Skip to content

Commit 7ee101b

Browse files
diamonwigginsdiamonwiggins
authored
Mlflow Release Pipeline (#44)
* add release pipeline for mlflow fix app slug fix helm setup action fix helm version add ct.yaml bump upload-artifact version fix kubeconfig retrieval bump chart testing version fix typo ci improvements and adding kots install remove conditional from create release pass license file as content pass license as content modify app slug for unstable channel fix slug move some steps to make targets add replicated helm install to pipeline fix chart testing setup fix makefile debug ci debug ci fix ci fix ci use jq to parse customer json debug ci try getting oci working with chart testing fix ct install kubeconfig offset channel name by run number in case of re-runs add more distros revert adding distros cleanup add test for mlflow fix mlflow test domain fix chart file path fix helm install stop using chart-testing fix helm install enable application tests debug remove make make sure helm dependencies get updated fix helm install bump timeout on test script and increase cluster size use kubectl port forward to access app fix port forward wait before port-forward fix authentication in test script fix test add application testing for kots install improve wait for mlfow service before starting test moving to taskfile moving to taskfile commit taskfile fix taskfile remove check deps fix get licenseid fix get licenseid fix get licenseid fix get licenseid fix get licenseid fix port forward fix port forward add missing dependency for tests refactor taskfile polish readme * add aks and gke to test coverage * remove instance type from create cluster * remove instance type from create cluster * set nodecount to 3 and use k3s instead of kind * remove aks * improve docs * improve docs * readme updates * break down port forward task * use requirements * use REPLICATED_APP env var * remove cache dependency path * remove cache from pythong setup * revert using requirements * revert using requirements * fix dependency install * fix dependency install
1 parent 8b3d469 commit 7ee101b

14 files changed

+3903
-91
lines changed

‎.github/workflows/mlflow-ci.yml

Lines changed: 664 additions & 0 deletions
Large diffs are not rendered by default.

