chart-tests

作者: astronomer

在编写、编辑、审查或运行Astronomer APC仓库的Helm chart测试时使用。涵盖pytest模式、render_chart()用法、子chart…

npx skills add https://github.com/astronomer/astronomer --skill chart-tests

Chart Test Writing Guide

Critical Rules

  1. Always run tests with uv run — never python3 -m pytest or python bin/...
  2. Sub-chart values MUST be nested under the sub-chart name (see Values Nesting)
  3. No helm unittest plugin — all tests are pytest-based using render_chart()
  4. One render_chart() call per test function — never call it multiple times in the same function to cover different value combinations (see One Condition Per Test)

Test Organization

tests/
├── chart_tests/              # Helm template rendering tests (main focus)
│   ├── test_<component>.py   # One file per component
│   ├── conftest.py           # Shared fixtures
│   └── test_data/            # Feature configs, expected outputs
├── functional/               # End-to-end cluster tests
├── k8s_schema/               # Cached Kubernetes API schemas
└── utils/
    ├── chart.py              # render_chart() and helpers
    ├── fixtures.py           # Common fixtures
    └── __init__.py           # get_containers_by_name(), get_all_features(), etc.

Writing Tests

Basic Pattern

import pytest
from tests import supported_k8s_versions
from tests.utils.chart import render_chart

DEPLOYMENT_FILE = "charts/grafana/templates/grafana-deployment.yaml"


@pytest.mark.parametrize("kube_version", supported_k8s_versions)
def test_some_feature(kube_version):
    """Brief description of what is being tested."""
    docs = render_chart(
        kube_version=kube_version,
        show_only=[DEPLOYMENT_FILE],
        values={"grafana": {"enabled": True}},
    )
    assert len(docs) == 1
    assert docs[0]["kind"] == "Deployment"

Sub-Chart Values Nesting — CRITICAL

Templates in charts/<subchart>/templates/ belong to a sub-chart. Values for those templates must be nested under the sub-chart's top-level key:

# ❌ WRONG — will not override sub-chart values
values = {"houston": {"replicas": 3}}

# ✅ CORRECT — nest under the sub-chart name
values = {"astronomer": {"houston": {"replicas": 3}}}

# ✅ EXAMPLE — disable a feature in the astronomer sub-chart
docs = render_chart(
    values={"astronomer": {"dpLink": {"enabled": False}}},
    show_only=["charts/astronomer/templates/dp-link/dp-link-deployment.yaml"],
)
assert len(docs) == 0

Top-level charts (e.g. nginx, grafana, prometheus) use their chart name directly:

values = {"nginx": {"serviceType": "LoadBalancer"}}
values = {"grafana": {"extraEnvVars": [...]}}

Using show_only

Always use show_only to target the specific template being tested:

docs = render_chart(
    show_only=[
        "charts/nginx/templates/controlplane/nginx-cp-service.yaml",
        "charts/nginx/templates/dataplane/nginx-dp-service.yaml",
    ]
)

Parametrized Tests

Always parametrize over supported_k8s_versions and over relevant values axes:

@pytest.mark.parametrize("kube_version", supported_k8s_versions)
@pytest.mark.parametrize("plane_mode,docs_count", [("control", 1), ("unified", 1), ("data", 0)])
def test_deployment_should_render(kube_version, plane_mode, docs_count):
    docs = render_chart(
        kube_version=kube_version,
        show_only=[DEPLOYMENT_FILE],
        values={"global": {"plane": {"mode": plane_mode}}},
    )
    assert len(docs) == docs_count

One Condition Per Test

Never call render_chart() more than once inside the same test function to check several value combinations (e.g. default, feature-enabled, feature-disabled). Each render_chart() call is its own test condition — hiding several behind one function name makes failures ambiguous (which call failed?) and hides the test matrix from pytest --collect-only / -k filtering. Use parametrization when the calls only differ by a value/expectation, or separate test functions when they differ conceptually:

# ❌ WRONG — three renders buried in one function
def test_secretstore_rule(kube_version):
    docs = render_chart(kube_version=kube_version, show_only=[ROLE_FILE])
    assert absent(docs[0])

    docs = render_chart(kube_version=kube_version, values={"global": {"dataPlaneFailover": {"enabled": True}}}, show_only=[ROLE_FILE])
    assert present(docs[0])

    docs = render_chart(kube_version=kube_version, values={"global": {"dataPlaneFailover": {"enabled": False}}}, show_only=[ROLE_FILE])
    assert absent(docs[0])
# ✅ CORRECT — parametrize when only the value/expectation changes
@pytest.mark.parametrize("kube_version", supported_k8s_versions)
@pytest.mark.parametrize(
    "dataplane_failover_enabled,rule_expected",
    [(None, False), (True, True), (False, False)],
    ids=["default", "enabled", "disabled"],
)
def test_secretstore_rule(kube_version, dataplane_failover_enabled, rule_expected):
    values = {}
    if dataplane_failover_enabled is not None:
        values = {"global": {"dataPlaneFailover": {"enabled": dataplane_failover_enabled}}}
    docs = render_chart(kube_version=kube_version, values=values, show_only=[ROLE_FILE])
    assert (expected_rule in docs[0]["rules"]) == rule_expected

Or, if an existing test function already renders with the exact values you need (e.g. a test for dataPlaneFailover.enabled: True that targets a related template), prefer adding your assertion to that template's doc in the existing show_only list over writing a new render_chart() call.

