docs(design): unified Go/Ginkgo E2E test suite by xgerman · Pull Request #346 · documentdb/documentdb-kubernetes-operator

xgerman · 2026-04-20T16:54:28Z

Summary

Replaces the four legacy E2E workflows with a single Go + Ginkgo v2 + Gomega + mongo-driver test suite that reuses CloudNative-PG's tests/utils/ packages as a library.

This PR ships both the design rationale (docs/designs/e2e-test-suite.md) and the full implementation under test/e2e/, and removes the legacy workflows + bash/JS/Python glue. (Originally opened as a design-only draft; scope has since grown to cover implementation.)

What this replaces

.github/workflows/test-integration.yml
.github/workflows/test-E2E.yml
.github/workflows/test-backup-and-restore.yml
.github/workflows/test-upgrade-and-rollback.yml
.github/actions/setup-port-forwarding/
operator/src/scripts/test-scripts/{test-mongodb-connection.sh,test-python-pymongo.sh,mongo-python-data-pusher.py,comprehensive_mongosh_tests.js,performance_test.js}

Key design decisions

Go + Ginkgo over Python + pytest. A spike against CNPG's tests/utils/ found ~20 packages directly reusable (DocumentDB wraps apiv1.Cluster/apiv1.Backup — same types CNPG uses). A Python rewrite would have re-implemented all of it.
Per-area package layout under test/e2e/tests/{lifecycle,scale,data,performance,backup,tls,feature_gates,exposure,status,upgrade}/ — each area grows independently.
Fixture tiers to dedup heavy lifting:
- SharedROCluster (1 instance, session) for data/, performance/, status/ — saves ~24 min of cluster creation. Read-only contract enforced via proxy.
- SharedScaleCluster (2 instances, session) for scale/.
- FreshDocumentDB per-spec only for genuinely mutating tests.
- Minio lazy session fixture, imported as-is from CNPG.
Operator-health gate (BeforeEach/AfterEach) auto-skips non-disruptive/non-upgrade specs if the operator pod churns — adapted from CNPG's pattern, retargeted to the documentdb-operator namespace.
Small, single-purpose specs — comprehensive_mongosh_tests.js split into 7 files; performance_test.js into 6; upgrade workflow into 3.
Two-layer parallelism: ginkgo -p (process per package) + label-grouped GitHub Actions jobs. Performance + upgrade jobs forced serial.
Labels + TEST_DEPTH for cheap PR runs vs deep nightly runs from one tree.
test/e2e/ as a separate Go module so test deps (Ginkgo, CNPG utils, mongo-driver) don't pollute the operator's runtime deps.

Running locally

cd test/e2e
ginkgo -r --label-filter=smoke ./tests/...          # smoke
TEST_DEPTH=4 ginkgo -r --procs=4 ./tests/...        # full sweep

See test/e2e/README.md for the full env-var table and troubleshooting.

WentingWu666666 · 2026-04-20T17:30:08Z

Suggestion: Remove internal version references (v4/v5) from the design doc

The "v4" and "v5" labels are internal drafting artifacts that have no meaning to a new reviewer. Specific issues:

"Why Go (v5 reversed from v4)" a reader has to figure out what v4 was before understanding the heading.
"v5 changes vs v4 (what flipped)" section useful to the author during iteration, but noise for reviewers. The key reasoning is already well-covered in the Spike Findings section.
"v4" / "v5" scattered throughout adds cognitive load without adding value.

Suggested changes:

Instead of	Write
`Why Go (v5 reversed from v4)`	`Why Go over Python`
`v5 changes vs v4 (what flipped)` section	Fold the key points (language flip, spike justification) into the existing Spike Findings section, then remove the diff table
`v4` / `v5` references in body text	Describe the current design directly; mention the Python alternative was considered and rejected based on the spike

The spike findings section already tells the story clearly: "we investigated CNPG's test utils, found ~20 packages reusable, so we chose Go." That's all a reviewer needs. The iteration history belongs in commit messages, not the design doc.

xgerman · 2026-04-20T19:39:14Z

Thanks for the feedback @WentingWu666666 — applied in 100b583:

Why Go (v5 — reversed from v4) → Why Go over Python
Design principles (unchanged from v4) → Design principles
Spike findings (informed v5 decision) → Spike findings
Table header Our plan (v5) → Our plan
Deleted the entire v5 changes vs v4 (what flipped) section — the Spike Findings already carry the story.

