Fork me on GitHub

RDB DocumentStore

The RDBDocumentStore is one of the backend implementations for the DocumentNodeStore. It uses relational databases to persist nodes as documents, mainly emulating the native capabilities of MongoDocumentStore.

Note that the API docs for RDBDocumentStore contain more implementation details.

Supported Databases

The code was written to be as database-agnostic as possible. That said, there are vendor-specific code paths. Adding support for a new database type however should be relatively straighforward. Most of the database-specific code resides in the RDBDocumentStoreDB class.

The following databases are supported in the sense that they are recognized and have been tested with:

For testing purposes:

  • Apache Derby
  • H2DB

For production use:

  • IBM DB2 (LUW)
  • Microsoft SQL Server
  • MySQL (MariaDB)
  • Oracle
  • PostgreSQL

For supported databases, RDBDocumentStoreDB has knowledge about supported versions (and likewise supported JDBC drivers). Watch out for log messages during system startup which might warn about outdated versions (the system will attempt to start anyway):

12:20:20.864 ERROR [main] RDBDocumentStore.java:1014        Unsupported Apache Derby version: 9.14, expected at least 10.11

Initialization

The recommended method to initialize a DocumentNodeStore with an RDBDocumentStore is using an OSGi container and configure the DocumentNodeStoreService. See corresponding Repository OSGi Configuration.

This will also require deploying the Sling DataSource provider and furthermore the associated JDBC driver as OSGi bundle. The details of the latter vary by database:

  1. If the JDBC driver already is an OSGI bundle, it can be deployed as is. This is the case for Apache Derby, H2DB, IBM DB2, Microsoft SQL Server, MySQL, and PostgreSQL. Some of these drivers also implement the OSGi Data Service Specification for JDBC , in which case org.osgi.service.jdbc-1.0.0.jar needs to be deployed as well (this is the case for IBM DB2, Microsoft SQL Server, and PostgreSQL).
  2. Otherwise (e.g., Oracle), an OSGi wrapper needs to be built. See below.

Alternatively an RDB based DocumentNodeStore can be created with the help of a RDBDocumentNodeStoreBuilder.

DataSource dataSource = RDBDataSourceFactory.forJdbcUrl(jdbcurl, user, pw);
DocumentNodeStore store = RDBDocumentNodeStoreBuilder().newRDBDocumentNodeStoreBuilder()
    .setRDBConnection(dataSource).build();
// do something with the store
NodeState root = store.getRoot();

// dispose it when done
store.dispose();

Example: Creating OSGi Bundle for Oracle JDBC driver

  1. Make sure to have a local copy of the JDBC driver, for instance ojdbc8-12.2.0.1.jar.
  2. Get BND command line tool from https://repo1.maven.org/maven2/biz/aQute/bnd/biz.aQute.bnd/4.3.1/biz.aQute.bnd-4.3.1.jar
  3. Create BND ora.bnd below:
 -classpath: ojdbc8-12.2.0.1.jar
 Bundle-SymbolicName: com.oracle.jdbc.ojdbc8
 ver: 12.2.0.1
 -output: ${bsn}-${ver}.jar
 Bundle-Version: ${ver}
 Include-Resource: @ojdbc8-12.2.0.1.jar
 Import-Package: *;resolution:=optional

Then run java -jar biz.aQute.bnd-4.3.1.jar ora.bnd; this should create the OSGi bundle com.oracle.jdbc.ojdbc8-12.2.0.1.jar.

Database Creation

RDBDocumentStore relies on JDBC, and thus, in general, can not create database instances (that said, certain DBs such as Apache Derby or H2DB can create the database automatically when it's not there yet - consult the DB documentation in general and the JDBC URL syntax specifically).

So in general, the administrator will have to take care of creating the database. There are only a few requirements for the database, but these are critical for the correct operation:

  • character fields must be able to store any Unicode code point - UTF-8 encoding is recommended
  • the collation for character fields needs to sort by Unicode code points
  • BLOBs need to support sizes of ~16MB

The subsections below give examples that have been found to work during the development of RDBDocumentStore.

