Merge pull request #112 from julienrf/strengthen-testing

Strengthen testing

.github/wait-for-port.sh

@@ -0,0 +1,11 @@
+#!/usr/bin/env bash
+
+port=$1
+echo "Waiting for port ${port}"
+attempts=0
+max_attempts=30
+while ! nc -z 127.0.0.1 $port && [[ $attempts < $max_attempts ]] ; do
+    attempts=$((attempts+1))
+    sleep 1;
+    echo "waiting... (${attempts}/${max_attempts})"
+done

.github/workflows/tests.yml

@@ -23,7 +23,12 @@ jobs:
       - name: Build migrator
         run: ./build.sh
       - name: Set up services
-        run: docker compose -f docker-compose-tests.yml up --wait
+        run: |
+          docker compose -f docker-compose-tests.yml up -d
+          .github/wait-for-port.sh 8000 # ScyllaDB
+          .github/wait-for-port.sh 8001 # DynamoDB
+          .github/wait-for-port.sh 8080 # Spark master
+          .github/wait-for-port.sh 8081 # Spark worker
       - name: Run tests
         run: sbt test
       - name: Stop services

CONTRIBUTING.md

@@ -41,3 +41,38 @@ sbt migrator/assembly
 ~~~
 
 And then re-run the tests.
+
+## Debugging
+
+The tests involve the execution of code on several locations:
+- locally (ie, on the machine where you invoke `sbt test`): tests initialization and assertions
+- on the Spark master node: the `Migrator` entry point
+- on the Spark worker node: RDD operations
+
+In all those cases, it is possible to debug them by using the Java Debug Wire Protocol.
+
+### Local Debugging
+
+Follow the procedure documented [here](https://stackoverflow.com/a/15505308/561721).
+
+### Debugging on the Spark Master Node
+
+1. In the file `MigratorSuite.scala`, uncomment the line that sets the
+   `spark.driver.extraJavaOptions`.
+2. Set up the remote debugger of your IDE to listen to the port 5005.
+3. Run a test
+4. When the test starts a Spark job, it waits for the remote debugger
+5. Start the remote debugger from your IDE.
+6. The test execution resumes, and you can interact with it from your debugger.
+
+### Debugging on the Spark Worker Node
+
+1. In the file `MigratorSuite.scala`, uncomment the line that sets the
+   `spark.executor.extraJavaOptions`.
+2. Set up the remote debugger of your IDE to listen to the port 5006.
+3. Run a test
+4. When the test starts an RDD operation, it waits for the remote debugger.
+   Note that the Spark master node will not display the output of the worker node,
+   but you can see it in the worker web UI: http://localhost:8081/.
+5. Start the remote debugger from your IDE.
+6. The test execution resumes, and you can interact with it from your debugger.

docker-compose-tests.yml

@@ -36,6 +36,7 @@ services:
     networks:
       - scylla
     expose:
+      - 5005
       - 7001
       - 7002
       - 7003
@@ -46,6 +47,7 @@ services:
       - 6066
     ports:
       - 4040:4040
+      - 5005:5005
       - 6066:6066
       - 7077:7077
       - 8080:8080
@@ -65,13 +67,15 @@ services:
     networks:
       - scylla
     expose:
+      - 5006
       - 7012
       - 7013
       - 7014
       - 7015
       - 7016
       - 8881
     ports:
+      - 5006:5006
       - 8081:8081
     depends_on:
       - spark-master

tests/src/test/scala/com/scylladb/migrator/MigratorSuite.scala

@@ -113,6 +113,9 @@ trait MigratorSuite extends munit.FunSuite {
         "--master", "spark://spark-master:7077",
         "--conf", "spark.driver.host=spark-master",
         "--conf", s"spark.scylla.config=/app/configurations/${migratorConfigFile}",
+        // Uncomment one of the following lines to plug a remote debugger on the Spark master or worker.
+        // "--conf", "spark.driver.extraJavaOptions=-agentlib:jdwp=transport=dt_socket,server=y,suspend=y,address=5005",
+        // "--conf", "spark.executor.extraJavaOptions=-agentlib:jdwp=transport=dt_socket,server=y,suspend=y,address=5006",
         "/jars/scylla-migrator-assembly-0.0.1.jar"
       )
     ).run().exitValue().tap { statusCode =>