Testing with All Features Enabled

get_all_features() enables as many compatible features as possible. Not all features can be enabled simultaneously due to incompatibilities.

from tests.utils import get_all_features

def test_with_all_features():
    docs = render_chart(values=get_all_features())
    kinds = [doc["kind"] for doc in docs]
    assert "Deployment" in kinds

Testing Probe Customization

Every container must support customizable livenessProbe and readinessProbe. When adding a new component:

  1. Add its probes to tests/chart_tests/test_data/enable_all_probes.yaml
  2. Run tests with that file to verify probes are rendered correctly

Cross-cutting invariants — ALWAYS guard with a cross-cutting test

Some requirements must hold for every container/pod/object the cluster renders, not just one component — typically because an admission controller (e.g. a Gatekeeper/OPA constraint) rejects the whole install if a single object violates them. Examples: every container must define a startupProbe (allow-with-probes, PINF-691), allowPrivilegeEscalation: false (PINF-585/713), no forbidden Service fields (PINF-692).

For requirements like these, a per-component test is not enough — a new component added later silently reintroduces the violation. Always add a single cross-cutting test that:

  1. Renders with get_all_features() (what a real install presents to the admission controller), and
  2. Iterates every container across every pod-manager kind and asserts the invariant, so any new component that violates it fails the suite automatically.

Use get_chart_containers() or get_containers_by_name() over the filtered doc list. test_probes.py (TestStartupProbes, TestCustomProbes) and test_security_context_override.py are the reference patterns. Write this test first (TDD): it should fail (red) listing every offending container before you implement the fix, then pass (green) once coverage is complete.


ConfigMap Scripts

Scripts embedded in ConfigMaps must follow these conventions:

  1. Static content only — scripts must not use Helm templating to conditionally modify their content based on chart values. The rendered output must be identical regardless of what values are passed.

  2. Environment variable inputs — all runtime configuration must be passed as environment variables defined in the container spec (via env or envFrom), not baked into the script at render time.

  3. Stored as files on disk — scripts must be committed as real files in the repository (e.g. under charts/<subchart>/files/) so they can be linted and reviewed like any other source file.

  4. Included via .Files.Get — scripts must be included in ConfigMap templates using .Files.Get, not inline Helm template blocks:

    # ✅ CORRECT
    apiVersion: v1
    kind: ConfigMap
    metadata:
      name: {{ include "chart.fullname" . }}-scripts
    data:
      my-script.sh: {{ .Files.Get "files/my-script.sh" | quote }}
    
    # ❌ WRONG — inline script with template logic
    data:
      my-script.sh: |
        #!/bin/sh
        {{- if .Values.someFlag }}
        do_something
        {{- end }}
    

Test Utilities

render_chart(values, show_only, kube_version, validate_objects)

Renders the chart via helm template and returns parsed YAML documents.

ParameterTypeDescription
valuesdictValues merged with chart defaults
show_onlylist[str]Templates to render (filters output)
kube_versionstrK8s version for schema validation
validate_objectsboolValidate against K8s schemas (default True)
from tests.utils.chart import render_chart

docs = render_chart(
    values={"nginx": {"enabled": True}},
    show_only=["charts/nginx/templates/controlplane/nginx-cp-service.yaml"],
    kube_version="1.31.0",
)

get_containers_by_name(doc, *, include_init_containers=False)

Returns {name: container_dict} for all containers in a pod manager doc (Deployment, StatefulSet, DaemonSet, Job, CronJob). Pass include_init_containers=True to also include init containers.

from tests.utils import get_containers_by_name

c_by_name = get_containers_by_name(doc, include_init_containers=True)
assert c_by_name["grafana"]["securityContext"] == {"readOnlyRootFilesystem": True}
assert c_by_name["bootstrapper"]["securityContext"] == {"readOnlyRootFilesystem": True}

get_all_features()

Returns a values dict with most components enabled.

from tests.utils import get_all_features

Other utilities in tests/utils/__init__.py

  • get_env_vars_dict(container_env) — converts env list to {name: value} dict
  • get_service_ports_by_name(doc) — returns service ports keyed by name
  • get_pod_template(doc) — extracts pod template from any pod manager
  • get_service_account_name_from_doc(doc) — returns the serviceAccountName
  • dot_notation_to_dict(dotted_string, default_value) — builds nested dict from dot notation

