[Brend-Smits]

Description

In v8.0.0 the lambda bundles contained several copies of @smithy/core and @aws-sdk/core. The SDK registers service exceptions in a TypeRegistry owned by @smithy/core and looks them up from @aws-sdk/core, so with duplicate copies the lookup missed and service failures reached the lambdas as a plain Error: the name is Error and the code is only in Code, __type or the message. Every check keyed on error.name then missed the error. main and v8.1.0 no longer have the duplicates, but nothing in the code or CI protects against them coming back, and the checks only trusted name.

This PR makes those checks tolerant of that shape and adds a guard for the root cause:

• New @aws-github-runner/aws-error-util lib with awsErrorCodes(), which reads the code from name (ignoring the generic Error), Code, __type (with or without a namespace) or a message that is a bare code, along the cause chain. It lives in its own lib rather than in aws-powertools-util/aws-ssm-util because many suites replace those modules with partial mocks, and so storage-providers and compute-providers can share it.
• Runner group cache store: a missing cache parameter is a cache miss again, so scale-up resolves the group from GitHub instead of launching an instance, terminating it and retrying forever (v8.0.0: runner-group cache miss treated as fatal because SSM ParameterNotFound loses its name in the lambda bundle #5571).
• requestFailureCodes() also records the derived codes as aws-name:<code>, so request-level EC2 failures such as UnauthorizedOperation match scale_errors instead of being dropped as non-retryable (v8.0.0: EC2 request errors lose their name in the lambda bundles, so scale_errors never match request-level failures (UnauthorizedOperation) #5565). Existing codes, including aws-name:Error, are unchanged.
• SSM runner config consumer: a missing parameter keeps polling and throttling is retried, using the same helper (this had the same name-only check).
• A test that fails when more than one installed copy of @smithy/core or @aws-sdk/core is reachable from the AWS clients used by the workspaces, which is the check requested in v8.0.0: EC2 request errors lose their name in the lambda bundles, so scale_errors never match request-level failures (UnauthorizedOperation) #5565.

A message only counts as a code when it is a single identifier, so ordinary messages such as Request failed with a socket hang up are not turned into failure codes, and unnamed AccessDeniedException, ThrottlingException or ParameterVersionNotFound errors are still treated as failures where they should be.

Test Plan

• Added unit tests for the lost-name shape in the cache store, the config consumer, and EC2 create (CreateFleet, RunInstances dedicated host, AMI lookup, and the resulting retryable classification through scale_errors), plus unit tests for awsErrorCodes() at 100% coverage. 11 of the new tests failed before the fix and pass now.
• The installed SDK no longer produces the exact v8.0.0 error shape, so these tests use hand-built errors in that shape. Two aws-ssm-util tests pin the error chain that getParameter really produces (GetParameterError with the SDK error as cause) so the fixtures match what the cache store sees.
• The dependency guard passes on main and fails on the v8.0.0 lockfile (13 copies of @smithy/core and 10 of @aws-sdk/core reachable from the lambdas).
• yarn format-check, yarn lint, yarn test (12 projects, uncached) and yarn build all pass in lambdas/.
• Not deployed to a live environment from this branch.

Related Issues

Fixes #5571. Related to #5565: this adds the guard against duplicate SDK cores that it asks for, and makes the affected checks robust.

[github-actions]

Dependency Review

The following issues were found:

• ✅ 0 vulnerable package(s)
• ✅ 0 package(s) with incompatible licenses
• ✅ 0 package(s) with invalid SPDX license definitions
• ⚠️ 2 package(s) with unknown licenses.

See the Details below.

License Issues

lambdas/libs/compute-providers/package.json

Package	Version	License	Issue Type
@aws-github-runner/aws-error-util		Null	Unknown License

lambdas/libs/storage-providers/package.json

Package	Version	License	Issue Type
@aws-github-runner/aws-error-util		Null	Unknown License

OpenSSF Scorecard

Package	Version	Score	Details
npm/@aws-github-runner/aws-error-util		Unknown	Unknown
npm/@aws-github-runner/aws-error-util		Unknown	Unknown

Scanned Files

• lambdas/libs/compute-providers/package.json
• lambdas/libs/storage-providers/package.json

[Copilot]

🟢 Approval recommended

No active review issues are reported for the changes addressing unnamed AWS SDK errors.

No open issues 🎉

What changed in this PR

Makes Lambda error handling resilient when AWS SDK errors lose their names, addressing runner-group cache failures in #5571 and missed EC2 retry classifications in #5565.

Changes:

• Adds a shared helper that extracts AWS error codes from error fields and cause chains.
• Uses derived codes for SSM cache misses, configuration polling and retries, and EC2 failure classification.
• Adds regression tests and a dependency guard against duplicate AWS SDK core installations.

File	Description
lambdas/​yarn.lock	Registers the helper workspace and consumer dependencies.
lambdas/​libs/​storage-providers/​package.json	Adds the shared error helper dependency.
lambdas/​libs/​storage-providers/​aws/​ssm/​runner-group-cache-store.ts	Recognizes missing cache parameters despite lost error names.
lambdas/​libs/​storage-providers/​aws/​ssm/​runner-group-cache-store.test.ts	Tests unnamed cache misses and propagation of other failures.
lambdas/​libs/​storage-providers/​aws/​ssm/​runner-config-consumer.ts	Recognizes unnamed missing-parameter errors while polling.
lambdas/​libs/​storage-providers/​aws/​ssm/​runner-config-consumer.test.ts	Tests unnamed polling, throttling, and access-error handling.
lambdas/​libs/​storage-providers/​aws/​ssm/​runner-config-consumer-common.ts	Uses derived codes to identify retryable errors.
lambdas/​libs/​compute-providers/​package.json	Adds the shared error helper dependency.
lambdas/​libs/​compute-providers/​aws/​ec2/​src/​runners.ts	Adds derived AWS codes to request failure classifications.
lambdas/​libs/​compute-providers/​aws/​ec2/​src/​runners.test.ts	Covers unnamed EC2 and AMI lookup failures, including retry classification.
lambdas/​libs/​aws-ssm-util/​src/​index.test.ts	Verifies preservation of SDK error causes and messages.
lambdas/​libs/​aws-powertools-util/​src/​aws-sdk-single-instance.test.ts	Guards against multiple reachable SDK core installations.
lambdas/​libs/​aws-error-util/​vitest.config.ts	Sets full coverage thresholds for the helper.
lambdas/​libs/​aws-error-util/​tsconfig.json	Configures the helper’s TypeScript source scope.
lambdas/​libs/​aws-error-util/​src/​index.ts	Extracts error codes with bounded, cycle-safe cause traversal.
lambdas/​libs/​aws-error-util/​src/​index.test.ts	Tests code extraction, input filtering, and traversal limits.
lambdas/​libs/​aws-error-util/​package.json	Defines the helper workspace and development scripts.

_{Review effort: Balanced}