Breaking Changes

💥 Worker Heartbeats

This release adds a new feature called Worker Heartbeating. This feature spins up a single nexus-only worker per Client/Namespace/Task Queue in the background and periodically sends some basic metrics over to server for all of the workers registered on the same Client/Namespace/Task Queue.

Server version 1.29.1 and newer supports this feature. This feature will be enabled by default, although this currently requires a server dynamic config flag, --dynamic-config-value frontend.WorkerHeartbeatsEnabled=true to enable. Currently, the only way to interact with this new data is to also enable the flag --dynamic-config-value frontend.ListWorkersEnabled=true, then use CLI commands temporal worker list and temporal worker describe to query the data. If heartbeating is enabled with a server version older or with the config flag off, a single warning log Worker heartbeating configured for runtime, but server version does not support it. on worker startup will emit.

💥 With the introduction of Worker Heartbeats, we’ve deprecated the contrib/resourcetuner lib, and instead created a new contrib/sysinfo lib. The resource tuner now lives in the worker package.

💥  Nexus Error Serialization Improvements

Note: Requires Temporal Server version 1.31.0 or later.

This release of the Go SDK includes a number of enhancements to error serialization for synchronous operations.

• nexus.HandlerError can now include its own message (and stack trace if the language supports it) independent from the cause
• nexus.OperationError can now include its own message (and stack trace if the language supports it) independent from the cause
• 💥 nexus.OperationError is now included in the error chain when a synchronous operation handler returns a nexus.OperationError . The nexus.OperationError will be translated to an appropriate Temporal error based on the state of the error. Previously this error was not included in the error chain.
  • OperationStateFailed will be translated to an ApplicationError
  • OperationStateCancelled will be translated to an CanceledError

These changes only effect synchronous operations. There is no change to errors for asynchronous operations like starting a workflow.

Highlights

Nexus Operation Timeouts

Note: Requires Temporal Server version 1.31.0 or later.

This release of the Go SDK adds more timeout options on NexusOperationOptions to control how long the caller wants to wait on different stages of the operation.

Schedule-to-Start timeout

The Schedule-to-Start timeout limits how long the caller will wait for the Operation to be started by the handler.
If not set, no Schedule-to-Start timeout is enforced.

fut := c.ExecuteOperation(ctx, service.HelloOperationName, service.HelloInput{Name: name, Language: language}, workflow.NexusOperationOptions{
	ScheduleToStartTimeout: 2 * time.Minute,
})

Start-to-Close timeout

The Start-to-Close timeout limits how long the caller will wait for an asynchronous Operation to complete after it has been started.
This timeout only applies to asynchronous Operations.
If not set, no Start-to-Close timeout is enforced.

fut := c.ExecuteOperation(ctx, service.HelloOperationName, service.HelloInput{Name: name, Language: language}, workflow.NexusOperationOptions{
	StartToCloseTimeout: 5 * time.Minute,
})

Add trampolining API for versioned continue-as-new

Added support for workflows to detect when their target deployment version
has changed and opt into upgrading via continue-as-new. New APIs:

• WorkflowInfo.GetTargetWorkflowDeploymentVersionChanged() — check if the
  workflow's target version changed
• WorkflowInfo.GetContinueAsNewSuggestedReasons() — get reasons why CAN is
  suggested
• ContinueAsNewVersioningBehaviorAutoUpgrade option for ContinueAsNewError
  — explicitly upgrade to the new version on CAN

This enables a "trampolining" pattern: a pinned workflow detects a version
change, then continues-as-new with auto-upgrade to seamlessly move to the
latest deployment version. Both detection and action are fully opt-in,
avoiding accidental infinite CAN loops.

What's Changed

• Nexus caller timeouts by @bergundy in #2146
• Fix cgroup detection condition by @yuandrew in #2179
• Introduce SDK flag to use user DC on Memo's by @yuandrew in #2121
• 💥Add worker heartbeat support by @yuandrew in #2186
• workflowcheck: Fix Go 1.26 false positives for SetFinalizer and godebug.Setting by @cretz in #2192
• Bump github.com/sirupsen/logrus from 1.9.0 to 1.9.1 in /contrib/sysinfo by @dependabot[bot] in #2194
• Change ShutdownWorker err log from Debug to Warn by @yuandrew in #2195
• Make search attribute type more lenient when parsing by @cretz in #2197
• Reduce test flakes by @yuandrew in #2196
• Selector Signal fix with multiple workers by @yuandrew in #2134
• Handle disabled TLS in envconfig ToClientOptions by @dandavison in #2205
• Fix canceling external workflows that are children in testsuite by @yuandrew in #1968
• Fix standalone activity result long-poll bug by @dandavison in #2198
• 💥 Use Temporal Failures for Nexus Error Serialization by @Quinn-With-Two-Ns in #2181
• Fix 2090 child workflow goroutine leak by @kritibehl in #2200
• Add trampolining API for versioned continue-as-new by @THardy98 in #2216
• Upgrade go.opentelemetry.io/otel/sdk to v1.40.0 by @yuandrew in #2218
• Revert "Upgrade go.opentelemetry.io/otel/sdk to v1.40.0" by @yuandrew in #2221

New Contributors

• @kritibehl made their first contribution in #2200

Full Changelog: v1.40.0...v1.41.0