DB2

Creating a database called OAK:

create database oak USING CODESET UTF-8 TERRITORY DEFAULT COLLATE USING IDENTITY;

To verify, check the INFO level log message written by RDBDocumentStore upon startup. For example:

14:47:20.332 INFO  [main] RDBDocumentStore.java:1065        RDBDocumentStore (SNAPSHOT) instantiated for database DB2/NT64 SQL11014 (11.1), using driver: IBM Data Server Driver for JDBC and SQLJ 4.19.77 (4.19), connecting to: jdbc:db2://localhost:50276/OAK, properties: {DB2ADMIN.CODEPAGE=1208, DB2ADMIN.COLLATIONSCHEMA=SYSIBM, DB2ADMIN.COLLATIONNAME=IDENTITY}, transaction isolation level: TRANSACTION_READ_COMMITTED (2), DB2ADMIN.NODES: ID VARCHAR(512), MODIFIED BIGINT, HASBINARY SMALLINT, DELETEDONCE SMALLINT, MODCOUNT BIGINT, CMODCOUNT BIGINT, DSIZE BIGINT, VERSION SMALLINT, SDTYPE SMALLINT, SDMAXREVTIME BIGINT, DATA VARCHAR(16384), BDATA BLOB(1073741824) /* {BIGINT=-5, BLOB=2004, SMALLINT=5, VARCHAR=12} */ /* index DB2ADMIN.NODES_MOD on DB2ADMIN.NODES (MODIFIED ASC) other (#0, p0), unique index DB2ADMIN.NODES_PK on DB2ADMIN.NODES (ID ASC) clustered (#0, p0), index DB2ADMIN.NODES_SDM on DB2ADMIN.NODES (SDMAXREVTIME ASC) other (#0, p0), index DB2ADMIN.NODES_SDT on DB2ADMIN.NODES (SDTYPE ASC) other (#0, p0), index DB2ADMIN.NODES_VSN on DB2ADMIN.NODES (VERSION ASC) other (#0, p0) */

MySQL

Creating a database called OAK:

create database oak DEFAULT CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;

Also make sure to configure the max_allowed_packet parameter for the server (mysqld) to a value greater than 4M (such as 8388608).

To verify, check the INFO level log message written by RDBDocumentStore upon startup. For example:

13:40:46.637 INFO  [main] RDBDocumentStore.java:1065        RDBDocumentStore (SNAPSHOT) instantiated for database MySQL 8.0.15 (8.0), using driver: MySQL Connector/J mysql-connector-java-8.0.15 (Revision: 79a4336f140499bd22dd07f02b708e163844e3d5) (8.0), connecting to: jdbc:mysql://localhost:3306/oak?serverTimezone=UTC, properties: {character_set_database=utf8mb4, character_set_client=utf8mb4, character_set_connection=utf8mb4, character_set_results=, max_allowed_packet=8388608, collation_database=utf8mb4_unicode_ci, character_set_system=utf8, collation_server=utf8mb4_0900_ai_ci, collation=utf8mb4_unicode_ci, character_set_filesystem=binary, character_set_server=utf8mb4, collation_connection=utf8mb4_0900_ai_ci}, transaction isolation level: TRANSACTION_REPEATABLE_READ (4), .nodes: ID VARBINARY(512), MODIFIED BIGINT(20), HASBINARY SMALLINT(6), DELETEDONCE SMALLINT(6), MODCOUNT BIGINT(20), CMODCOUNT BIGINT(20), DSIZE BIGINT(20), VERSION SMALLINT(6), SDTYPE SMALLINT(6), SDMAXREVTIME BIGINT(20), DATA VARCHAR(16000), BDATA LONGBLOB(2147483647) /* {BIGINT=-5, LONGBLOB=-4, SMALLINT=5, VARBINARY=-3, VARCHAR=12} */ /* unique index oak.PRIMARY on nodes (ID ASC) other (#0, p0), index oak.NODES_MOD on nodes (MODIFIED ASC) other (#0, p0), index oak.NODES_SDM on nodes (SDMAXREVTIME ASC) other (#0, p0), index oak.NODES_SDT on nodes (SDTYPE ASC) other (#0, p0), index oak.NODES_VSN on nodes (VERSION ASC) other (#0, p0) */

Oracle

Creating a database called OAK:

…is different compared to other databases. See https://docs.oracle.com/cd/B28359_01/server.111/b28310/create003.htm for more information. Defaults should work.

