Skip to content

Latest commit

 

History

History
108 lines (68 loc) · 3.33 KB

developer_docs.md

File metadata and controls

108 lines (68 loc) · 3.33 KB
Raw

tranSMART Batch

Build Status

tranSMART pipeline alternative to ETL, using Spring Batch.

Building

./gradlew shadowJar

The executable will now be at build/libs/transmart-batch-<version>.jar.

To run it, execute:

java -jar build/libs/transmart-batch-<version>.jar

or use the script transmart-batch.sh.

Configuring

Create a file named batchdb.properties. Sample:

batch.jdbc.driver=org.postgresql.Driver
batch.jdbc.url=jdbc:postgresql://localhost:5432/transmart
batch.jdbc.user=glopes
batch.jdbc.password=glopes

By default, transmart-batch will look for this file in the working directory. You can override this behavior with the -c option. Please note that on newer versions of Postgres, the default port is 5433.

Preparation of the Database

This step is no longer needed in new PostgreSQL instances. Transmart-data already creates the necessary tables for spring-batch.

On old databases (or Oracle), this needs to be done only once per database. It creates the spring-batch schema. Running it with the capsule executable is probably possible, but not supported. Run:

./gradlew setupSchema

Specifying an alternative config file can be done with the Java system property propertySource. A user with administrative privileges must be configured.

Running

In its simplest form:

./transmart-batch.sh -p /path/to/STUDY_NAME/clinical.params

Beware: the parent directory of the clinical.params file will be study name. See the usage help for more information.

Worthy of mention is that transmart-batch will refuse to re-run a job with the same parameters (which depend on the clinical.params file contents). Use -n to force the job to be re-run.

Details on how to prepare your datam, mapping and parameter files for transmart-batch can be found in the docs folder.

To restart a job, take a note of the job execution id at the beginning of the failed job:

org...BetterExitMessageJobExecutionListener - Job id is 1186, execution id is 1271

Then run:

./transmart-batch.sh -p /path/to/STUDY_NAME/<type>.params -r -j <execution id>

Logging can be customized by creating a logback.groovy file in the working directory. You can copy the original logback.groovy and remove the part before the comment CUT HERE:

bsdtar -O -xvf transmart-batch-capsule.jar logback.groovy | \
  vim -c '/CUT HERE' -c '1,+1d' -c 'wq logback.groovy' -

Development

Running can also be done without the shadow jar:

./gradlew run -Pargs='-p studies/GSE8581/clinical.params -n' --debug-jvm

Before committing, run:

./gradlew check

To run the functional tests, you need to have a dedicated PostgreSQL database setup. After setting it up with transmart-data, you will also need to run the functionalTestPrepare target. So:

// only once
make -j8 postgres // in transmart-data
./gradlew functionalTestPrepare

// everytime one wants to run the tests
./gradlew functionalTest

To test on Oracle, the procedure is similar. There is one small difference: while the preparation step must be run with an administrator account (like with PostgreSQL), the tests themselves must be run as tm_cz.