No remaining v4/v5 references in the doc. Ready for further feedback.

Replaces the four legacy end-to-end workflows (test-integration.yml, test-E2E.yml, test-backup-and-restore.yml, test-upgrade-and-rollback.yml) and their bash/mongosh/pymongo glue with a single Go/Ginkgo v2/Gomega suite under test/e2e/. Highlights: - test/e2e/ Go module with pkg/e2eutils/ helpers (testenv, namespaces, portforward, fixtures, mongo, seed, assertions, timeouts, tlscerts) - specs organised by CRD area: lifecycle, scale, data, performance, backup, tls, feature-gates, exposure, status, upgrade - labels in test/e2e/labels.go (areas + smoke/basic/destructive/ disruptive/slow + needs-* capability labels) - TEST_DEPTH=0..4 (Highest..Lowest) selects spec density - fixtures pipeline: renderDocumentDB preprocesses templates with documentdb.DropEmptyVarLines before envsubst to tolerate empty vars (CNPG's envsubst treats empty as missing) - image architecture preserved: CNPG pg18 base + DocumentDB extension via image-library + independent gateway sidecar — harness leaves image env vars empty so operator composes layered defaults - upgrade/image spec requires explicit old/new env vars for both extension and (optionally) gateway — never conflates the two - lifecycle/update_image spec skips unless E2E_DOCUMENTDB_IMAGE_NEXT is a genuinely different tag from DOCUMENTDB_IMAGE (no silent no-op) - new .github/workflows/test-e2e.yml is the single PR-level gate - three release_*.yml workflows no longer reference the deleted legacy test workflows; dead run_tests inputs removed from release_operator.yml / release_images.yml Design: docs/designs/e2e-test-suite.md Runbook: test/e2e/README.md Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ager renewal Rationale from PR documentdb#346 gate: 'Ready=true' + a bare reconcile success alone can mask a broken gateway sidecar or a cert chain that only validates at issue-time. Add two minimal data-plane probes: 1. lifecycle/deploy: after the 1-instance cluster reaches Ready and owner-refs are verified, open a mongo-driver v2 client via the existing NewFromDocumentDB helper (port-forward + credentials) and Ping. Covers the 'the gateway actually answers' invariant during the most common bring-up path. 2. tls/cert-manager: after the initial CA-verified ping succeeds, trigger a reissue by deleting the cert-manager-generated Secret, wait for a new tls.crt to land, and reconnect with the freshly extracted CA. With the self-signed Issuer used here this exercises a full leaf+CA rotation, so a passing re-ping proves the gateway remounted the renewed material rather than serving a pinned copy. Both use mongohelper.Ping — the canonical 'hello' probe — so the existing connect retry + TLS verification paths are reused without introducing bespoke command builders. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Adds design doc for consolidating the four existing E2E workflows (test-integration, test-E2E, test-backup-and-restore, test-upgrade-and-rollback) into a single Ginkgo v2 + Gomega + go.mongodb.org/mongo-driver/v2 suite that reuses ~20 CNPG tests/utils packages. Key decisions for discussion: - Go + Ginkgo over Python + pytest (spike found large CNPG util reuse) - Per-area package layout under test/e2e/tests/ - Shared session cluster for read-only tests (data/perf/status) with per-spec Mongo DB isolation - Fresh DocumentDB per spec for mutating tests (lifecycle/tls/backup/upgrade) - Lazy MinIO session fixture, imported directly from CNPG - Operator-health gate auto-skips non-disruptive specs on operator churn - Label + imported-level (TEST_DEPTH) selection See doc for full fixture tiers, parallelism strategy, CI layout, and 28-item implementation plan. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Address review feedback: the design doc should read as a current-state proposal, not a diff log. Remove version labels from headings and drop the 'v5 changes vs v4' section since the Spike Findings section already tells the story. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Replaces the four legacy end-to-end workflows (test-integration.yml, test-E2E.yml, test-backup-and-restore.yml, test-upgrade-and-rollback.yml) and their bash/mongosh/pymongo glue with a single Go/Ginkgo v2/Gomega suite under test/e2e/. Highlights: - test/e2e/ Go module with pkg/e2eutils/ helpers (testenv, namespaces, portforward, fixtures, mongo, seed, assertions, timeouts, tlscerts) - specs organised by CRD area: lifecycle, scale, data, performance, backup, tls, feature-gates, exposure, status, upgrade - labels in test/e2e/labels.go (areas + smoke/basic/destructive/ disruptive/slow + needs-* capability labels) - TEST_DEPTH=0..4 (Highest..Lowest) selects spec density - fixtures pipeline: renderDocumentDB preprocesses templates with documentdb.DropEmptyVarLines before envsubst to tolerate empty vars (CNPG's envsubst treats empty as missing) - image architecture preserved: CNPG pg18 base + DocumentDB extension via image-library + independent gateway sidecar — harness leaves image env vars empty so operator composes layered defaults - upgrade/image spec requires explicit old/new env vars for both extension and (optionally) gateway — never conflates the two - lifecycle/update_image spec skips unless E2E_DOCUMENTDB_IMAGE_NEXT is a genuinely different tag from DOCUMENTDB_IMAGE (no silent no-op) - new .github/workflows/test-e2e.yml is the single PR-level gate - three release_*.yml workflows no longer reference the deleted legacy test workflows; dead run_tests inputs removed from release_operator.yml / release_images.yml Design: docs/designs/e2e-test-suite.md Runbook: test/e2e/README.md Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ager renewal Rationale from PR documentdb#346 gate: 'Ready=true' + a bare reconcile success alone can mask a broken gateway sidecar or a cert chain that only validates at issue-time. Add two minimal data-plane probes: 1. lifecycle/deploy: after the 1-instance cluster reaches Ready and owner-refs are verified, open a mongo-driver v2 client via the existing NewFromDocumentDB helper (port-forward + credentials) and Ping. Covers the 'the gateway actually answers' invariant during the most common bring-up path. 2. tls/cert-manager: after the initial CA-verified ping succeeds, trigger a reissue by deleting the cert-manager-generated Secret, wait for a new tls.crt to land, and reconnect with the freshly extracted CA. With the self-signed Issuer used here this exercises a full leaf+CA rotation, so a passing re-ping proves the gateway remounted the renewed material rather than serving a pinned copy. Both use mongohelper.Ping — the canonical 'hello' probe — so the existing connect retry + TLS verification paths are reused without introducing bespoke command builders. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…olish Addresses code-review-agent findings on 07fec09: - critical: shared renewCtx in the Eventually closure could exhaust the MongoConnect budget on the first iteration, starving later retries. Use per-iteration attemptCtx and widen the outer poll window to DocumentDBReady so cert-manager + gateway remount have real time to converge. - bytes.Clone replaces the ad-hoc make+append pattern for origCrt. - extra By(...) markers around reissue/reconnect make the failure narrative obvious. - clarifying comment on the explicit Ping in lifecycle/deploy documenting why we keep it despite NewFromDocumentDB pinging internally. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

