Mozilla Pipeline Schemas

This repository contains schemas for Mozilla's data ingestion pipeline and data lake outputs.

The JSON schemas are used to validate incoming submissions at ingestion time. They also are used as the source of truth for defining metadata about how each document type should be handled in the pipeline (see the metaschema).

The jsonschema [Python] and everit-org/json-schema [Java] library (using draft 4) are used for JSON Schema Validation in this repository's tests. This has implications for what kinds of string patterns are supported, see the Conformance section in the linked document for further details. Note that as of 2019, the data pipeline uses the everit-org/json-schema library for validation in production (see #302).

To learn more about writing JSON Schemas, Understanding JSON Schema is a great resource.

Adding a new schema

• Create the JSON Schema in the templates directory first. Make use of common schema components from the templates/include directory where possible, including things like the telemetry environment, clientId, application block, or UUID patterns. The filename should be templates/<namespace>/<doctype>/<doctype>.<version>.schema.json.
• Build the rendered schemas using the instructions below, and check those artifacts (in the schemas directory) in to the git repo as well. See the rationale for this in the "Notes" section below.
• Add one or more example JSON documents to the validation directory.
• Run the tests (either via Docker or directly) using the instructions below.
• Once all tests pass, submit a PR to the github repository against the main branch. See also the notes on contributions.

Note that Pioneer studies have a slightly amended process.

Build

Prerequisites

• CMake (3.0+)
• jq (1.5+)
• python (3.6+)
• Optional: java 11, maven
• Optional: Docker

On MacOS, these prerequisites can be installed using homebrew:

brew install cmake
brew install jq
brew install python
brew cask install docker

CMake Build Instructions

git clone https://github.com/mozilla-services/mozilla-pipeline-schemas.git
cd mozilla-pipeline-schemas
mkdir release
cd release

cmake ..  # this is the build process (the schemas are built with cmake templates)

Running Tests via Docker (optional)

You can generally skip this step if you're just make a small change to an existing schema: tests are automatically run via continuous integration.

The tests expect example pings to be in the validation/<namespace>/ subdirectory, with files named in the form <ping type>.<version>.<test name>.pass.json for documents expected to be valid, or <ping type>.<version>.<test name>.fail.json for documents expected to fail validation. The test name should match the pattern [0-9a-zA-Z_]+

To run the tests, make use of the wrapper scripts:

./scripts/mps-build
./scripts/mps-test

Packaging and integration tests (optional)

Follow the CMake Build Instructions above to update the schemas directory. To run the unit-tests, run the following commands:

# optional: activate a virtual environment with python3.6+
python3 -m venv venv
source venv/bin/activate

# install python dependencies, if they haven't already
pip install -r requirements-dev.txt
pip install .

# run the tests, with 8 parallel processes
pytest -n 8

# run tests for a specific namespace and doctype
pytest -k telemetry/main.4

# run java tests only (if Java is configured)
pytest -k java

To generate a diff of BigQuery schemas, use the mps command-line tool.

# optionally, enter the mozilla-pipeline-schemas environment
# for jsonschema-transpiler and python3 dependencies
./script/mps-shell

# generate an integration folder, the options will default to HEAD and main
# respectively
mps bigquery diff --base-ref main --head-ref HEAD

This generates an integration folder:

integration
├── bq_schema_f59ca95-d502688.diff
├── d502688
│   ├── activity-stream.events.1.bq
│   ├── activity-stream.impression-stats.1.bq
...
│   └── webpagetest.webpagetest-run.1.bq
└── f59ca95
    ├── activity-stream.events.1.bq
    ├── activity-stream.impression-stats.1.bq
    ...
    └── webpagetest.webpagetest-run.1.bq

Pushes to the main repo will trigger integration tests in CircleCI that directly compare the revision to the main branch. These tests do not run for forked PRs in order to protect data and credentials, but reviewers can trigger tests to run by pushing the PR's revisions to a branch of the main repo. We provide a script for this:

# Before running, double check that the PR doesn't make any changes to
# .circleci/config.yml that could spill sensitive environment variables
# or data contents to the public CircleCI logs.
./.github/push-to-trigger-integration <username>:<branchname>

For details on how to compare two arbitrary revisions, refer to the integration job in .circleci/config.yml. For more documentation, see mozilla-services/edge-validator.

mps command-line tool

The repository has an mps command-line tool for checking on the output of schema transformations used for BigQuery. Enter the shell using scripts/mps-shell.

To transpile a schema for Bigquery:

schema=schemas/glean/glean/glean.1.schema.json
mps bigquery transpile $schema

It may be useful to look at a compact version of the output:

schema=schemas/glean/glean/glean.1.schema.json
mps bigquery transpile $schema | mps bigquery columns /dev/stdin

The output of the ingestion sink can be viewed for validation documents.

validation=validation/glean/glean.1.baseline.pass.json
mps bigquery transform $validation | jq

Any value that is not captured in the schema is put into additional_properties.

validation=validation/glean/glean.1.baseline.pass.json
mps bigquery transform $validation | jq '.additional_properties'
"{\"$schema\":\"moz://mozilla.org/schemas/glean/ping/1\"}"

Releases

There is a daily series of tasks run by Airflow (see the probe_scraper DAG) that uses the main branch of this repository as input and ends up pushing final JSONSchema and BigQuery schema files to the generated-schemas branch. As of January 2020, deploying schema changes still requires manual intervention by a member of the Data Ops team, but you can generally expect schemas to be deployed to production BigQuery tables several times a week.

Contributions

• All non trivial contributions should start with a bug or issue being filed (if it is a new feature please propose your design/approach before doing any work as not all feature requests are accepted).
• If updating the glean schemas, be sure to update the changelog in include/glean/CHANGELOG.md.
• This repository is configured to auto-assign a reviewer on PR submission. If you do not receive a response within a few business days (or your request is urgent), please followup in the #fx-metrics slack channel.
• If updating schemas associated with certain restricted-access datasets (specified in CODEOWNERS), a CODEOWNER (usually SRE) will automatically be assigned to review the PR. Please follow additional change control procedures for PRs referencing these schemas. The CODEOWNER will be responsible for merging the PR once it has been approved.
• If your PR is associated with a bugzilla bug, please title it Bug XXX - Description of change, that way the Bugzilla PR Linker will automatically add an attachment with your PR to bugzilla, for future reference.

Notes

All schemas are generated from the 'templates' directory and written into the 'schemas' directory (i.e., the artifacts are generated/saved back into the repository) and validated against the draft 4 schema a copy of which resides in the 'tests' directory. The reason for this is twofold:

1. It lets us easily see and refer to complete schemas as they are actually used. This means that the schemas can be referenced directly in bugs and such, as well as being fetched directly from the repo for testing other schema consumers (test being important here, as any production use should be using the installable packages).
2. It gives us a changelog for each schema, rather than having to reason about changes to templated external pieces and when/how that impacted a given doctype's schema over time. This means that it should be easy to look back in time for the provenance of different parts of the schema for each doctype.

We have a number of scripts to keep the schemas in sync with various in-tree definitions. See the contents of the scripts subdirectory.