openova/platform/temporal

Temporal

Durable workflow orchestration for microservices.

Status: Accepted | Updated: 2026-02-09

---

Overview

Temporal is a durable execution platform that makes it simple to build reliable, long-running workflows and microservice orchestrations. Unlike traditional message queues or job schedulers, Temporal provides durable execution: workflow code survives process crashes, node failures, and even entire cluster restarts without losing state. Developers write workflows as ordinary code in their language of choice, and Temporal handles retries, timeouts, and state persistence transparently.

Within OpenOva, Temporal serves as the workflow orchestration engine for the Fabric data and integration product. It handles saga patterns for distributed transactions, long-running business processes, scheduled jobs, and any operation that needs reliable execution across multiple services. Temporal replaces fragile combinations of message queues, cron jobs, and custom state machines with a single, battle-tested platform.

Temporal's architecture separates the server (which manages workflow state) from workers (which execute workflow and activity code). Workers are stateless and can be scaled independently. The server persists all workflow state to a database, ensuring that workflows survive any infrastructure failure. SDKs are available for Go, Java, Python, and TypeScript, making Temporal accessible to polyglot teams.

---

Architecture

flowchart TB
    subgraph Clients["Workflow Clients"]
        API[API Services]
        Scheduler[Scheduled Jobs]
        Events[Event Handlers]
    end

    subgraph Temporal["Temporal Server"]
        Frontend[Frontend Service]
        History[History Service]
        Matching[Matching Service]
        Worker_svc[Internal Worker]
    end

    subgraph Persistence["Persistence"]
        PG[PostgreSQL / CNPG]
        ES[Elasticsearch / OpenSearch]
    end

    subgraph Workers["Application Workers"]
        W1[Order Worker]
        W2[Payment Worker]
        W3[Notification Worker]
    end

    Clients --> Frontend
    Frontend --> History
    Frontend --> Matching
    History --> PG
    Worker_svc --> ES
    Matching --> W1
    Matching --> W2
    Matching --> W3

Saga Pattern

sequenceDiagram
    participant C as Client
    participant T as Temporal
    participant O as Order Service
    participant P as Payment Service
    participant I as Inventory Service
    participant N as Notification Service

    C->>T: Start OrderSaga
    T->>O: CreateOrder
    O-->>T: Order Created
    T->>P: ProcessPayment
    P-->>T: Payment OK
    T->>I: ReserveInventory
    I-->>T: Inventory Reserved
    T->>N: SendConfirmation
    N-->>T: Notification Sent
    T-->>C: Saga Complete

    Note over T,I: If ReserveInventory fails:
    T->>P: CompensatePayment (refund)
    T->>O: CancelOrder

---

Key Features

Feature	Description
Durable Execution	Workflows survive process/node/cluster failures
Saga Orchestration	Coordinate distributed transactions with compensation
Retry Policies	Configurable retry with exponential backoff per activity
Timeouts	Start-to-close, schedule-to-start, and heartbeat timeouts
Cron Workflows	Replace cron jobs with reliable scheduled workflows
Versioning	Deploy new workflow logic without breaking running instances
Signals & Queries	Send data to and read state from running workflows
Child Workflows	Compose complex workflows from smaller building blocks
Visibility	Search and filter workflows by custom attributes

---

Configuration

Helm Values

temporal:
  server:
    replicas: 3
    config:
      persistence:
        default:
          driver: sql
          sql:
            driver: postgres12
            host: temporal-postgres.databases.svc
            port: 5432
            database: temporal
            user: temporal
            password: ${PG_PASSWORD}  # From ESO
        visibility:
          driver: sql
          sql:
            driver: postgres12
            host: temporal-postgres.databases.svc
            port: 5432
            database: temporal_visibility
            user: temporal
            password: ${PG_PASSWORD}  # From ESO

    resources:
      requests:
        cpu: 500m
        memory: 1Gi
      limits:
        cpu: 2
        memory: 4Gi

  admintools:
    enabled: true

  web:
    enabled: true
    ingress:
      enabled: true
      hosts:
        - temporal.fuse.<domain>

  prometheus:
    enabled: true

Namespace Setup

# Create Temporal namespace for workload isolation
tctl namespace register \
  --namespace orders \
  --retention 30d \
  --description "Order processing workflows"

