bvaradar commented on a change in pull request #626: Adding documentation for
hudi test suite
URL: https://github.com/apache/incubator-hudi/pull/626#discussion_r344022879
##########
File path: docs/test_suite.md
##########
@@ -0,0 +1,155 @@
+---
+title: Test Suite
+keywords: test suite
+sidebar: mydoc_sidebar
+permalink: test_suite.html
+toc: false
+summary: In this page, we will discuss the Hudi Test suite to perform end to
end tests
+
+This page describes in detail how to run end to end tests on a hudi dataset
that helps in improving our confidence
+in a release as well as perform large scale performance benchmarks.
+
+### Objectives
+
+1. Test with different versions of core libraries and components such as
`hdfs`, `parquet`, `spark`,
+`hive` and `avro`.
+2. Generate different types of workloads across different dimensions such as
`payload size`, `number of updates`,
+`number of inserts`, `number of partitions`
+3. Perform multiple types of operations such as `insert`, `bulk_insert`,
`upsert`, `compact`, `query`
+4. Support custom post process actions and validations
+
+### High Level Design
+
+{% include image.html file="hudi_test_suite_design.png"
alt="hudi_test_suite_design.png" %}
+
+The Hudi test suite runs as a long running spark job. The suite is divided
into the following high level components :
+
+##### Workload Generation
+
+This component does the work of generating the workload; `inserts`, `upserts`
etc.
+
+##### Workload Scheduling
+
+Depending on the type of workload generated, data is either ingested into the
target hudi
+dataset or the corresponding workload operation is executed. For example
compaction does not necessarily need a workload
+to be generated/ingested but can require an execution.
+
+### Usage instructions
+
+
+##### Entry class to the test suite
+
+```
+org.apache.hudi.bench.job.HudiTestSuiteJob.java - Entry Point of the hudi test
suite job. This
+class wraps all the functionalities required to run a configurable integration
suite.
+```
+
+##### Configurations required to run the job
+```
+org.apache.hudi.bench.job.HudiTestSuiteConfig - Config class that drives the
behavior of the
+integration test suite. This class extends from
com.uber.hoodie.utilities.DeltaStreamerConfig. Look at
+link#HudiDeltaStreamer page to learn about all the available configs
applicable to your test suite.
+```
+
+##### Generating a custom Workload Pattern
+```
+There are 2 ways to generate a workload pattern
+
+1. Programatically
+Choose to write up the entire DAG of operations programatically, take a look
at WorkflowDagGenerator class.
+Once you're ready with the DAG you want to execute, simply pass the class name
as follows
+spark-submit
+...
+...
+--class org.apache.hudi.bench.job.HudiTestSuiteJob
+--workload-generator-classname
org.apache.hudi.bench.dag.scheduler.<your_workflowdaggenerator>
+...
+2. YAML file
+Choose to write up the entire DAG of operations in YAML, take a look at
complex-workload-dag-cow.yaml or
+complex-workload-dag-mor.yaml.
+Once you're ready with the DAG you want to execute, simply pass the yaml file
path as follows
+spark-submit
+...
+...
+--class org.apache.hudi.bench.job.HudiTestSuiteJob
+--workload-yaml-path /path/to/your-workflow-dag.yaml
+...
+```
+
+#### Building the test suite
+
+The test suite can be found in the `hudi-bench` module. Use the
`prepare_integration_suite.sh` script to build
+the test suite, you can provide different parameters to the script.
+
+```
+shell$ ./prepare_integration_suite.sh --help
Review comment:
Great !!
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