ChangeStreams is an experimental DocumentDB feature that requires the -changestream image variant, which is not part of the default e2e image set. Running the wal_level-translation assertion on the default image implied support the shipped pipeline cannot honour, so this change: - removes tests/feature_gates/changestreams_test.go - removes manifests/mixins/feature_changestreams.yaml.template - removes the corresponding render check in pkg/e2eutils/fixtures/fixtures_test.go - leaves a TODO in feature_gates_suite_test.go and at the fixtures deletion site explaining status and the steps to re-enable (capability label + preflight + best-effort mongo Watch on top). Operator API (previewv1.FeatureGateChangeStreams) and the operator's wal_level=logical translation are untouched. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…e Ping Strengthen connection_string_test with a four-layer assertion: 1. Shape (existing AssertConnectionStringMatches helper). 2. Template: assert the secret name in the $(kubectl get secret …) userinfo matches the CR's credentials secret, catching drift between the status template and the actual secret. 3. Semantic: regex-split userinfo from the URI (url.Parse can't handle the shell subshell in userinfo) and assert host, port, and query params (tls, replicaSet, optional tlsAllowInvalidCertificates) against the published Service and status.TLS state. 4. Live: open a port-forward to the gateway Service and Ping via mongo.NewFromDocumentDB, proving the advertised string is not just well-formed but actually usable. Covers the failure mode where status.connectionString parses but points at the wrong host/port/params, which the previous shape-only check silently allowed. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: German <geeichbe@microsoft.com>

The documentdb org tracks issue priority via the single-select 'Priority' field on project boards documentdb#6 and documentdb#4 (options P0/P1/P2), not via repo labels. Record the project/field/option IDs and the gh CLI workflow so future assistants don't invent P0 labels that drift from the actual triage signal. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: German <geeichbe@microsoft.com>

Record the DCO requirement and the rebase-exec recipe for retrofitting Signed-off-by trailers, so future assistants don't ship commits that the DCO bot will block on the PR. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: German <geeichbe@microsoft.com>

Copilot

Pull request overview

This PR introduces (and documents) a unified Go/Ginkgo-based E2E test suite under test/e2e/, replacing the legacy multi-workflow + bash/JS/Python E2E approach across the repository’s CI and contributor docs.