To verify, check the INFO level log message written by RDBDocumentStore upon startup. For example:

13:26:37.073 INFO  [main] RDBDocumentStore.java:1067        RDBDocumentStore (SNAPSHOT) instantiated for database Oracle Oracle Database 12c Enterprise Edition Release 12.2.0.1.0 - 64bit Production (12.2), using driver: Oracle JDBC driver 12.2.0.1.0 (12.2), connecting to: jdbc:oracle:thin:@localhost:1521:orcl, properties: {NLS_CHARACTERSET=AL32UTF8, NLS_COMP=BINARY}, transaction isolation level: TRANSACTION_READ_COMMITTED (2), .: ID VARCHAR2(512), MODIFIED NUMBER, HASBINARY NUMBER, DELETEDONCE NUMBER, MODCOUNT NUMBER, CMODCOUNT NUMBER, DSIZE NUMBER, VERSION NUMBER, SDTYPE NUMBER, SDMAXREVTIME NUMBER, DATA VARCHAR2(4000), BDATA BLOB(-1) /* {BLOB=2004, NUMBER=2, VARCHAR2=12} */ /* index NODES_MOD on SYSTEM.NODES (MODIFIED) clustered (#0, p0), index NODES_SDM on SYSTEM.NODES (SDMAXREVTIME) clustered (#0, p0), index NODES_SDT on SYSTEM.NODES (SDTYPE) clustered (#0, p0), index NODES_VSN on SYSTEM.NODES (VERSION) clustered (#0, p0), unique index SYS_C008093 on SYSTEM.NODES (ID) clustered (#0, p0) */

PostgreSQL

Creating a database called OAK:

CREATE DATABASE "oak" TEMPLATE = template0 ENCODING = 'UTF8' LC_COLLATE = 'C' LC_CTYPE = 'C';

To verify, check the INFO level log message written by RDBDocumentStore upon startup. For example:

16:26:28.172 INFO  [main] RDBDocumentStore.java:1065        RDBDocumentStore (SNAPSHOT) instantiated for database PostgreSQL 10.6 (10.6), using driver: PostgreSQL JDBC Driver 42.2.5 (42.2), connecting to: jdbc:postgresql:oak, properties: {datcollate=C, pg_encoding_to_char(encoding)=UTF8}, transaction isolation level: TRANSACTION_READ_COMMITTED (2), .nodes: id varchar(512), modified int8, hasbinary int2, deletedonce int2, modcount int8, cmodcount int8, dsize int8, version int2, sdtype int2, sdmaxrevtime int8, data varchar(16384), bdata bytea(2147483647) /* {bytea=-2, int2=5, int8=-5, varchar=12} */ /* index nodes_mod on public.nodes (modified ASC) other (#0, p1), unique index nodes_pkey on public.nodes (id ASC) other (#0, p1), index nodes_sdm on public.nodes (sdmaxrevtime ASC) other (#0, p1), index nodes_sdt on public.nodes (sdtype ASC) other (#0, p1), index nodes_vsn on public.nodes (version ASC) other (#0, p1) */

SQL Server

Creating a database called OAK:

create database OAK;

To verify, check the INFO level log message written by RDBDocumentStore upon startup. For example:

