Durable Workflow Server
Durable Workflow Server is the self-hosted, language-neutral runtime for
durable workflows. It records workflow state and history, matches tasks to
workers, fires timers and schedules, manages namespaces, and resumes execution
after process or infrastructure restarts.
Applications connect through first-party PHP, Python, and Rust SDKs. Workers
run in your application environment and can scale independently from Server.
For a managed runtime, use Durable Workflow Cloud.
Laravel applications can also run the engine directly in
embedded mode.
Documentation | Sample App | CLI | Waterline
Run Server
PHP Server 2.11.8 is retired and contains Laravel dependencies affected by
published advisories. Existing PHP installations should follow the
PHP retirement and security notice
for immediate exposure reduction and the supported native migration guidance.
For the 21 October 2026 critical PHP update, follow the
PHP runtime security notice and verify
the patched runtime in every deployed role.
The published Compose stack starts the native Server with MySQL. Native is a
single process: it creates its schema and the default namespace at startup and
runs its own maintenance, so there is no Redis, queue worker or scheduler. The
PHP 2.x stack remains available as
docker-compose.php.yml for installations that pin
it; see the native takeover runbook to
move an existing PHP database to native.
curl -fsSLO https://raw.githubusercontent.com/durable-workflow/server/main/docker-compose.published.yml
export DW_AUTH_TOKEN=dev-token
docker compose -f docker-compose.published.yml up -d --wait
curl http://localhost:8080/api/health
curl http://localhost:8080/api/ready
curl -H "Authorization: Bearer $DW_AUTH_TOKEN" \
http://localhost:8080/api/cluster/info
The file requires DW_AUTH_TOKEN and publishes the API on 127.0.0.1 only;
set SERVER_BIND_ADDRESS to a protected private interface for remote access and
terminate HTTPS at a reverse proxy. This development configuration uses the
shared dev-token. Production
deployments must set a strong DW_AUTH_TOKEN (native accepts it for every
role), should add role-scoped worker, operator, and administrator tokens, and
should pin the Server image by version or digest. See the
self-hosting reference for SQLite, production
Compose, authentication, backup, upgrade, API, and configuration guidance.
For a tested recovery sequence, see the
self-hosted backup and restore procedure.
Readiness checks database connectivity and setup. A successful check does not
verify database write capacity. See database storage recovery
for full disks, tablespaces and temporary storage.
See PHP database ownership checks for Rust
schema refusal, database permissions and SQLite crash-recovery capacity.
For terminal summaries that undercount history bytes, follow the
selected-run summary repair guide.
Install the current CLI for server administration and workflow inspection:
curl -fsSL https://durable-workflow.com/install.sh | sh
export PATH="$HOME/.local/bin:$PATH"
Choose an SDK
SDK clients start and inspect workflows. SDK workers register workflow and
activity types, poll named task queues, execute user code, and report results
to Server. Type names and portable Avro values form the cross-language
contract, so a workflow in one language can dispatch activities to another.
Deployment Paths
Server supports SQLite for a single-node runtime and MySQL or PostgreSQL for
shared durable state. Multi-node deployments use shared Redis for queue and
coordination state. The database remains authoritative for workflow history.
Namespaces using external payloads in a multi-node deployment should use the
built-in S3-compatible shared storage
instead of node-local files.
Capabilities
- Durable workflow and activity execution with retries, timeouts, and heartbeats
- Signals, queries, updates, timers, schedules, and child workflows
- Search attributes, memo, history export, and namespace retention policies
- Worker sessions, task-queue routing, build rollout controls, and backpressure
- Role-scoped authentication and machine-readable capability discovery
- External payload storage for large encoded values
- Waterline-compatible operational APIs and Prometheus metrics
Capacity
Published throughput uses DW Standard Workflow v1,
a versioned one-activity workflow with fixed Avro payloads, history shape,
concurrency, warmup, measurement, and qualification rules. Timer, signal,
query, replay, fanout, and mixed workloads are reported separately so a
generic "workflows per second" number cannot hide the workload being measured.
Reference
Optional trace context
A configured application image can persist bounded W3C context across compatible
SDK workers. Enable the optional API and discover the actual capability as
explained in PHP Server trace context. The guide
covers nullable schema migrations, baggage controls and runtime boundaries.
Development
git clone https://github.com/durable-workflow/server.git
cd server
cp .env.example .env
docker compose up -d --wait
Repository checks and focused test commands are documented in
CONTRIBUTING.md. Bugs and feature requests belong in
GitHub Issues. The
native publication and alias-promotion runbook
describes the separate qualification and protected release phases.
Published-image qualification runs from the release tag itself, so the runner
and the producer are the same commit. The qualification subject is the tag, its
commit and the published index, plus the pinned MySQL and QEMU images committed
in scripts/ci/native-qualification-subject.json. V2 receipts retain each actual
dependency image, the owned resource scope, body observations and verified
cleanup. The four Compose cells
retain engine target IDs, platform manifests and config digests distinctly, and
run on amd64 hosts; their arm64 server cells use the pinned QEMU helper. These
receipts do not select alias promotion or a GitHub Release.
License
Durable Workflow Server is released under the MIT License.