Changes:

Add a new test/e2e/ Go/Ginkgo suite with per-area test packages (lifecycle/scale/data/performance/tls/status/exposure/feature-gates/upgrade), shared fixtures, and reusable helper utilities.
Add embedded and templated manifests (test/e2e/manifests) plus helper packages (timeouts, port-forwarding, TLS cert generation, namespaces, operator health gating, etc.).
Remove legacy integration workflow and python glue; update docs/changelog and adjust release workflows to assume the unified E2E workflow gate.

Reviewed changes

Copilot reviewed 115 out of 127 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
test/e2e/tests/upgrade/upgrade_suite_test.go	Upgrade area Ginkgo suite bootstrap/teardown.
test/e2e/tests/upgrade/upgrade_control_plane_test.go	Upgrade control-plane E2E scenario (helm uninstall/install/upgrade + data retention check).
test/e2e/tests/upgrade/rollback_test.go	Placeholder/pending rollback spec to document future intent.
test/e2e/tests/upgrade/helpers_test.go	Upgrade-area env gating and helper utilities (helm presence, manifests path, namespace/secret helpers).
test/e2e/tests/tls/tls_suite_test.go	TLS area Ginkgo suite + operator-unchanged gate.
test/e2e/tests/tls/tls_selfsigned_test.go	TLS self-signed mode E2E coverage.
test/e2e/tests/tls/tls_provided_test.go	TLS provided-secret mode E2E coverage (custom CA + hostname verification).
test/e2e/tests/tls/tls_disabled_test.go	TLS disabled mode E2E coverage.
test/e2e/tests/status/status_suite_test.go	Status area Ginkgo suite + operator-unchanged gate.
test/e2e/tests/status/pv_name_test.go	Status validation for PVC discovery via CNPG labels (CRD discrepancy handling).
test/e2e/tests/status/mount_options_test.go	Status validation for PGDATA mount presence on CNPG pods (CRD discrepancy handling).
test/e2e/tests/scale/scale_up_test.go	Scale-up scenarios using shared scale fixture + CNPG primary checks.
test/e2e/tests/scale/scale_suite_test.go	Scale area Ginkgo suite + operator-unchanged gate.
test/e2e/tests/scale/scale_down_test.go	Scale-down scenarios including primary re-election health assertion.
test/e2e/tests/performance/performance_suite_test.go	Performance area Ginkgo suite + operator-unchanged gate.
test/e2e/tests/performance/perf_update_test.go	Bulk update performance tripwire spec.
test/e2e/tests/performance/perf_sort_test.go	Indexed sort performance tripwire spec.
test/e2e/tests/performance/perf_insert_test.go	Bulk insert performance tripwire spec.
test/e2e/tests/performance/perf_helpers_test.go	Shared RO connection + per-spec DB + latency logging helpers for perf specs.
test/e2e/tests/performance/perf_delete_drop_test.go	Bulk delete + drop performance tripwire spec.
test/e2e/tests/performance/perf_count_range_test.go	Indexed range count performance tripwire spec.
test/e2e/tests/performance/perf_aggregation_test.go	Aggregation pipeline performance tripwire spec.
test/e2e/tests/lifecycle/update_storage_test.go	Lifecycle storage resize spec (with storageclass capability probe).
test/e2e/tests/lifecycle/update_loglevel_test.go	Lifecycle logLevel patch propagation spec.
test/e2e/tests/lifecycle/update_image_test.go	Lifecycle documentDBImage rollout spec.
test/e2e/tests/lifecycle/lifecycle_suite_test.go	Lifecycle area Ginkgo suite + operator-unchanged gate.
test/e2e/tests/lifecycle/helpers_test.go	Lifecycle helper utilities (vars, namespace/secret creation, refetch helper).
