Build Resilient Distributed Systems with Temporal Go SDK

Temporal Go SDK (temporal-golang-pro) is an expert-level guide for building resilient, scalable, deterministic distributed systems with the Temporal Go SDK, turning vague orchestration requirements into production-grade Go implementations focused on durable execution, strict determinism, and enterprise-scale worker configuration.

Use this when designing distributed systems that need durable state and reliable orchestration, implementing complex long-running workflows spanning days to months or Saga patterns, optimizing worker concurrency, mTLS security, or custom interceptors, ensuring reliability via idempotent activities and retry policies, or versioning running workflows and doing zero-downtime worker updates. Do not use it for Temporal in other SDKs (Python, Java, TypeScript each have their own -pro skill), simple request/response tasks with no durability or coordination need, or high-level design without implementation, which is covered by workflow-orchestration-patterns instead.

The process: gather context proactively, including the target cluster (Cloud vs self-hosted) and namespace, task queue names and expected throughput, security requirements like mTLS paths, and failure modes with retry/timeout policy; verify workflow code against 5 determinism rules before suggesting it - no native goroutines, no native time (time.Now/time.Sleep), no non-deterministic map iteration (sort keys), no direct external I/O, and no non-deterministic random numbers; implement incrementally, starting with shared Protobuf/data classes, then activities, then workflows, then workers; and pull in the implementation playbook or testing strategies resources for advanced patterns like Sagas, interceptors, or replay testing. Worked examples cover a versioned subscription workflow using workflow.GetVersion to safely change billing logic (28 vs 30-day cycles) mid-flight, a full mTLS worker setup dialing the cluster with a loaded client certificate and CA pool, and a selector-based approval workflow racing a signal channel against a 72-hour timer.

Capabilities span worker management (worker.Options fields like MaxConcurrentActivityTaskPollers, WorkerStopTimeout, StickyScheduleToStartTimeout), interceptors for cross-cutting concerns such as logging, tracing, and auth, and custom data converters for Protobuf or encrypted payloads; advanced workflow patterns like durable concurrency via workflow.Go/workflow.Channel/workflow.Selector instead of native primitives, safe versioning via workflow.GetVersion, ContinueAsNew for history-size limits (default 50MB or 50K events), and child-workflow lifecycle and signal management; and testing and observability via WorkflowTestSuite, activity and child-workflow mocking, replay testing against production event histories, and Prometheus/OpenTelemetry metrics. It names three related skills - grpc-golang for transport and Protobuf design, golang-pro for general Go performance, and workflow-orchestration-patterns for language-agnostic strategy - and explicitly does not cover Temporal Cloud UI navigation, non-Go SDKs, or experimental features like Nexus or multi-cluster replication, assuming Temporal Server v1.20+ and Go SDK v1.25+.

Go developers building production Temporal workflows who need concrete, determinism-safe code patterns and troubleshooting guidance for specific failures: a determinism-mismatch panic usually traces to a logic change made without workflow.GetVersion or hidden non-determinism like native map iteration, a history-size-exceeded error means ContinueAsNew isn't implemented, and a worker hang means checking WorkerStopTimeout and activity context-cancellation handling.