16:59:12.726 INFO  [main] RDBDocumentStore.java:1067        RDBDocumentStore (SNAPSHOT) instantiated for database Microsoft SQL Server 13.00.5081 (13.0), using driver: Microsoft JDBC Driver 7.2 for SQL Server 7.2.1.0 (7.2), connecting to: jdbc:sqlserver://localhost:1433;useBulkCopyForBatchInsert=false;cancelQueryTimeout=-1;sslProtocol=TLS;jaasConfigurationName=SQLJDBCDriver;statementPoolingCacheSize=0;serverPreparedStatementDiscardThreshold=10;enablePrepareOnFirstPreparedStatementCall=false;fips=false;socketTimeout=0;authentication=NotSpecified;authenticationScheme=nativeAuthentication;xopenStates=false;sendTimeAsDatetime=true;trustStoreType=JKS;trustServerCertificate=false;TransparentNetworkIPResolution=true;serverNameAsACE=false;sendStringParametersAsUnicode=true;selectMethod=direct;responseBuffering=adaptive;queryTimeout=-1;packetSize=8000;multiSubnetFailover=false;loginTimeout=15;lockTimeout=-1;lastUpdateCount=true;encrypt=false;disableStatementPooling=true;databaseName=OAK;columnEncryptionSetting=Disabled;applicationName=Microsoft JDBC Driver for SQL Server;applicationIntent=readwrite;, properties: {collation_name=Latin1_General_CI_AS}, transaction isolation level: TRANSACTION_READ_COMMITTED (2), .: ID varbinary(512), MODIFIED bigint, HASBINARY smallint, DELETEDONCE smallint, MODCOUNT bigint, CMODCOUNT bigint, DSIZE bigint, VERSION smallint, SDTYPE smallint, SDMAXREVTIME bigint, DATA nvarchar(4000), BDATA varbinary(2147483647) /* {bigint=-5, nvarchar=-9, smallint=5, varbinary=-3} */ /* index NODES.NODES_MOD on dbo.NODES (MODIFIED ASC) other (#0, p0), unique index NODES.NODES_PK on dbo.NODES (ID ASC) clustered (#0, p0), index NODES.NODES_SDM on dbo.NODES (SDMAXREVTIME ASC) other (#0, p0), index NODES.NODES_SDT on dbo.NODES (SDTYPE ASC) other (#0, p0), index NODES.NODES_VSN on dbo.NODES (VERSION ASC) other (#0, p0) */

Table Creation

The implementation will try to create all tables and indices when they are not present yet. Of course this requires that the configured database user actually has permission to do so. Example from system log:

12:20:22.705 INFO  [main] RDBDocumentStore.java:1063        RDBDocumentStore (SNAPSHOT) instantiated for database Apache Derby 10.14.2.0 - (1828579) (10.14), using driver: Apache Derby Embedded JDBC Driver 10.14.2.0 - (1828579) (10.14), connecting to: jdbc:derby:./target/derby-ds-test, transaction isolation level: TRANSACTION_READ_COMMITTED (2), SA.NODES: ID VARCHAR(512), MODIFIED BIGINT, HASBINARY SMALLINT, DELETEDONCE SMALLINT, MODCOUNT BIGINT, CMODCOUNT BIGINT, DSIZE BIGINT, VERSION SMALLINT, SDTYPE SMALLINT, SDMAXREVTIME BIGINT, DATA VARCHAR(16384), BDATA BLOB(1073741824) /* {BIGINT=-5, BLOB=2004, SMALLINT=5, VARCHAR=12} */ /* index NODES_MOD on SA.NODES (MODIFIED ASC) other (#0, p0), index NODES_SDM on SA.NODES (SDMAXREVTIME ASC) other (#0, p0), index NODES_SDT on SA.NODES (SDTYPE ASC) other (#0, p0), index NODES_VSN on SA.NODES (VERSION ASC) other (#0, p0), unique index SQL190131122022490 on SA.NODES (ID ASC) other (#0, p0) */
12:20:22.705 INFO  [main] RDBDocumentStore.java:1070        Tables created upon startup: [CLUSTERNODES, NODES, SETTINGS, JOURNAL]

If it does not, the system will not start up and provide diagnostics in the log file.

Administrators who want to create tables upfront can do so. The DDL statements for the supported databases can be dumped using RDBHelper or, more recently, using oak-run rdbddldump (see below).

Upgrade from earlier versions

As of Oak 1.8, the database layout has been slightly extended (see API docs for RDBDocumentStore for details).

Upon startup on an “old” database instance, RDBDocumentStore will try to upgrade the tables. Example (for NODES):

12:05:54.146 INFO  [main] RDBDocumentStore.java:1369        Upgraded NODES to DB level 1 using 'alter table NODES add VERSION smallint'

12:05:54.166 INFO  [main] RDBDocumentStore.java:1369        Upgraded NODES to DB level 2 using 'alter table NODES add SDMAXREVTIME bigint'
12:05:54.167 INFO  [main] RDBDocumentStore.java:1369        Upgraded NODES to DB level 2 using 'create index NODES_VSN on NODES (VERSION)'
12:05:54.167 INFO  [main] RDBDocumentStore.java:1369        Upgraded NODES to DB level 2 using 'create index NODES_SDT on NODES (SDTYPE)'
12:05:54.167 INFO  [main] RDBDocumentStore.java:1369        Upgraded NODES to DB level 2 using 'create index NODES_SDM on NODES (SDMAXREVTIME)'