‎applications/mlflow/DEVELOPMENT.md

Lines changed: 215 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,215 @@
1+
# MLflow Development Guide
2+
3+
This document provides information about developing, testing, and releasing the MLflow application using the included Taskfile.
4+
5+
## Development Workflow
6+
7+
The MLflow application includes a Taskfile.yml that provides tasks for developing, testing, and publishing the application.
8+
9+
### Prerequisites
10+
11+
- [Task](https://taskfile.dev/#/installation) command line tool
12+
- Kubernetes cluster configured in your current context
13+
- kubectl, helm, and python3 installed
14+
15+
### Local Development
16+
17+
Follow this workflow for development:
18+
19+
1. Add required Helm repositories and update dependencies:
20+
```bash
21+
task add:repos:helm
22+
task update:deps:helm
23+
```
24+
25+
2. Lint charts to check for issues:
26+
```bash
27+
task lint
28+
```
29+
30+
3. Template charts to verify the rendered manifests:
31+
```bash
32+
task template
33+
```
34+
35+
4. Install charts for development:
36+
```bash
37+
# Installs with Replicated SDK disabled
38+
task install:helm:local
39+
40+
# Optionally specify a custom values file
41+
MLFLOW_VALUES=./my-values.yaml task install:helm:local
42+
```
43+
44+
> **Note:** For local development, the Replicated SDK is explicitly disabled (`replicated.enabled=false`). This allows development without requiring access to the Replicated platform.
45+
>
46+
> This task automatically sets up port forwarding from localhost:5000 to the MLflow service in the cluster, making the application available for testing.
47+
>
48+
> The Helm releases are created with names `infra` and `mlflow` in the `mlflow` namespace.
49+
50+
5. Run application tests:
51+
```bash
52+
task run:tests:app
53+
```
54+
55+
6. Make changes to your charts and repeat steps 2-5 as needed
56+
57+
This workflow allows rapid iteration without needing to publish to the Replicated registry.
58+
59+
### Task Reference
60+
61+
Tasks follow a `verb:resource[:subresource]` naming convention for clarity:
62+
63+
```bash
64+
# Validation and verification
65+
task lint # Lint Helm charts
66+
task template # Render templates to stdout (SDK disabled)
67+
task check:versions # Verify Chart.yaml and KOTS manifest versions match
68+
69+
# Repository and dependency management
70+
task add:repos:helm # Add required Helm repositories
71+
task update:deps:helm # Update Helm chart dependencies
72+
73+
# Packaging and versioning
74+
task update:versions:chart # Update chart version refs in KOTS manifests
75+
task package:charts # Package Helm charts for distribution
76+
task extract:version:chart # Extract current MLflow chart version
77+
78+
# Installation
79+
task install:helm:local # Install charts for local development (SDK disabled)
80+
81+
# Testing
82+
task test:install:helm # Test with charts from Replicated registry
83+
task test:install:kots # Test KOTS installation
84+
task run:tests:app # Run application tests against running MLflow
85+
task run:tests:all # Run all tests (Helm install + app tests)
86+
87+
# Release management
88+
task create:release # Create a Replicated release
89+
90+
# Cleanup
91+
task clean:files:charts # Clean packaged chart files
92+
task clean:all # Clean all generated files
93+
```
94+
95+
## Releasing
96+
97+
### Updating Documentation
98+
99+
Before creating a release, ensure the documentation is up-to-date:
100+
101+
1. Update version information in `charts/mlflow/Chart.yaml` if needed.
102+
103+
2. Update the changelog in `charts/mlflow/README_CHANGELOG.md.gotmpl` with details about the new release.
104+
105+
3. Generate documentation using helm-docs:
106+
```bash
107+
# From the mlflow chart directory
108+
cd charts/mlflow
109+
110+
# If helm-docs is installed locally
111+
helm-docs
112+
113+
# Or use Docker
114+
docker run --rm -v "$(pwd):/helm-docs" -u $(id -u) jnorwood/helm-docs:latest
115+
```
116+
117+
4. Verify the generated documentation:
118+
- `README.md` - Main chart documentation
119+
- `README_CHANGELOG.md` - Changelog
120+
- `README_CONFIG.md` - Configuration reference
121+
122+
### Publishing Replicated Releases
123+
124+
When you're ready to publish your changes to the Replicated platform:
125+
126+
1. Update the version in `charts/mlflow/Chart.yaml` if necessary.
127+
128+
2. Update documentation:
129+
```bash
130+
# If helm-docs is not installed
131+
cd charts/mlflow
132+
docker run --rm -v "$(pwd):/helm-docs" -u $(id -u) jnorwood/helm-docs:latest
133+
```
134+
135+
3. Set up the required environment variables:
136+
```bash
137+
# Replicated API token for authentication
138+
export REPLICATED_API_TOKEN=your_api_token
139+
140+
# App and channel to publish to
141+
export REPLICATED_APP=app_slug
142+
export REPLICATED_CHANNEL=channel_name
143+
```
144+
145+
4. Package the charts and update version references:
146+
```bash
147+
# This updates KOTS manifests with the current chart versions
148+
# and packages the charts as .tgz files
149+
task package:charts
150+
```
151+
152+
5. Create a release in Replicated:
153+
```bash
154+
# This uploads the packaged charts and creates a new release
155+
task create:release
156+
```
157+
158+
6. Verify the release was created successfully in the Replicated vendor portal
159+
160+
### Testing Replicated Releases
161+
162+
This workflow tests the full Replicated release and distribution process:
163+
164+
1. After publishing a release, login to the registry with a license ID:
165+
```bash
166+
# Set license ID for registry authentication
167+
export REPLICATED_LICENSE_ID=your_license_id
168+
export REPLICATED_APP=app_slug
169+
export REPLICATED_CHANNEL=channel_name
170+
171+
# Login to the registry
172+
task login:registry
173+
```
174+
175+
2. Test the Helm installation from the Replicated registry:
176+
```bash
177+
# This pulls charts from the Replicated registry with SDK enabled
178+
task test:install:helm
179+
```
180+
181+
> **Note:** This creates Helm releases named `infra` and `mlflow` in the `mlflow` namespace.
182+
183+
3. Verify the installation with application tests:
184+
```bash
185+
task run:tests:app
186+
```
187+
188+
You can also run the complete test suite after setting up environment variables:
189+
```bash
190+
task run:tests:all
191+
```
192+
193+
This workflow validates the entire release pipeline from publishing to installation, ensuring that your charts work correctly when distributed through the Replicated platform.
194+
195+
## CI/CD Pipeline
196+
197+
This application includes a CI/CD pipeline implemented with GitHub Actions. The pipeline handles:
198+
199+
- Linting and validating Helm chart templates
200+
- Creating releases in Replicated
201+
- Testing Helm installation with charts from the Replicated registry
202+
- Installing the application via KOTS
203+
204+
The pipeline workflow:
205+
1. `lint-and-template`: Validates chart syntax and templates (SDK disabled)
206+
2. `create-release`: Packages charts and creates a release in Replicated
207+
3. `helm-install-test`: Tests Helm installation with charts from Replicated registry (SDK enabled)
208+
4. `kots-install-test`: Tests KOTS installation
209+
5. `cleanup-test-release`: Cleans up test resources
210+
211+
The pipeline is triggered on:
212+
- Pull requests affecting the MLflow application
213+
- Pushes to the main branch
214+
215+
For more details, see the workflow definition in [.github/workflows/mlflow-ci.yml](../../.github/workflows/mlflow-ci.yml).

‎applications/mlflow/Makefile

Lines changed: 0 additions & 57 deletions
This file was deleted.

0 commit comments

Comments
 (0)