three decimal places by default. With numberDecimalcount decimals can be changed. Especially if sql types DECIMAL or NUMERIC are used for sqltype.NUMBER, rounding can be disabled by setting numberDecimalcount=-1.

# Maintenance

Some maintenance tools are provided as console commands.

# List Tables

Tables and corresponding items can be listed with the command jdbc tables list. Per default only tables with some kind of problem are listed. To list all tables, use the command jdbc tables list all.

The list contains table name, item name, row count and status, which can be one of:

  • Valid: Table is consistent.
  • Item missing: Table has no corresponding item.
  • Table missing: Referenced table does not exist.
  • Item and table missing: Referenced table does not exist nor has corresponding item.
  • Orphan table: Mapping for table does not exist in index.

# Clean Inconsistent Items

Some issues can be fixed automatically using the command jdbc tables clean (all items having issues) or jdbc tables clean <itemName> (single item). This cleanup operation will remove items from the index (table Items) if the referenced table does not exist.

If the item does not exist, the table will be physically deleted, but only if it's empty. This precaution is taken because items may have existed previously, and the data might still be valuable. For example, an item for a lost or repurposed sensor could have been deleted from the system while preserving persisted data. To skip this check for a single item, use jdbc tables clean <itemName> force with care.

Prior to performing a jdbc tables clean operation, it's recommended to review the result of jdbc tables list.

Fixing integrity issues can be useful before performing a migration to another naming scheme. For example, when migrating to tableCaseSensitiveItemNames, an index will no longer exist after the migration:

Before migration:

Table Row count Item Status
ActualItem 0 Orphan table
TableNotBelonging 0 Orphan table
item0077 0 MyItem Table missing

After migration:

Table Row count Item Status
ActualItem 0 ActualItem Valid
TableNotBelonging 0 TableNotBelonging Item missing

This happened:

  • ActualItem was missing in the index and became valid because it was left untouched, not being a part of the migration. After the migration, it happened to match the name of an existing item, thus it became valid.
  • TableNotBelonging was also not part of the migration, but since now assumed to match an item, status changed since no item with that name exists.
  • item0077, being the only correct table name according to previous naming scheme, disappeared from the list since it didn't have a corresponding table, and is now no longer part of any index.

In other words, extracting this information from the index before removing it, can be beneficial in order to understand the issues and possible causes.

# Reload Index/Schema

Manual changes in the index table, Items, will not be picked up automatically for performance reasons. The same is true when manually adding new item tables or deleting existing ones. After making such changes, the command jdbc reload can be used to reload the index.

# Check/fix Schema

Use the command jdbc schema check to perform an integrity check of the schema.

Identified issues can be fixed automatically using the command jdbc schema fix (all items having issues) or jdbc schema fix <itemName> (single item).

Issues than can be identified and possibly fixed:

  • Wrong column name case (time and name).
  • Wrong column type. Before fixing this, make sure that time-zone is correctly configured.
  • Unexpected column (identify only).

# For Developers

  • Clearly separated source files for the database-specific part of openHAB logic.
  • Code duplication by similar services is prevented.
  • Integrating a new SQL and JDBC enabled database is fairly simple.

# Performance Tests

Not necessarily representative of the performance you may experience.

DATABASE FIRST RUN AVERAGE FASTEST SIZE AFTER COMMENT
Derby 7.829 6.892 5.381 5.36 MB local embedded
H2 1.797 2.080 1.580 0.96 MB local embedded
hsqldb 3.474 2.104 1.310 1.23 MB local embedded
mysql 11.873 11.524 10.971 - ext. Server VM
postgresql 8.147 7.072 6.895 - ext. Server VM
sqlite 2.406 1.249 1.137 0.28 MB local embedded
  • Each test ran about 20 Times every 30 seconds.
  • openHAB 1.x has ready started for about a Minute.
  • the data in seconds for the evaluation are from the console output.

Used a script like this:

var count = 0;
rule "DB STRESS TEST"
when
	Time cron "30 * * * * ?"
then
	if( count = 24) count = 0
	count = count+1
	if( count > 3 && count < 23){
		for( var i=500; i>1; i=i-1){
			postUpdate( NUMBERITEM, i)
			SWITCHITEM.previousState().state
			postUpdate( DIMMERITEM, OFF)
			NUMBERITEM.changedSince( now().minusMinutes(1))
			postUpdate( DIMMERITEM, ON)
		}
	}
end
Caught a mistake or want to contribute to the documentation? Edit this page on GitHub (opens new window)

← MapDB →