If this fails, it will continue using the “old” layout, and log diagnostics about the failed upgrade:

12:05:56.746 INFO  [main] RDBDocumentStore.java:1379        Attempted to upgrade NODES to DB level 1 using 'alter table NODES add VERSION smallint', but failed with SQLException 'table alter statement rejected: alter table NODES add VERSION smallint' (code: 17/state: ABCDE) - will continue without.

12:05:56.955 INFO  [main] RDBDocumentStore.java:1379        Attempted to upgrade NODES to DB level 2 using 'alter table NODES add SDTYPE smallint', but failed with SQLException 'table alter statement rejected: alter table NODES add SDTYPE smallint' (code: 17/state: ABCDE) - will continue without.
12:05:56.955 INFO  [main] RDBDocumentStore.java:1379        Attempted to upgrade NODES to DB level 2 using 'alter table NODES add SDMAXREVTIME bigint', but failed with SQLException 'table alter statement rejected: alter table NODES add SDMAXREVTIME bigint' (code: 17/state: ABCDE) - will continue without.
12:05:56.964 INFO  [main] RDBDocumentStore.java:1379        Attempted to upgrade NODES to DB level 2 using 'create index NODES_SDT on NODES (SDTYPE)', but failed with SQLException ''SDTYPE' is not a column in table or VTI 'NODES'.' (code: 20000/state: 42X14) - will continue without.
12:05:56.964 INFO  [main] RDBDocumentStore.java:1379        Attempted to upgrade NODES to DB level 2 using 'create index NODES_SDM on NODES (SDMAXREVTIME)', but failed with SQLException ''SDMAXREVTIME' is not a column in table or VTI 'NODES'.' (code: 20000/state: 42X14) - will continue without.

The upgrade can then be done at a later point of time by executing the required DDL statements.

oak-run rdbddldump

@since Oak 1.8.12 @since Oak 1.10.1 @since Oak 1.12

The rdbddldump prints out the DDL statements that Oak would use to create or update a database. It can be used to create the tables upfront, or to obtain the DDL statements needed to upgrade to a newer schema version.

By default, it will print out the DDL statements for all supported databases, with a target of the latest schema version.

The --db switch can be used to specify the database type (note that precise spelling is needed, otherwise the code will fall back to a generic database type).

The --initial switch selects the initial database schema (and defaults to the most recent one).

The --upgrade switch selects the target database schema (and defaults to the most recent one).

Selecting a higher “upgrade” version then the “initial” version causes the tool to create separate DDL statements for the initial table schema (which may already be there), and then to add individual statements for the upgrade to the target schema.

For instance:

java -jar oak-run-*.jar rdbddldump --db DB2 --initial 0 --upgrade 2

will dump statements for DB2, initially creating schema version 0 tables, and then include DDL statements to upgrade to version 2 (the latter would be applicable if an installation needed to be upgraded from an Oak version older than 1.8 to 1.8 or newer).

-- DB2

  -- creating table CLUSTERNODES for schema version 0
  create table CLUSTERNODES (ID varchar(512) not null, MODIFIED bigint, HASBINARY smallint, DELETEDONCE smallint, MODCOUNT bigint, CMODCOUNT bigint, DSIZE bigint, DATA varchar(16384), BDATA blob(1073741824))
  create unique index CLUSTERNODES_pk on CLUSTERNODES ( ID ) cluster
  alter table CLUSTERNODES add constraint CLUSTERNODES_pk primary key ( ID )
  create index CLUSTERNODES_MOD on CLUSTERNODES (MODIFIED)
  -- upgrading table CLUSTERNODES to schema version 1
  alter table CLUSTERNODES add VERSION smallint
  -- upgrading table CLUSTERNODES to schema version 2
  alter table CLUSTERNODES add SDTYPE smallint
  alter table CLUSTERNODES add SDMAXREVTIME bigint
  create index CLUSTERNODES_VSN on CLUSTERNODES (VERSION)
  create index CLUSTERNODES_SDT on CLUSTERNODES (SDTYPE) exclude null keys
  create index CLUSTERNODES_SDM on CLUSTERNODES (SDMAXREVTIME) exclude null keys

