Slony-I 2.2.10 Documentation | ||||
---|---|---|---|---|
Prev | Fast Backward | Chapter 6. Additional Utilities | Fast Forward | Next |
6.7. Clustertest Test Framework
6.7.1. Introduction and Overview
The clustertest framework is implemented in Java, where tests are implemented in the interpreted JavaScript language. The use of Java made it much easier to implement tests involving concurrent activities, both in terms of inducing test load, and in, concurrently changing configuration of the replication cluster.
It consists of two physical portions:
A framework, implemented in Java
This software is available at clustertest-framework @ GitHub .
This framework makes use of libraries from several other open source projects:
js.jar
This is for org.mozilla.javascript, the Mozilla JavaScript interpreter
junit-4.8.1.jar
JUnit, a unit test framework.
log4j-1.2.15.jar
Log4J is a popular Java-based framework for generating event logs.
postgresql-8.4-701.jdbc3.jar
This is the PostgreSQL JDBC driver.
To build the framework, it is necessary to have a Java compiler and the build tool, Ant, installed. To build all the .jar files used by the framework, one will run the command, with output similar to the following:
% ant jar Buildfile: /var/lib/postgresql/PostgreSQL/clustertest-framework/build.xml compile-common: [mkdir] Created dir: /var/lib/postgresql/PostgreSQL/clustertest-framework/build/classes [javac] /var/lib/postgresql/PostgreSQL/clustertest-framework/build.xml:23: warning: 'includeantruntime' was not set, defaulting to build.sysclasspath=last; set to false for repeatable builds compile-testcoordinator: [javac] /var/lib/postgresql/PostgreSQL/clustertest-framework/build.xml:44: warning: 'includeantruntime' was not set, defaulting to build.sysclasspath=last; set to false for repeatable builds [javac] Compiling 25 source files to /var/lib/postgresql/PostgreSQL/clustertest-framework/build/classes [javac] Note: /var/lib/postgresql/PostgreSQL/clustertest-framework/src/info/slony/clustertest/testcoordinator/script/ClientScript.java uses or overrides a deprecated API. [javac] Note: Recompile with -Xlint:deprecation for details. [copy] Copying 1 file to /var/lib/postgresql/PostgreSQL/clustertest-framework/build/classes/info/slony/clustertest/testcoordinator jar-common: [mkdir] Created dir: /var/lib/postgresql/PostgreSQL/clustertest-framework/build/jar [jar] Building MANIFEST-only jar: /var/lib/postgresql/PostgreSQL/clustertest-framework/build/jar/clustertest-common.jar compile-client: [javac] /var/lib/postgresql/PostgreSQL/clustertest-framework/build.xml:30: warning: 'includeantruntime' was not set, defaulting to build.sysclasspath=last; set to false for repeatable builds [javac] Compiling 1 source file to /var/lib/postgresql/PostgreSQL/clustertest-framework/build/classes [copy] Copying 2 files to /var/lib/postgresql/PostgreSQL/clustertest-framework/build/classes/info/slony/clustertest/client jar-client: [jar] Building jar: /var/lib/postgresql/PostgreSQL/clustertest-framework/build/jar/clustertest-client.jar jar-coordinator: [jar] Building jar: /var/lib/postgresql/PostgreSQL/clustertest-framework/build/jar/clustertest-coordinator.jar jar: BUILD SUCCESSFUL Total time: 2 seconds
At this time, there is no "regression test" for the test framework; to validate that it works requires running tests that use it.
It includes classes supporting PostgreSQL- and Slony-I-specific functionality such as:
CreateDbScript
Creates a database
DropDbScript
Drops a database
LogShippingDaemon
Starts up Slony-I logshipping daemon
LogShippingDumpScript
Dumps and loads logshipping-based schema
PgCommand
Run a PostgreSQL shell command (such as psql, createdb, and such)
PgDumpCommand
Dump a PostgreSQL database
PsqlCommandExec
Run SQL
ShellExecScript
Run a shell script/command
SlonLauncher
Start up a slon process
SlonikScript
Run a slonik script
Tests integrated into the Slony-I software distribution, that consist of a combination of shell scripts, JavaScript, and SQL scripts.
See the directory clustertest in the Slony-I software distribution, which has two sets of tests:
6.7.2. DISORDER - DIStributed ORDER test
The DISORDER or DIStributed ORDER test is intended to provide concurrency tests involving a reasonably sophisticated schema to validate various aspects of Slony-I behavior under concurrent load.
It consists of:
A schema for an inventory management application.
Major objects include customers, inventory items, orders, order lines, and shipments.
There are foreign key relationships between the various items, as well as triggers that maintain inventory and customer balances. Some of these relationships involve ON DELETE CASCADE, and so some actions may induce large numbers of cascaded updates.
Stored procedures to induce creation of the various sorts of objects, purchases, shipments, and additions and removals of customers and products.
Some tests are intended to be run against replicas, validating that balances add up. We believe that PostgreSQL applies changes in a transactional fashion such that they will always COMMIT leaving the visible state consistent; certain of the tests look for inconsistencies.
There are JavaScript test scripts that induce all sorts of manipulations of replication clusters to validate that replication configuration changes succeed and fail as expected.
6.7.2.1. Configuring DISORDER
DISORDER test configuration may be found in the following files:
conf/disorder.properties.sample
This file contains Java style properties indicating how to connect to the various databases used by the DISORDER tests, including paths to tools such as slon and slonik
The sample file is to be copied to conf/disorder.properties, and customized to indicate your local configuration. By using a .sample file, a developer may run tests within a Git tree, and not need to worry about their customizations interfering with the "canonical" sample configuration provided.
conf/java.conf.sample
This is a shell script containing a path indicating where the clustertest Java code (e.g. - the clustertest-coordinator.jar file) may be found. This is also used, indirectly to determine where additional Java .jar files such as the JDBC driver are located.
As with the disorder properties, above, this needs to be copied to conf/java.conf, and customized to indicate one's own local configuration.
conf/log4j.properties
See documentation for Log4J for more details as to how this is configured; the defaults provided likely do not need to be altered.
6.7.3. Regression Tests
These tests represent a re-coding of the tests previously implemented as shell scripts using the clustertest framework.
These tests have gradually been enhanced to provide coverage of scenarios with which Slony-I has had problems; it is to be expected that new bugs may lead to the addition of further tests.
6.7.3.1. Configuring Regression Tests
Similar to the Section 6.7.2.1 for DISORDER tests, there are three configuration parameters:
conf/slonyregress.properties.sample
This file contains Java style properties indicating how to connect to the various databases used by the regression tests, including paths to tools such as slon and slonik
The sample file is to be copied to conf/slonyregress.properties, and customized to indicate your local configuration. By using a .sample file, a developer may run tests within a Git tree, and not need to worry about their customizations interfering with the "canonical" sample configuration provided.
conf/java.conf.sample
This is a shell script containing a path indicating where the clustertest Java code (e.g. - the clustertest-coordinator.jar file) may be found. This is also used, indirectly to determine where additional Java .jar files such as the JDBC driver are located.
conf/log4j.properties
Identical to configuration for DISORDER.