* ci: adopt canonical hanzo.yml + hanzoai/ci reusable (arc, no GHA minutes)
Build/test/deploy now driven by a root hanzo.yml read by the reusable
hanzoai/ci workflow, imported via a ~7-line .github/workflows/cicd.yml
(uses: hanzoai/ci/.github/workflows/build.yml@main, secrets: inherit).
Runs on the self-hosted arc pool (hanzo-build-linux-amd64) — zero
GitHub-hosted minutes; GHCR push via the automatic workflow token.
Two self-building images (compile in-image, no CI pre-build step):
- collector -> cmd/o11yotelcollector/Dockerfile.selfbuild
ghcr.io/hanzoai/otel-collector (tag-suffix zap-native)
- schema-migrator -> cmd/o11yschemamigrator/Dockerfile.selfbuild (new)
ghcr.io/hanzoai/o11y-schema-migrator
* ci: remove broken/superseded workflows
Removed (all fail today):
- build-ghcr.yaml, build-staging.yaml, build.yaml: broken reusables
(hanzoai/.github docker-build.yml -> runtime-validation 'jobs: 0';
o11y/primus.workflows -> inaccessible upstream org).
- gor-signozotelcollector.yaml, gor-signozschemamigrator.yaml: goreleaser
on GitHub-hosted ubuntu/macos runners (violates arc-only) + primus.
- ci.yaml, postci.yaml: o11y/primus.workflows go-test/label; the go-test
gate is folded into hanzo.yml test: instead.
Kept workflow-sanity.yml (unrelated Hanzo lint gate). cicd.yml supersedes.
* docs: canonical CI pattern note (hanzo.yml + hanzoai/ci on arc)
* ci: remove workflow-sanity.yml (broken hanzoai/.github reusable, jobs:0)
Same runtime-validation failure as the docker-build.yml reusable; fails at
0s on every PR touching .github/workflows. cicd.yml is the one CI path now.
* ci: temporarily pin hanzoai/ci@fix/provision-parse-toolchain (pending #4)
hanzoai/ci@main fails on the stock arc runner (no PyYAML/jq); hanzoai/ci#4
provisions them. Pin the fix branch so this repo's arc build goes green now;
flip back to @main (canonical) once #4 merges.
* ci: gate the ZAP surface in the build stage (bare/DinD runner has no Go)
The reusable runs hanzo.yml test: on a bare arc runner with no Go toolchain
and a DinD daemon (bind-mounts don't cross), so 'go test' there fails with
'go: command not found'. Move the ZAP receiver+exporter test into
Dockerfile.selfbuild's build stage — where the toolchain + source already are
and a failing test fails the image — and drop the runner-based test:.
---------
Co-authored-by: hanzo-dev <[email protected]>
1.9 KiB
Hanzo O11y Otel Collector
The Hanzo distribution of the OpenTelemetry Collector. Ingests traces / metrics /
logs over ZAP-native (zap-proto/http, :4319) plus OTLP (:4317/:4318)
and writes to the one hanzoai/datastore (ClickHouse). Adds receiver/zapreceiver
exporter/zapexporterover upstream.
CI/CD — canonical pattern (one and only one way)
Build/test/deploy is driven by a single root hanzo.yml + a ~7-line
.github/workflows/cicd.yml that imports the reusable
hanzoai/ci workflow:
# .github/workflows/cicd.yml
jobs:
cicd:
uses: hanzoai/ci/.github/workflows/build.yml@main
secrets: inherit
All build/test/deploy logic lives once in hanzoai/ci. It reads hanzo.yml
(images: / test: / deploy: / kms:), builds + pushes each image, runs the
tests, and rolls the deployment. The build runs on our self-hosted arc pool
(hanzo-build-linux-amd64) — zero GitHub-hosted minutes; GitHub only
orchestrates, our infra does the work. GHCR push uses the automatic workflow
token; KMS (kms.hanzo.ai) supplies only the deploy KUBECONFIG.
This repo builds two images from self-building multi-stage Dockerfiles (so no CI pre-build step is needed):
| image | dockerfile | tags |
|---|---|---|
ghcr.io/hanzoai/otel-collector |
cmd/o11yotelcollector/Dockerfile.selfbuild |
:zap-native-latest, :sha-<short>-amd64-zap-native |
ghcr.io/hanzoai/o11y-schema-migrator |
cmd/o11yschemamigrator/Dockerfile.selfbuild |
:latest, :sha-<short>-amd64 |
This is the going-forward pattern for all native-PaaS repos. To onboard a new
repo: add a root hanzo.yml and the cicd.yml above — nothing else. Do not add
per-repo build logic, GitHub-hosted runners, or the legacy hanzoai/.github
docker-build.yml / o11y/primus.workflows reusables (both removed here).