conductor-oss--conductor/e2e

Conductor E2E Tests

End-to-end tests for conductor-oss covering workflow execution, task types, control flow, event handling, metadata operations, and data processing.

Prerequisites

• Docker and Docker Compose
• Java 21+
• Gradle (use the ./gradlew wrapper at the repo root)

Quick Start

The recommended way to run the full suite is via a convenience script. Each script starts the required Docker services, waits for the server to be healthy, runs all tests, then tears down.

From the repo root:

# Postgres + Elasticsearch 7 (recommended for this project)
./e2e/run_tests-postgres.sh

# Other backends:
./e2e/run_tests.sh                  # Redis + Elasticsearch 7 (default)
./e2e/run_tests-es8.sh              # Redis + Elasticsearch 8
./e2e/run_tests-postgres-es7.sh     # Postgres + Elasticsearch 7 (explicit)
./e2e/run_tests-redis-os2.sh        # Redis + OpenSearch 2.x
./e2e/run_tests-redis-os3.sh        # Redis + OpenSearch 3.x
./e2e/run_tests-mysql.sh            # MySQL

The server listens on http://localhost:8000 by default. Override with:

SERVER_ROOT_URI=http://localhost:9090 ./e2e/run_tests-postgres.sh

Manual Setup

If you already have a Conductor server running, skip the scripts and run Gradle directly:

# Using the environment variable (recommended)
RUN_E2E=true SERVER_ROOT_URI=http://localhost:6000 ./gradlew :conductor-e2e:test

# Or using Gradle properties
./gradlew :conductor-e2e:test -PrunE2E -DSERVER_ROOT_URI=http://localhost:8000

Tests are skipped by default during a normal ./gradlew build to avoid requiring a running server. The RUN_E2E=true env var or -PrunE2E flag is required to enable them.

Building a local server image

If you need to test against a locally built server (e.g., after changing core/ code):

# 1. Build the server JAR
./gradlew :conductor-server:build -x test

# 2. Build the Docker image
docker build -t conductor:server -f docker/server/Dockerfile .

# 3. Start with the e2e compose file (maps port 6000 → 8080)
docker compose -f docker/docker-compose-postgres-e2e.yaml up -d

# 4. Run tests
RUN_E2E=true SERVER_ROOT_URI=http://localhost:6000 ./gradlew :conductor-e2e:test

Test Options

Run a specific test class or method

RUN_E2E=true SERVER_ROOT_URI=http://localhost:6000 \
  ./gradlew :conductor-e2e:test --tests "io.conductor.e2e.control.SubWorkflowTests"

# Single method
RUN_E2E=true SERVER_ROOT_URI=http://localhost:6000 \
  ./gradlew :conductor-e2e:test --tests "io.conductor.e2e.control.DoWhileTests.testDoWhileSetVariableFix"

Run the server-backed agent lifecycle suite

RUN_E2E=true SERVER_ROOT_URI=http://localhost:8080 \
  ./gradlew :conductor-e2e:test --tests "io.conductor.e2e.task.AgentTaskTests"

This deterministic suite needs no LLM credentials. It covers basic and nested agent execution, measured callback/poll intervals, manually completed long-running work, multi-turn human input, concurrent calls, parent and explicit CANCEL_AGENT cancellation, child cancellation, wrapper deadlines, agent workflow timeouts, internal failures, and missing agents.