test/e2e/tests/lifecycle/deploy_test.go	Lifecycle deploy smoke + ownerRef check + mongo connectivity smoke.
test/e2e/tests/lifecycle/delete_reclaim_test.go	Lifecycle reclaim-policy Retain behavior validation.
test/e2e/tests/feature_gates/helpers_test.go	Feature-gates area helpers for fresh clusters + manifests wiring.
test/e2e/tests/feature_gates/feature_gates_suite_test.go	Feature-gates suite bootstrap + operator-unchanged gate + roadmap comments.
test/e2e/tests/exposure/loadbalancer_test.go	Exposure LoadBalancer spec with controller probe/skip behavior.
test/e2e/tests/exposure/helpers_test.go	Exposure area helpers for fresh clusters + vars/credentials wiring.
test/e2e/tests/exposure/exposure_suite_test.go	Exposure suite bootstrap + operator-unchanged gate.
test/e2e/tests/exposure/clusterip_test.go	Exposure ClusterIP spec (service type + port-forward ping).
test/e2e/tests/data/update_ops_test.go	Data-plane update operator coverage ($set/$inc/$unset/$push).
test/e2e/tests/data/sort_limit_skip_test.go	Data-plane sort/limit/skip coverage using deterministic dataset.
test/e2e/tests/data/query_test.go	Data-plane query filter coverage ($eq/$gt/$in/$and/$or/$regex).
test/e2e/tests/data/pipeline_test.go	Data-plane aggregation pipeline coverage ($lookup/$unwind/$group).
test/e2e/tests/data/helpers_test.go	Data area shared RO connect helper.
test/e2e/tests/data/delete_ops_test.go	Data-plane delete operator coverage (DeleteOne/DeleteMany).
test/e2e/tests/data/data_suite_test.go	Data suite bootstrap + operator-unchanged gate.
test/e2e/tests/data/crud_test.go	Basic CRUD coverage using shared RO fixture with per-spec DB.
test/e2e/tests/data/aggregation_test.go	Basic aggregation coverage using deterministic dataset.
test/e2e/suite_test.go	Root E2E suite bootstrap/teardown + logging (depth/run-id guidance).
test/e2e/runid_test.go	Unit tests for RunID env override + generation behavior.
test/e2e/runid.go	RunID implementation for fixture namespacing across processes/binaries.
test/e2e/pkg/e2eutils/tlscerts/tlscerts_test.go	Unit tests for generated TLS bundle correctness/validity.
test/e2e/pkg/e2eutils/tlscerts/tlscerts.go	Test-only CA/server cert generator for TLS “Provided” specs.
test/e2e/pkg/e2eutils/timeouts/timeouts_test.go	Unit tests asserting timeout coverage and poll interval sanity.
test/e2e/pkg/e2eutils/timeouts/timeouts.go	Central timeout registry with CNPG timeout aliasing + defaults.
test/e2e/pkg/e2eutils/timeouts/.keep	Keeps empty dir in VCS.
test/e2e/pkg/e2eutils/testenv/env_test.go	Unit tests for scheme registration + default constants.
test/e2e/pkg/e2eutils/testenv/env.go	CNPG TestingEnvironment constructor with DocumentDB/CNPG/client-go schemes + POSTGRES_IMG defaulting.
test/e2e/pkg/e2eutils/testenv/.keep	Keeps empty dir in VCS.
test/e2e/pkg/e2eutils/seed/datasets_test.go	Unit tests for deterministic dataset generators.
test/e2e/pkg/e2eutils/seed/datasets.go	Deterministic datasets for data/perf tests (Small/Medium/Sort/Agg).
test/e2e/pkg/e2eutils/seed/.keep	Keeps empty dir in VCS.
test/e2e/pkg/e2eutils/portforward/portforward_test.go	Unit tests for gateway service naming + nil-guard behavior.
test/e2e/pkg/e2eutils/portforward/portforward.go	CNPG forwardconnection wrapper specialized for DocumentDB gateway Service.
