Files
cairnobs/deploy/helm/sentry
jcoffey-dev 1d57e697b1 Phase 4: real per-tenant ClickHouse isolation via a new enterprise-api binary
Closes the threat model's headline finding for the SQL query path:
enterprise/internal/tenantprovision does real CREATE DATABASE/USER/GRANT
against ClickHouse, and enterprise/internal/chrunner is a per-tenant
connection registry implementing api's SQLRunner interface, resolving
the tenant from the authenticated request identity -- never a
caller-suppliable parameter. Both are wired into a new binary,
enterprise/cmd/enterprise-api, alongside the unchanged single-tenant
api/cmd/api, since AGPL core can never import enterprise/ and Go's own
internal/ package visibility rules meant enterprise/ couldn't implement
core's SQLRunner interface without importing the package that defines
it. That required moving api/internal/{authz,queryapi,dashboards,
querylang/executor,searchclient,httpserver} out of internal/ -- the
minimal set enterprise-api needs to import; querylang's compiler
internals (planner/lexer/parser/ast/ir) and api's own config stay
internal, since nothing outside api needs them directly.

Also finally wires enterprise/internal/audit into queryapi.AuditLogger
(nil since Phase 4 task 4) via a new adapter, and adds live-ClickHouse
integration tests for two of the four adversarial probes named in
docs/phase-4-isolation-design.md's verification plan.

Corrected several overclaims in the docs while writing this up: an
earlier claim that rbacstore's CRUD was "verified against a live
Postgres" was never actually true in this environment (only
internal/audit was, earlier in this phase, before Docker access was
lost) -- threat-model.md, phase-4-runbook.md, CLAUDE.md, and
enterprise/README.md all now distinguish "a real integration test
exists" from "this was confirmed against a live database."

Still not built: Tantivy/free-text tenant isolation
(enterprise/internal/searchclient), and any deployment-topology
mechanism that actually routes traffic to enterprise-api instead of
plain api -- both binaries exist side by side today with nothing
enforcing or flagging which one a deployment runs.
2026-08-13 22:48:38 -07:00
..

deploy/helm/sentry

A Helm chart covering every docker-compose.yml service (Redpanda, ClickHouse, Postgres, ingest, search, api, alerting, web) plus, when enterprise.enabled: true: enterprise-auth, the deploy/operator tenant-operator, and Tenant CRs from values.tenants. See /deploy/README.md for what "multi-tenant-aware" does and doesn't mean at this layer, and its verification-status section before trusting this against a real cluster.

This chart never builds images -- push every image its values.yaml references to a registry the cluster can pull from first, same division of labor as docker compose build vs. docker compose up.

Startup ordering

docker-compose.yml uses depends_on: condition: service_healthy / service_completed_successfully to sequence startup (e.g. api waits for clickhouse-migrate to actually finish, not just for clickhouse to be reachable). This chart approximates that more loosely:

  • Migration Jobs (clickhouse-migrate, metadata-migrate, redpanda-provision) are plain Job resources (not Helm hooks -- making the StatefulSets they depend on into hooks too, to get ordering, would break helm upgrade/helm uninstall's normal ownership tracking of stateful resources, a worse tradeoff), with backoffLimit: 6 so they retry a few times if their dependency isn't up yet.
  • App Deployments get an initContainer that busy-waits for their dependency's TCP port, not for a specific Job's completion (see templates/_helpers.tpl's sentry.waitForTCP) -- this covers "is ClickHouse/Postgres/Redpanda up" but not "has the migration Job actually finished."
  • The gap that leaves (a pod starts before its migration has completed) is covered by every Go service here already calling os.Exit(1) on a failed startup DB ping (see e.g. api/cmd/api/main.go) -- Kubernetes' pod restart policy retries with backoff until the schema is ready. This is a real, working, but looser guarantee than docker-compose's explicit ordering -- documented here rather than implied to be equivalent.

Trying the two-tenant example

# Quote each --set value -- zsh globs an unquoted tenants[0] as a
# pattern and fails with "no matches found."
helm install sentry . --include-crds \
  --set enterprise.enabled=true \
  --set tenantOperator.enabled=true \
  --set 'tenants[0].name=acme' --set 'tenants[0].displayName=Acme Corp' \
  --set 'tenants[1].name=globex' --set 'tenants[1].displayName=Globex Corporation'

kubectl get tenants
kubectl get secret sentry-tenant-acme-clickhouse sentry-tenant-globex-clickhouse

This proves the K8s-side half of Phase 4's "two tenants... with their own users, roles, dashboards" exit criteria (/CLAUDE.md) -- a real per-tenant credential Secret exists for each. It does not by itself give either tenant a working login, dashboard, or ClickHouse database: those need the OIDC/SAML login handlers, internal/tenantprovision, and internal/rbacstore wiring the Phase 4 task 5 summary names as deferred.

web's image needs rebuilding per environment

web is a static SvelteKit build (adapter-static) -- its three API base URLs (VITE_API_BASE_URL/VITE_ALERTING_API_BASE_URL/ VITE_ENTERPRISE_AUTH_BASE_URL) are baked in at image build time (web/Dockerfile's build args), not read from the container's environment at runtime. values.yaml's web.builtWithApiBaseURL etc. document what the image you point web.image at needs to have been built with (an Ingress hostname, a LoadBalancer IP, etc.) -- this chart has no Ingress resources and can't itself act on those values; rebuild web's image with the right build args for wherever this release is actually reachable from a browser before pointing real users at it.

Validating without a cluster

helm lint .
helm template sentry . --include-crds > /tmp/rendered.yaml

See /deploy/README.md's verification section for what was actually checked this way (and what wasn't -- no live cluster was available).