Running Tests

Correct examples

# Full suite in parallel (fastest — use for full runs)
uv run pytest tests/chart_tests/ -n auto --quiet

# Full suite, verbose
uv run pytest tests/chart_tests/ --verbose

# Single file
uv run pytest tests/chart_tests/test_grafana.py --verbose

# Tests matching a pattern
uv run pytest tests/chart_tests/ -k "test_service" --verbose

# Single test
uv run pytest tests/chart_tests/test_grafana.py::test_deployment_should_render --verbose

# Verbose output, stop on first failure
uv run pytest tests/chart_tests/ -vv --capture=no --maxfail=1

# Iterate on failures: re-run only last-failed tests
uv run pytest tests/chart_tests/ --maxfail=1 --lf

Tip: -n auto uses all CPU cores. Omit it when running a single file to avoid subprocess overhead.

Incorrect examples

# ❌ WRONG — we do not need to activate the venv manually, and we do not run pytest without `uv run`
.venv/bin/activate && pytest tests/chart_tests/

# ❌ WRONG — do not create virtual environment with python, do not use pip install. Use `uv run` to run all python files in the repo.
python3 -m venv .venv && .venv/bin/pip install -r requirements-dev.txt && .venv/bin/python3 -m pytest tests/chart_tests/

Kubernetes Schema Validation

Tests validate rendered manifests against cached K8s OpenAPI schemas in tests/k8s_schema/v<version>-standalone/. Validation runs by default; disable with validate_objects=False.

def test_custom_resource():
    docs = render_chart(
        show_only=["charts/airflow-operator/templates/crds/airflow.yaml"],
        validate_objects=True,
    )
    for doc in docs:
        assert doc["kind"] == "CustomResourceDefinition"

来自 astronomer 的更多技能

airflow
astronomer
查询、管理和排查Apache Airflow的DAG、运行记录、任务及系统配置。支持30多种命令,涵盖DAG检查、运行管理、任务日志、配置查询及直接REST API访问。通过持久化配置管理多个Airflow实例;自动发现本地和Astro部署。同步(等待完成)或异步触发DAG运行,诊断故障,清除运行记录以重试,并通过重试/映射索引过滤访问任务日志。输出...
official
airflow-hitl
astronomer
在Airflow DAG中使用可延迟操作符实现人工审批关卡、表单输入和分支。四种操作符类型:用于批准/拒绝决策的ApprovalOperator、带表单的多选项选择HITLOperator、人工驱动的任务路由HITLBranchOperator,以及表单数据收集HITLEntryOperator。所有操作符均为可延迟设计,在通过Airflow UI的"必需操作"标签页或REST API等待人工响应时释放工作槽位。支持包括自定义在内的可选功能...
official
airflow-state-store
astronomer
Persists task and asset state across retries and DAG runs using Airflow 3.3's AIP-103 key/value stores (`task_state_store`, `asset_state_store`) and the…
official
analyzing-data
astronomer
查询数据仓库,利用缓存的模式和概念映射来回答业务问题。支持对重复问题类型进行模式查找和缓存,并通过记录结果来改进后续查询。包含概念到表的映射缓存,以及通过INFORMATION_SCHEMA或代码库grep进行表结构发现。提供run_sql()和run_sql_pandas()内核函数,返回Polars或Pandas DataFrame用于分析。提供CLI命令用于管理概念、模式和表缓存,以及...
official
annotating-task-lineage
astronomer
使用入口和出口为Airflow任务标注数据血缘。支持使用OpenLineage Dataset对象、Airflow Assets和Airflow Datasets定义跨数据库、数据仓库及云存储的输入输出。当运算符缺少内置OpenLineage提取器时作为备用方案;遵循四级优先级系统,其中自定义提取器和OpenLineage方法优先。包含针对Snowflake、BigQuery、S3和PostgreSQL的数据集命名辅助工具,以确保一致性...
official
authoring-dags
astronomer
创建Apache Airflow DAG的引导式工作流,集成验证与测试。采用六阶段结构化方法:发现环境与现有模式、规划DAG结构、遵循最佳实践实现、通过af CLI命令验证、经用户同意测试、迭代修复。用于发现(af config connections、af config providers、af dags list)和验证(af dags errors、af dags get、af dags explore)的CLI命令可提供DAG的即时反馈...
official
authoring-go-sdk-tasks
astronomer
Writes Airflow task logic in Go using the Airflow Go SDK. Use when the user wants to implement Airflow tasks in Go, asks about `BundleProvider`/`RegisterDags`,…
official
authoring-java-sdk-tasks
astronomer
使用 Airflow Java SDK 编写 Java、Kotlin 或任何 JVM 语言的 Airflow 任务逻辑。当用户想要在 Java/JVM 中实现 Airflow 任务时使用,询问……
official