test/e2e/pkg/e2eutils/portforward/.keep	Keeps empty dir in VCS.
test/e2e/pkg/e2eutils/operatorhealth/gate_test.go	Unit tests for operator health gate drift detection + sentinel.
test/e2e/pkg/e2eutils/operatorhealth/.keep	Keeps empty dir in VCS.
test/e2e/pkg/e2eutils/namespaces/namespaces_test.go	Unit tests for deterministic DNS-safe per-spec namespace naming.
test/e2e/pkg/e2eutils/namespaces/namespaces.go	Deterministic per-spec namespace naming keyed by area/run-id/proc/spec hash.
test/e2e/pkg/e2eutils/mongo/.keep	Keeps empty dir in VCS.
test/e2e/pkg/e2eutils/helmop/helmop_test.go	Unit tests for helmop arg validation + deterministic flag ordering.
test/e2e/pkg/e2eutils/fixtures/teardown_test.go	Unit tests for fixture teardown selection by run-id + collision handling.
test/e2e/pkg/e2eutils/fixtures/shared_scale.go	Shared “scale” fixture creation/reset + CNPG readyInstances convergence checks.
test/e2e/pkg/e2eutils/fixtures/shared_ro.go	Shared “read-only CR” fixture creation and accessor-only handle.
test/e2e/pkg/e2eutils/fixtures/.keep	Keeps empty dir in VCS.
test/e2e/pkg/e2eutils/documentdb/.keep	Keeps empty dir in VCS.
test/e2e/pkg/e2eutils/assertions/.keep	Keeps empty dir in VCS.
test/e2e/manifests/mixins/tls_selfsigned.yaml.template	TLS SelfSigned manifest mixin template.
test/e2e/manifests/mixins/tls_provided.yaml.template	TLS Provided manifest mixin template.
test/e2e/manifests/mixins/tls_disabled.yaml.template	TLS Disabled manifest mixin template.
test/e2e/manifests/mixins/tls_certmanager.yaml.template	TLS CertManager manifest mixin template.
test/e2e/manifests/mixins/storage_custom.yaml.template	Storage override mixin template.
test/e2e/manifests/mixins/reclaim_retain.yaml.template	Retain reclaim policy mixin template.
test/e2e/manifests/mixins/exposure_loadbalancer.yaml.template	LoadBalancer exposure mixin template.
test/e2e/manifests/mixins/exposure_clusterip.yaml.template	ClusterIP exposure mixin template.
test/e2e/manifests/mixins/.keep	Keeps empty dir in VCS.
test/e2e/manifests/embed.go	Embeds manifest templates via `embed.FS`.
test/e2e/manifests/base/documentdb.yaml.template	Base DocumentDB CR template used across suites.
test/e2e/manifests/base/.keep	Keeps empty dir in VCS.
test/e2e/levels_test.go	Unit tests for TEST_DEPTH parsing/defaults + ordering semantics.
test/e2e/levels.go	Runtime depth gating (SkipUnlessLevel) and level label utilities.
test/e2e/labels.go	Area/cross-cutting/capability/level label definitions.
operator/src/scripts/test-scripts/mongo-python-data-pusher.py	Removed legacy python pymongo test helper script.
CHANGELOG.md	Adds release notes describing the unified E2E suite migration and contributor breaking change.
AGENTS.md	Adds contributor documentation on running the new E2E suite and labels/depth model.
.github/workflows/test-integration.yml	Removed legacy integration workflow (python-based).
.github/workflows/release_operator.yml	Removes pre-release legacy test gates; documents reliance on unified E2E workflow.
.github/workflows/release_images.yml	Removes pre-release legacy test gates; documents reliance on unified E2E workflow.
.github/workflows/release_documentdb_images.yml	Removes sed updates for deleted legacy workflows; updates comments accordingly.
.github/copilot-instructions.md	Adds repo process guidance (priority via projects, DCO sign-off requirement).

