e2e-testing

name	e2e-testing
description	KubeBlocks E2E testing infrastructure: kbcli-test.yml reusable workflow, kbcli-test-k8s.yml, cloud-e2e-*.yml, terraform-init, terraform-destroy, test-type, random-suffix. Use when adjusting kbcli-test parameters, adding E2E entry points, or changing cloud-e2e / terraform cluster lifecycle.

E2E Testing

This skill covers generic E2E testing infrastructure: the kbcli-test reusable workflow, cloud-e2e trigger chains, and terraform cluster create/destroy.

Related Files

Workflows

Workflow	Purpose
.github/workflows/kbcli-test.yml	Core reusable workflow: run kbcli tests on an existing K8s cluster
.github/workflows/kbcli-test-k8s.yml	kbcli test entry point that creates the K8s cluster first
.github/workflows/kbcli-pre-test-k8s.yml	K8s pre-test
.github/workflows/kbcli-test-pre.yml	kbcli pre-test preparation
.github/workflows/cloud-e2e-engine.yml	Engine E2E on cloud
.github/workflows/cloud-e2e-api.yml	API E2E on cloud
.github/workflows/cloud-e2e-web.yml	Web E2E on cloud
.github/workflows/cloud-e2e-installer.yml	Installer E2E on cloud
.github/workflows/cloud-e2e-schedule-dev.yml	Scheduled E2E for dev environment
.github/workflows/cloud-e2e-schedule-idc.yml	Scheduled E2E for IDC
.github/workflows/cloud-e2e-schedule-web.yml	Scheduled Web E2E
.github/workflows/terraform-init.yml	Create cloud K8s cluster (also in cloud-deploy)
.github/workflows/terraform-destroy.yml	Destroy cloud K8s cluster (also in cloud-deploy)

Utils

Script	Purpose
.github/utils/cloud_e2e_installer.sh	Install cloud E2E environment

kbcli-test.yml Key Inputs

• test-type: test type number, e.g. 0 (install), 2 (postgresql), 5 (redis).
• test-type-name: display name for job/reports.
• test-args: arguments passed to kbcli-test, commonly --service-version, --topology, --replicas.
• release-version: KubeBlocks version.
• k8s-cluster-name / artifact-name / random-suffix: cluster and artifact correlation.

Call Chain

kbcli-test-k8s.yml
   ├── terraform-init.yml
   └── kbcli-test.yml

kubeblocks-e2e-test.yml
   ├── terraform-init.yml
   ├── test-kubeblocks (test-type 0)
   ├── kbcli-test.yml (per engine/version)
   ├── send-message-* jobs
   └── terraform-destroy.yml

Common Tasks

1. Add a standalone E2E entry point

1. Decide whether a new K8s cluster is needed: reuse kbcli-test-k8s.yml if yes, otherwise reuse kbcli-test.yml.
2. Determine the test-type number (add in kbcli-test code if necessary).
3. Write a workflow that calls kbcli-test.yml with the correct test-args.
4. If scheduled, add a cloud-e2e-schedule-*.yml.

2. Modify K8s cluster creation logic

terraform-init.yml creates the cluster via terraform and outputs k8s-cluster-name. When modifying:

• Keep inputs consistent: cloud-provider, region, cluster-version, instance-type.
• terraform-destroy.yml is usually orchestrated to run with always().

3. Adjust kbcli-test arguments

test-args can be freely combined:

test-args: "--service-version 8.4.3 ${{ inputs.ARGS }}"
test-args: "--service-version 4.2 --topology distribution --replicas 3 ${{ inputs.ARGS }}"

4. Test result aggregation

utils.sh --type 36 aggregates test results (used in kubeblocks-e2e-test.yml send-message jobs). See kubeblocks-e2e-workflow for details.

Troubleshooting

Engine job never starts

• Check inputs.TEST_TYPE filtering. The if condition uses contains(..., '<engine>|') and endsWith(..., '<engine>'); for sub-variants (mongodb + mongodb-shard) both must be listed.
• Ensure the prerequisite wave completed successfully. A Wave 2 job needs Wave 1 jobs in needs.

kbcli-test.yml fails with missing inputs

• Any required input added to kbcli-test.yml must also be added to all callers, including kubeblocks-e2e-test.yml, kbcli-test-k8s.yml, and cloud-e2e-*.yml.

Test result is empty in send-message job

• Confirm the engine job ID in the send-message needs matches the actual test job ID.
• Verify the engine job actually sets outputs.test-result and outputs.test-result-report.
• Check that utils.sh --type 36 receives the expected TEST_RESULT format.

Cluster creation fails in terraform-init.yml

• Validate CLOUD_PROVIDER, REGION, CLUSTER_VERSION, and INSTANCE_TYPE inputs.
• Ensure cloud credentials and terraform state backend are configured for the target provider.

terraform-destroy-k8s runs before send-message

• Add the missing send-message-* job to terraform-destroy-k8s.needs.
• The destroy job uses if: always(), so it will run as soon as all explicit needs complete, including failed jobs.

Related Skills

• kubeblocks-e2e-workflow — maintain the large engine matrix in kubeblocks-e2e-test.yml
• cloud-deploy — terraform cluster lifecycle
• utils-common — full utils.sh usage