fix(node-dependency-matrix): recalibrate dedup expectedTestcasesExact 18->19 for agent v3.5.44#100
Merged
slayerjain merged 1 commit intoJun 10, 2026
Conversation
The kube-regression dedup assertion (enterprise-ui kube-flow.spec.ts test 8) checks the recorded test-case count against this contract value. It was calibrated against the old keploy-agent (enterprise v3.2.74) the fixture chart pinned; k8s-proxy now injects v3.5.44 (distroless + shell-free probes), which records 19. Verified this is correct, not a dedup regression: static dedup still collapses every business duplicate exactly as the contract specifies (GET /dedup/catalog 5->1, POST /dedup/order 4->1; all 16 business endpoints recorded once). The recorded 19 = 16 deduped business endpoints + 3 non-business probe/preflight buckets: GET /health (x1) and GET /expectations (x2 — its request-header-name set differs between the curl preflight and the Playwright readiness probe, and the agent's static-dedup key includes request-header names). v3.2.74 recorded fewer only because it undercaptured that probe traffic (an upstream OSS chunked/no-Content-Length capture bug since fixed). Paired with enterprise-ui spec change (toBeLessThan -> toBeLessThanOrEqual) and keploy/k8s-proxy#512 / keploy/enterprise-ui#1493. Signed-off-by: slayerjain <shubhamkjain@outlook.com>
charankamarapu
pushed a commit
that referenced
this pull request
Jul 2, 2026
…ration (0 → 2) (#101) * Add node dependency matrix sample * Add compose setup for node matrix sample * docs: tighten node dependency matrix k8s flow * Stabilize node dependency matrix recording flows * Document node dependency matrix staging runbook * Implement async catalog synchronization and deduplication endpoints in the node dependency matrix. Update documentation to reflect new features, including async workflows and enhanced deduplication for GET and POST requests. Adjust traffic recording scripts and improve error handling in the application logic. * Update Kubernetes traffic recording instructions and enhance scripts for better usability. Modify README and STAGING_RUNBOOK to clarify port-forwarding steps and add hints for using the hosted UI. Improve error handling in traffic recording scripts to guide users when using NodePorts. * Refactor Kafka and SQS scenario functions to improve error handling and response structure. Update Kafka scenario to include cluster and topic details, and enhance generic scenario with better socket management and timeout handling. * Enhance deploy-kind.sh to support custom node images for Kind clusters. Introduced a new KIND_NODE_IMAGE variable to allow users to specify a custom image when creating the cluster, improving flexibility in deployment configurations. * fix: update expected-values.json to match actual UI dedup behavior - expectedTestcasesExact: 23 → 18 (post-dedup count, UI shows unique) - expectedAdditionalTestcasesExact: 2 → 0 (gRPC not counted in UI total) - Dedup entries: use path-only format without query params to match how the Static dedup section groups entries in the UI - /dedup/catalog duplicateCount: 4 (all alpha+beta grouped) - /dedup/order duplicateCount: 3 (all orders grouped) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: update requiredKinds to include PostgresV2 and Redis The enterprise agent (v3.2.47) reliably captures Http, MySQL, PostgresV2, and Redis mock kinds. Add PostgresV2 and Redis to requiredKinds so the kube-regression tests assert all currently supported parsers. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: address Copilot review feedback for node-dependency-matrix - Route ERROR logs to stderr via console.error for proper log routing - Cache CA bundle read to avoid blocking the event loop on repeated calls - Ensure entrypoint.sh always creates the CA bundle file with a warning if no certs found - Align @types/express version with express v4 runtime dependency Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: reduce requiredKinds to Http only — mock kinds vary between runs Mock kinds captured by the sidecar are non-deterministic across runs (depends on pod timing and sidecar readiness). Only Http is guaranteed in every run. Other kinds are logged as informational via preferredKinds. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: sync package-lock.json with package.json npm ci fails in CI because @types/express and related packages were updated in package.json without regenerating the lock file. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: use ECR public mirrors to avoid Docker Hub rate limits Docker Hub 429 rate limits cause Kind cluster deployment failures in CI. Switch mysql, postgres, mongo, and redis images to public.ecr.aws/docker/library/ mirrors which have no rate limits. This matches the pattern used by api-server and enterprise-ui CI pipelines (playwright-setup.sh). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: use docker pull + kind load instead of crictl pull crictl pull goes directly to the registry (no cache), hitting rate limits. docker pull uses the Docker daemon cache (restored from MinIO by CI). kind load transfers cached images into the Kind node's containerd without any registry access. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: configure Kind with GCR mirror to avoid all registry rate limits Both Docker Hub and ECR public are rate-limited. Configure Kind's containerd to use mirror.gcr.io (Google's Docker Hub mirror) which has no rate limits. Reverted ECR image references back to docker.io since the GCR mirror handles them transparently inside Kind. This matches the approach used by k8s-proxy CI (setup-cluster.sh). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: use Keploy's local registry mirror (192.168.116.165:5000) All Keploy CI repos use a local Docker registry mirror at 192.168.116.165:5000 to avoid rate limits. Updated Kind containerd config to use this mirror (primary) with mirror.gcr.io as fallback. Matches the pattern in k8s-proxy/scripts/ci/setup-cluster.sh. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: load Kind images one at a time to avoid all-or-nothing failure kind load docker-image with multiple images fails entirely if one image has issues. Load each image individually so failures are isolated. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(node-dependency-matrix): allow Mongo TLS MITM hostname mismatch Signed-off-by: Asish Kumar <officialasishkumar@gmail.com> * fix(node-dependency-matrix): pin Mongo TLS to direct endpoint Signed-off-by: Asish Kumar <officialasishkumar@gmail.com> * fix(node-dependency-matrix): poll Mongo server monitor through proxy Signed-off-by: Asish Kumar <officialasishkumar@gmail.com> * fix(node-dependency-matrix): recalibrate expectedTestcasesExact 18 -> 19 (#100) The kube-regression dedup assertion (enterprise-ui kube-flow.spec.ts test 8) checks the recorded test-case count against this contract value. It was calibrated against the old keploy-agent (enterprise v3.2.74) the fixture chart pinned; k8s-proxy now injects v3.5.44 (distroless + shell-free probes), which records 19. Verified this is correct, not a dedup regression: static dedup still collapses every business duplicate exactly as the contract specifies (GET /dedup/catalog 5->1, POST /dedup/order 4->1; all 16 business endpoints recorded once). The recorded 19 = 16 deduped business endpoints + 3 non-business probe/preflight buckets: GET /health (x1) and GET /expectations (x2 — its request-header-name set differs between the curl preflight and the Playwright readiness probe, and the agent's static-dedup key includes request-header names). v3.2.74 recorded fewer only because it undercaptured that probe traffic (an upstream OSS chunked/no-Content-Length capture bug since fixed). Paired with enterprise-ui spec change (toBeLessThan -> toBeLessThanOrEqual) and keploy/k8s-proxy#512 / keploy/enterprise-ui#1493. Signed-off-by: slayerjain <shubhamkjain@outlook.com> * fix(node-dependency-matrix): count gRPC test cases in recording calibration k8s-proxy now records gRPC test cases in the recording session (previously dropped because processTestCase only counted entries with a non-empty HTTPReq.URL). The two calls in send_grpc_traffic.sh (Ping, RunDependencyScenario) each declare expectedAdditionalTestcases: 1, so the aggregate must reflect them. Bump grpcTrafficScript.expectedAdditionalTestcasesExact 0 -> 2 so getExpectedTotalTestcases() settles at 19 (HTTP) + 2 (gRPC) = 21, matching the recorded total once k8s-proxy counts gRPC cases. The kube-regression assertion is a <= ceiling, so this stays backward-compatible with released k8s-proxy (records 19; 19 <= 21). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Signed-off-by: Asish Kumar <officialasishkumar@gmail.com> Signed-off-by: slayerjain <shubhamkjain@outlook.com> Co-authored-by: Asish Kumar <officialasishkumar@gmail.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: Shubham Jain <shubhamkjain@outlook.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
The kube-regression dedup assertion (enterprise-ui
kube-flow.spec.tstest 8) readshttpTrafficScript.expectedTestcasesExactas the expected recorded test-case total. It was calibrated against the old keploy-agent (enterprise v3.2.74) that the fixture chart pinned. k8s-proxy #512 now injects v3.5.44 (distroless + shell-free probes), which records 19 → the< 18assertion fails.This is correct, not a dedup regression (verified)
Static dedup still collapses every business duplicate exactly as this contract specifies:
GET /dedup/catalog: 5 raw (alpha×3 + beta×2) → 1POST /dedup/order: 4 raw → 1The recorded 19 = 16 deduped business endpoints + 3 non-business probe/preflight buckets:
GET /health(×1) andGET /expectations(×2)./expectationssplits into two buckets because its request-header-name set differs between therecord_traffic.shcurl preflight and the Playwright readiness probe, and the agent's static-dedup key (enterprise/pkg/service/agent/staticDedup.goExtractOverallSchema) includes request-header names. v3.2.74 recorded fewer only because it undercaptured that probe traffic (an upstream OSS chunked/no-Content-Length capture bug, since fixed) — not because it deduped better.Change
expectedTestcasesExact: 18 → 19. Paired with the enterprise-ui spec relaxingtoBeLessThan→toBeLessThanOrEqualso it asserts the recorded total ≤ the calibrated count.Pairs with keploy/k8s-proxy#512, keploy/enterprise-ui#1493, and the enterprise-ui spec PR.
🤖 Generated with Claude Code