Copilot · 2026-04-22T21:03:32Z

+			_ = client.ObjectKeyFromObject(dd)
+		})


The final line _ = client.ObjectKeyFromObject(dd) appears to exist only to silence the unused controller-runtime/pkg/client import. This makes the spec harder to maintain and can hide real unused-import issues. Please remove the unused import and this no-op line (or use the client package for a real assertion/lookup if needed).

Fixed in b7803d5. Removed the no-op and dropped the now-unused controller-runtime client import.

Copilot · 2026-04-22T21:03:33Z

+			docs := seed.SortDataset()
+			any := make([]any, len(docs))
+			for i := range docs {
+				any[i] = docs[i]
+			}


any is used as a local variable name here (any := make([]any, len(docs))), which shadows Go's predeclared any identifier (alias for interface{}). This is legal but confusing and can lead to harder-to-read code; rename the slice variable to something like docsAny/payload to avoid shadowing.

Fixed in b7803d5. Renamed to docsAny to avoid shadowing the predeclared any.

Copilot · 2026-04-22T21:03:33Z

+// DocumentDB-specific operations. When adding an entry here, also
+// extend the switch in For and PollInterval.
+const (


The comment says "extend the switch in For" but For is implemented via cnpgAlias/documentDBDefaults maps (there is no switch). This is likely stale and can mislead future contributors; update the comment to reflect the current implementation (e.g., "extend documentDBDefaults/cnpgAlias and PollInterval").

Fixed in b7803d5. Updated the comment to point at documentDBDefaults/cnpgAlias.

Copilot · 2026-04-22T21:03:34Z

+### Running E2E tests
+
+The end-to-end suite lives in [`test/e2e/`](test/e2e/) as a separate Go module
+and replaces the legacy `test-integration.yml`, `test-E2E.yml`,
+`test-backup-and-restore.yml`, and `test-upgrade-and-rollback.yml` workflows
+(and their bash / JavaScript / Python glue). It is a Go / Ginkgo v2 / Gomega
+suite that drives the operator end-to-end and speaks the Mongo wire protocol
+via `go.mongodb.org/mongo-driver/v2`.
+
+**Prereqs:** kind + the DocumentDB operator already installed in the target
+cluster. In CI, the `.github/actions/setup-test-environment` composite action
+handles cluster creation and operator install (via `make deploy`). Locally,
+`operator/src/scripts/development/deploy.sh` is the equivalent entry point.


PR description says this is a draft design-doc-only proposal with "no implementation yet", but this change set adds a substantial Go/Ginkgo E2E implementation and removes legacy CI workflows/scripts. Please align the PR metadata/description with the actual scope (or split design vs implementation into separate PRs) so reviewers and release notes consumers have an accurate expectation.

Fixed in b7803d5. Acknowledged — the PR description is stale from an earlier design-only iteration; I'll refresh it. Scope is correctly implementation + docs + legacy workflow removal as reviewed.

- tests/exposure/clusterip: remove no-op client.ObjectKeyFromObject(dd) and drop the now-unused controller-runtime client import. - tests/data/sort_limit_skip: rename the local slice that shadowed the predeclared 'any' identifier to docsAny. - pkg/e2eutils/timeouts: refresh the 'when adding an op' comment to point at documentDBDefaults/cnpgAlias (there is no switch). - test/e2e/go.mod: bump go directive to 1.25.9 to match operator/src after rebase onto upstream/main. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: German <geeichbe@microsoft.com>

Ger I and others added 9 commits April 22, 2026 13:54

xgerman force-pushed the developer/e2e-suite branch from 6fc1356 to de201de Compare April 22, 2026 20:57

xgerman marked this pull request as ready for review April 22, 2026 20:58

xgerman requested a review from alaye-ms as a code owner April 22, 2026 20:58

Copilot AI review requested due to automatic review settings April 22, 2026 20:58

xgerman requested review from WentingWu666666 and hossain-rayhan as code owners April 22, 2026 20:58

Copilot started reviewing on behalf of xgerman April 22, 2026 20:58 View session

Copilot AI reviewed Apr 22, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

docs(design): unified Go/Ginkgo E2E test suite#346

docs(design): unified Go/Ginkgo E2E test suite#346
xgerman wants to merge 10 commits intodocumentdb:mainfrom
xgerman:developer/e2e-suite

xgerman commented Apr 20, 2026 •

edited

Loading

Uh oh!

WentingWu666666 commented Apr 20, 2026

Uh oh!

xgerman commented Apr 20, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 22, 2026

Uh oh!

xgerman Apr 22, 2026

Uh oh!

Copilot AI Apr 22, 2026

Uh oh!

xgerman Apr 22, 2026

Uh oh!

Copilot AI Apr 22, 2026

Uh oh!

xgerman Apr 22, 2026

Uh oh!

Copilot AI Apr 22, 2026

Uh oh!

xgerman Apr 22, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

xgerman commented Apr 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

What this replaces

Key design decisions

Running locally

Uh oh!

WentingWu666666 commented Apr 20, 2026

Uh oh!

xgerman commented Apr 20, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

xgerman Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

xgerman Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

xgerman Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

xgerman Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

xgerman commented Apr 20, 2026 •

edited

Loading