  -- creating table JOURNAL for schema version 0
  create table JOURNAL (ID varchar(512) not null, MODIFIED bigint, HASBINARY smallint, DELETEDONCE smallint, MODCOUNT bigint, CMODCOUNT bigint, DSIZE bigint, DATA varchar(16384), BDATA blob(1073741824))
  create unique index JOURNAL_pk on JOURNAL ( ID ) cluster
  alter table JOURNAL add constraint JOURNAL_pk primary key ( ID )
  create index JOURNAL_MOD on JOURNAL (MODIFIED)
  -- upgrading table JOURNAL to schema version 1
  alter table JOURNAL add VERSION smallint
  -- upgrading table JOURNAL to schema version 2
  alter table JOURNAL add SDTYPE smallint
  alter table JOURNAL add SDMAXREVTIME bigint
  create index JOURNAL_VSN on JOURNAL (VERSION)
  create index JOURNAL_SDT on JOURNAL (SDTYPE) exclude null keys
  create index JOURNAL_SDM on JOURNAL (SDMAXREVTIME) exclude null keys

  -- creating table NODES for schema version 0
  create table NODES (ID varchar(512) not null, MODIFIED bigint, HASBINARY smallint, DELETEDONCE smallint, MODCOUNT bigint, CMODCOUNT bigint, DSIZE bigint, DATA varchar(16384), BDATA blob(1073741824))
  create unique index NODES_pk on NODES ( ID ) cluster
  alter table NODES add constraint NODES_pk primary key ( ID )
  create index NODES_MOD on NODES (MODIFIED)
  -- upgrading table NODES to schema version 1
  alter table NODES add VERSION smallint
  -- upgrading table NODES to schema version 2
  alter table NODES add SDTYPE smallint
  alter table NODES add SDMAXREVTIME bigint
  create index NODES_VSN on NODES (VERSION)
  create index NODES_SDT on NODES (SDTYPE) exclude null keys
  create index NODES_SDM on NODES (SDMAXREVTIME) exclude null keys

  -- creating table SETTINGS for schema version 0
  create table SETTINGS (ID varchar(512) not null, MODIFIED bigint, HASBINARY smallint, DELETEDONCE smallint, MODCOUNT bigint, CMODCOUNT bigint, DSIZE bigint, DATA varchar(16384), BDATA blob(1073741824))
  create unique index SETTINGS_pk on SETTINGS ( ID ) cluster
  alter table SETTINGS add constraint SETTINGS_pk primary key ( ID )
  create index SETTINGS_MOD on SETTINGS (MODIFIED)
  -- upgrading table SETTINGS to schema version 1
  alter table SETTINGS add VERSION smallint
  -- upgrading table SETTINGS to schema version 2
  alter table SETTINGS add SDTYPE smallint
  alter table SETTINGS add SDMAXREVTIME bigint
  create index SETTINGS_VSN on SETTINGS (VERSION)
  create index SETTINGS_SDT on SETTINGS (SDTYPE) exclude null keys
  create index SETTINGS_SDM on SETTINGS (SDMAXREVTIME) exclude null keys

   -- creating blob store tables
  create table DATASTORE_META (ID varchar(64) not null primary key, LVL int, LASTMOD bigint)
  create table DATASTORE_DATA (ID varchar(64) not null primary key, DATA blob(2097152))

Using oak-run

The oak-run JAR file does not include the JDBC driver needed to access the database. Thus, a small amount of classpath surgery is needed.

Assuming the following two JAR files are in the current directory:

  • oak-run-1.14.0.jar
  • db2-4.19.77.jar

…the invocation would be:

$ java -cp "oak-run-1.14.0.jar:db2-4.19.77.jar" org.apache.jackrabbit.oak.run.Main

(where the path separator under Windows would be “;”).

In general, all commands applicable to a MongoDocumentStore should be available for RDBDocumentStore as well. Simply substitute the “mongdb:…” identifier by the JDBC “URL”, and also specify DB credentials using --rdbjdbcuser and --rdbjdbcpasswd.

