2076540 - Don't validate Glean metrics files against the schema in automation builds

Status: ASSIGNED

(Data Platform and Tools :: Glean: SDK, task, P2)

Description

[Julien Cristau [:jcristau]]

Generating metrics_yamls.cached validates all the metrics.yaml files against the schema, which is most of the time spent parsing them (2.5s -> 0.7s locally when skipped). In artifact builds, it is the longest job of the pre-export tier. In automation, the glean-parser lint job already validates these files whenever they change (it's always-target with files-changed optimization), so the builds don't need to do it again.

This needs a skip_validation parser option in glean_parser (applied to the vendored copy in the patch, but it needs to go upstream), which run_glean_parser.parse() sets when MOZ_AUTOMATION is set. The parsed output is identical.

This relies on the lint job actually reporting validation errors, which it currently doesn't (bug 2076535).

Comment 1

[Julien Cristau [:jcristau]]

Attachment: WIP: Bug 2076540 - Don't validate Glean metrics files against the schema in automation builds

Validating all the metrics.yaml files against the schema is most of the
time spent parsing them, and makes metrics_yamls.cached the longest job
of the pre-export tier in artifact builds (5.2s on CI, 2.5s locally
versus 0.7s without validation). In automation, the glean-parser lint
job already validates them whenever they change, so skip it there.

On CI, metrics_yamls.cached now takes 2.8s.

This adds a skip_validation option to glean_parser, which needs to go
upstream.

Comment 2

[BMO Github Automation]

Attachment: [mozilla/glean_parser] Bug 2076540 - Add a skip_schema_validation option to bypass schema validation (#867)

(via GitHub webhook, authenticated as shared-secret)

Comment 3

[Jan-Erik Rediger [:janerik]]

badboy merged PR [mozilla/glean_parser]: Bug 2076540 - Add a skip_schema_validation option to bypass schema validation (#867) in 7fa1bf8.

This will need a moment to land in a release we can pull in here.