Exclude a test class

./gradlew :conductor-e2e:test -PrunE2E -PexcludeTests="**/HTTPTaskTests*"

# Multiple patterns (comma-separated)
./gradlew :conductor-e2e:test -PrunE2E -PexcludeTests="**/HTTPTaskTests*,**/GraaljsTests*"

Parallelism

Tests run with maxParallelForks = 4 by default. Reduce if the server is under-resourced:

./gradlew :conductor-e2e:test -PrunE2E --max-workers=1

Test Suite Overview

Package	What it covers
control	DO_WHILE, SWITCH, SUB_WORKFLOW, DYNAMIC_FORK
task	AGENT lifecycle, WAIT, HTTP, concurrency limit, task timeout, backoff
workflow	Retry, restart, rerun, search, priority, failure workflows
processing	GraalJS inline tasks, SET_VARIABLE, JSON_JQ
event	Event handlers
metadata	Workflow/task definition CRUD, event handler registration

Disabled Tests

18 tests are @Disabled due to conductor-oss behavioural differences or infrastructure constraints. All other skips during ./gradlew build (without RUN_E2E=true) are simply the suite waiting for a server — those tests are not broken.

Rerun behaviour (conductor-oss does not support rerun from non-terminal or complex task states)

Class	Method	Reason
WorkflowRerunTests	testRerunFromWaitTask	Rerunning a RUNNING workflow is not allowed; conductor-oss requires a terminal state first
WorkflowRerunTests	testRerunForkJoinWorkflow	Rerun from a completed fork-branch task does not re-schedule sibling branches
WorkflowRerunTests	testRerunForkJoinWorkflowWithLoopTask	Rerun from a DO_WHILE task inside a fork terminates the workflow instead of resuming
WorkflowRerunTests	testRerunForkJoinWorkflowWithLoopTask2	Rerun from a DO_WHILE task inside a fork terminates the workflow instead of resuming
WorkflowRerunTests	testRerunForkJoinWorkflowWithLoopOverTask	Rerun from a DO_WHILE iteration task inside a fork terminates the workflow instead of resuming
WorkflowRerunTests	testRerunSubWorkflowInsideFork	Rerun from a SUB_WORKFLOW task inside a fork terminates the workflow instead of resuming
WorkflowRerunTests	testRerunSubWorkflowInsideFork_SequentialBranch	Rerun from a SUB_WORKFLOW task inside a fork terminates the workflow instead of resuming
WorkflowRerunTests	testDoWhileRerun	DO_WHILE task rerun leaves the workflow TERMINATED; sync task status is not reset to SCHEDULED
WorkflowRerunTests	testSwitchTaskRerun	SWITCH task rerun leaves the workflow TERMINATED; sync task status is not reset to SCHEDULED
WorkflowRerunTests	testRerunForkJoinWithWaitAndSwitchTasks	Rerun from fork-join with WAIT/SWITCH tasks terminates the workflow instead of resuming
WorkflowRerunTests	switchRerunIssue	SWITCH task rerun leaves the workflow TERMINATED; sync task status is not reset to SCHEDULED
WorkflowRerunTests	switchRerunIssue2	SWITCH inside DO_WHILE rerun leaves the workflow TERMINATED

Infrastructure / external dependencies

Class	Method	Reason
HTTPTaskTests	HTTPAsyncCompleteTest	Requires httpbin-server internal service (http://httpbin-server:8081) not in the conductor-oss e2e docker setup
SyncWorkflowExecutionTest	testSyncWorkflowExecution6	Depends on external HTTP services (orkes-api-tester.orkesconductor.com) not reliably accessible from conductor-oss e2e
PollTimeoutTests	testPollTimeout	Postgres-backed queue does not drain tasks from terminated workflows within the required window; cleanup timing is non-deterministic

SDK / test isolation

Class	Method	Reason
JavaSDKTests	testSDK	Shared AnnotatedWorkerExecutor thread pool is shut down by SwitchTests.@AfterAll when tests run in the same JVM; subsequent worker tasks are rejected
SwitchTests	testSwitchNegetive	SDK-based executeDynamic with a switch default case does not complete within the timeout in conductor-oss

Postgres-specific timing

The postgres-backed WAIT task sweeper adds roughly 10 seconds of overhead on top of the configured wait duration. Several tests account for this with extended timeouts (e.g., a "2 second" WAIT task may take up to 15 seconds end-to-end). If tests are flaky on slower machines, check whether a timeout needs to be increased further.