git.stella-ops.org/deploy

Deployment Profiles

This directory contains deterministic deployment bundles for the core Stella Ops stack. All manifests reference immutable image digests and map 1:1 to the release manifests stored under deploy/releases/.

Structure

• releases/ – canonical release manifests (edge, stable, airgap) used to source image digests.
• compose/ – Docker Compose bundles for dev/stage/airgap targets plus .env seed files.
• compose/docker-compose.mirror.yaml – managed mirror bundle for *.stella-ops.org with gateway cache and multi-tenant auth.
• compose/docker-compose.telemetry.yaml – optional OpenTelemetry collector overlay (mutual TLS, OTLP pipelines).
• compose/docker-compose.telemetry-storage.yaml – optional Prometheus/Tempo/Loki stack for observability backends.
• helm/stellaops/ – multi-profile Helm chart with values files for dev/stage/airgap.
• telemetry/ – shared OpenTelemetry collector configuration and certificate artefacts (generated via tooling).
• tools/validate-profiles.sh – helper that runs docker compose config and helm lint/template for every profile.

Workflow

1. Update or add a release manifest under releases/ with the new digests.
2. Mirror the digests into the Compose and Helm profiles that correspond to that channel.
3. Run deploy/tools/validate-profiles.sh (requires Docker CLI and Helm) to ensure the bundles lint and template cleanly.
4. If telemetry ingest is required for the release, generate development certificates using ./ops/devops/telemetry/generate_dev_tls.sh and run the collector smoke test with python ./ops/devops/telemetry/smoke_otel_collector.py to verify the OTLP endpoints.
5. Commit the change alongside any documentation updates (e.g. install guide cross-links).

Maintaining the digest linkage keeps offline/air-gapped installs reproducible and avoids tag drift between environments.

Surface.Env rollout warnings

• Compose (deploy/compose/env/*.env.example) and Helm (deploy/helm/stellaops/values-*.yaml) now seed SCANNER_SURFACE_* variables so the worker and web service resolve cache roots, Surface.FS endpoints, and secrets providers through StellaOps.Scanner.Surface.Env.
• During rollout, watch for structured log messages (and readiness output) prefixed with surface.env.—for example, surface.env.cache_root_missing, surface.env.endpoint_unreachable, or surface.env.secrets_provider_invalid.
• Treat these warnings as deployment blockers: update the endpoint/cache/secrets values or permissions before promoting the environment, otherwise workers will fail fast at startup.
• Air-gapped bundles default the secrets provider to file with /etc/stellaops/secrets; connected clusters default to kubernetes. Adjust the provider/root pair if your secrets manager differs.

Mongo2Go OpenSSL prerequisites

• Linux runners that execute Mongo2Go-backed suites (Excititor, Scheduler, Graph, etc.) must expose OpenSSL 1.1 (libcrypto.so.1.1, libssl.so.1.1). The canonical copies live under tests/native/openssl-1.1/linux-x64.
• Export LD_LIBRARY_PATH="$(git rev-parse --show-toplevel)/tests/native/openssl-1.1/linux-x64:${LD_LIBRARY_PATH:-}" before invoking dotnet test. Example:
  LD_LIBRARY_PATH="$(pwd)/tests/native/openssl-1.1/linux-x64" dotnet test src/Excititor/__Tests/StellaOps.Excititor.WebService.Tests/StellaOps.Excititor.WebService.Tests.csproj --nologo.
• CI agents or Dockerfiles that host these tests should either mount the directory into the container or copy the two .so files into a directory that is already on the runtime library path.

Additional tooling

• deploy/tools/check-channel-alignment.py – verifies that Helm/Compose profiles reference the exact images listed in a release manifest. Run it for each channel before promoting a release.
• ops/devops/telemetry/generate_dev_tls.sh – produces local CA/server/client certificates for Compose-based collector testing.
• ops/devops/telemetry/smoke_otel_collector.py – sends OTLP traffic and asserts the collector accepted traces, metrics, and logs.
• ops/devops/telemetry/package_offline_bundle.py – packages telemetry assets (config/Helm/Compose) into a signed tarball for air-gapped installs.
• docs/modules/devops/runbooks/deployment-upgrade.md – end-to-end instructions for upgrade, rollback, and channel promotion workflows (Helm + Compose).

CI smoke checks

The .gitea/workflows/build-test-deploy.yml pipeline includes a notify-smoke stage that validates scanner event propagation after staging deployments. Configure the following repository secrets (or environment-level secrets) so the job can connect to Redis and the Notify API:

• NOTIFY_SMOKE_REDIS_DSN – Redis connection string (redis://user:pass@host:port/db).
• NOTIFY_SMOKE_NOTIFY_BASEURL – Base URL for the staging Notify WebService (e.g. https://notify.stage.stella-ops.internal).
• NOTIFY_SMOKE_NOTIFY_TOKEN – OAuth bearer token (service account) with permission to read deliveries.
• NOTIFY_SMOKE_NOTIFY_TENANT – Tenant identifier used for the smoke validation requests.
• (Optional) NOTIFY_SMOKE_NOTIFY_TENANT_HEADER – Override for the tenant header name (defaults to X-StellaOps-Tenant).

Define the following repository variables (or secrets) to drive the assertions performed by the smoke check:

• NOTIFY_SMOKE_EXPECT_KINDS – Comma-separated event kinds the checker must observe (for example scanner.report.ready,scanner.scan.completed).
• NOTIFY_SMOKE_LOOKBACK_MINUTES – Time window (in minutes) used when scanning the Redis stream for recent events (for example 30).

All of the above values are required—the workflow fails fast with a descriptive error if any are missing or empty. Provide the variables at the organisation or repository scope before enabling the smoke stage.