---

Workflow Examples

Go SDK - Order Saga

package workflows

import (
    "time"
    "go.temporal.io/sdk/temporal"
    "go.temporal.io/sdk/workflow"
)

func OrderSagaWorkflow(ctx workflow.Context, order Order) (OrderResult, error) {
    retryPolicy := &temporal.RetryPolicy{
        InitialInterval:    time.Second,
        BackoffCoefficient: 2.0,
        MaximumInterval:    time.Minute,
        MaximumAttempts:    5,
    }
    activityOpts := workflow.ActivityOptions{
        StartToCloseTimeout: 30 * time.Second,
        RetryPolicy:         retryPolicy,
    }
    ctx = workflow.WithActivityOptions(ctx, activityOpts)

    // Step 1: Create Order
    var orderID string
    err := workflow.ExecuteActivity(ctx, CreateOrder, order).Get(ctx, &orderID)
    if err != nil {
        return OrderResult{}, err
    }

    // Step 2: Process Payment (with compensation)
    var paymentID string
    err = workflow.ExecuteActivity(ctx, ProcessPayment, orderID, order.Amount).Get(ctx, &paymentID)
    if err != nil {
        // Compensate: cancel the order
        _ = workflow.ExecuteActivity(ctx, CancelOrder, orderID).Get(ctx, nil)
        return OrderResult{}, err
    }

    // Step 3: Reserve Inventory (with compensation)
    err = workflow.ExecuteActivity(ctx, ReserveInventory, orderID, order.Items).Get(ctx, nil)
    if err != nil {
        // Compensate: refund payment, cancel order
        _ = workflow.ExecuteActivity(ctx, RefundPayment, paymentID).Get(ctx, nil)
        _ = workflow.ExecuteActivity(ctx, CancelOrder, orderID).Get(ctx, nil)
        return OrderResult{}, err
    }

    // Step 4: Send Confirmation
    _ = workflow.ExecuteActivity(ctx, SendConfirmation, orderID).Get(ctx, nil)

    return OrderResult{OrderID: orderID, PaymentID: paymentID, Status: "completed"}, nil
}

Python SDK - Data Pipeline

from temporalio import workflow, activity
from datetime import timedelta

@activity.defn
async def extract_data(source: str) -> dict:
    # Extract data from source system
    ...

@activity.defn
async def transform_data(raw_data: dict) -> dict:
    # Apply business transformations
    ...

@activity.defn
async def load_data(transformed: dict) -> str:
    # Write to destination
    ...

@workflow.defn
class DataPipelineWorkflow:
    @workflow.run
    async def run(self, source: str) -> str:
        raw = await workflow.execute_activity(
            extract_data, source,
            start_to_close_timeout=timedelta(minutes=10),
        )
        transformed = await workflow.execute_activity(
            transform_data, raw,
            start_to_close_timeout=timedelta(minutes=30),
        )
        result = await workflow.execute_activity(
            load_data, transformed,
            start_to_close_timeout=timedelta(minutes=10),
        )
        return result

---

Worker Deployment

apiVersion: apps/v1
kind: Deployment
metadata:
  name: order-workflow-worker
  namespace: fuse
spec:
  replicas: 3
  template:
    spec:
      containers:
        - name: worker
          image: registry.<domain>/fuse/order-worker:latest
          env:
            - name: TEMPORAL_HOST
              value: temporal-frontend.temporal.svc:7233
            - name: TEMPORAL_NAMESPACE
              value: orders
            - name: TEMPORAL_TASK_QUEUE
              value: order-processing
          resources:
            requests:
              cpu: 250m
              memory: 512Mi
            limits:
              cpu: 1
              memory: 1Gi

---

Monitoring

Metric	Description
temporal_workflow_started_total	Workflows started
temporal_workflow_completed_total	Workflows completed successfully
temporal_workflow_failed_total	Workflows failed
temporal_workflow_task_queue_depth	Pending tasks per queue
temporal_activity_execution_latency	Activity execution duration
temporal_workflow_endtoend_latency	Total workflow duration
temporal_schedule_missed_catchup_window	Missed cron schedule executions

---

Consequences

Positive:

• Durable execution eliminates custom retry/state-machine code across all services
• Saga pattern support simplifies distributed transaction management
• Multi-language SDKs (Go, Java, Python, TypeScript) suit polyglot teams
• Workflow versioning enables safe deployments without breaking running instances
• Built-in visibility and search make debugging production workflows practical
• Cron workflows replace fragile crontab/CronJob setups with reliable scheduling

Negative:

• Requires PostgreSQL for persistence, adding a database dependency
• Temporal server itself needs careful sizing and operational attention
• Workflow determinism constraints require developer discipline (no random, no system clock)
• Learning curve for understanding event sourcing and replay semantics
• Debugging replayed workflows requires familiarity with Temporal's execution model
• Large workflow histories can impact performance without proper archival configuration

---

Part of OpenOva Fabric - Data & Integration