Like that:

$ java -cp "oak-run-1.14.0.jar:db2-4.19.77.jar" org.apache.jackrabbit.oak.run.Main clusternodes jdbc:db2://localhost:50276/OAK --rdbjdbcuser user --rdbjdbcpasswd passwd --verbose

Apache Jackrabbit Oak 1.14.0
Id    State          Started LeaseEnd Left RecoveryBy      LastRootRev    OakVersion
 1 INACTIVE 20190125T110237Z        -    -          - r16884ad047c-0-1 1.12-SNAPSHOT

Note that in Oak versions prior to June 2019, oak-run also does not contain the artefacts tomcat-jdbc and tomcat-juli, which thus need to be added to the classpath as well (see OAK-8341 for details).

Reading Log Files

There are certain log messages to look out for when investigating problems, and also some that may cause unneeded confusion.

See below.

Startup Message

Such as:

16:26:28.172 INFO  [main] RDBDocumentStore.java:1065        RDBDocumentStore (SNAPSHOT) instantiated for database PostgreSQL 10.6 (10.6), using driver: PostgreSQL JDBC Driver 42.2.5 (42.2), connecting to: jdbc:postgresql:oak, properties: {datcollate=C, pg_encoding_to_char(encoding)=UTF8}, transaction isolation level: TRANSACTION_READ_COMMITTED (2), .nodes: id varchar(512), modified int8, hasbinary int2, deletedonce int2, modcount int8, cmodcount int8, dsize int8, version int2, sdtype int2, sdmaxrevtime int8, data varchar(16384), bdata bytea(2147483647) /* {bytea=-2, int2=5, int8=-5, varchar=12} */ /* index nodes_mod on public.nodes (modified ASC) other (#0, p1), unique index nodes_pkey on public.nodes (id ASC) other (#0, p1), index nodes_sdm on public.nodes (sdmaxrevtime ASC) other (#0, p1), index nodes_sdt on public.nodes (sdtype ASC) other (#0, p1), index nodes_vsn on public.nodes (version ASC) other (#0, p1) */

The information dumped here is essential for diagnosing issues and should be included in all bug reports. It includes:

  • Oak's version number
  • Database type and version
  • JDBC driver and version
  • JDBC “URL”
  • certain DB-specific properties
  • JDBC transaction isolation level
  • schema for “NODES” table (other tables are not reported; they ought to be the same)
  • information on database indices

“Long Running Queries”

RDBDocumentStore will log an INFO message when a query takes longer than 10s, which frequently indicates a configuration problem.

INFO  ... org.apache.jackrabbit.oak.plugins.document.rdb.RDBDocumentStoreJDBC - Long running query on NODES with 100 hits (limited to 100), elapsed time 11361ms (configured QUERYTIMELIMIT 10000), params minid 'null' maxid 'null' excludeKeyPatterns [] conditions [_modified >= 1554909530] limit 100. Result range: '0:/'...'12:/...'. Read 26126 chars from DATA and 0 bytes from BDATA. Check calling method.
java.lang.Exception: call stack

The call stack is dumped to identify the piece of code that executed the query. It is important to understand that this is not an error, it is just logged for diagnostic purposes.

Tomcat JDBC Pool Interceptor Messages

When using the Tomcat JDBC connection pool, by default “slow” and “failed” queries will be logged on WARN level. The latter category is a bit problematic, as what Tomcat thinks is a failure might be completely normal and expected behavior. For instance:

WARN [...] org.apache.tomcat.jdbc.pool.interceptor.SlowQueryReport Failed Query Report SQL=update NODES set MODIFIED = case when ? > MODIFIED then ? else MODIFIED end, HASBINARY = ?, DELETEDONCE = ?, MODCOUNT = ?, CMODCOUNT = ?, DSIZE = DSIZE + ?, VERSION = 2, DATA = CASE WHEN LEN(DATA) < ? THEN (DATA + CAST(? AS nvarchar(4000))) ELSE (DATA + CAST(DATA AS nvarchar(max))) END where ID = ? and MODCOUNT = ?; time=1 ms;

This just indicates that an update operation that was made conditional did not happen because the condition was not met.

Another example are insert operations, which are done “optimistically”, and will automatically be retried as updates upon failure.