Files

T

jcoffey-dev 3eb0f4c589 Phase 4: SSO scaffolding, RBAC enforcement, tenant-scoped dashboards, audit logging, K8s deployment

RBAC (api/internal/authz) is live on /query and /dashboards, backed by a
new enterprise/ module (session issuance, audit logging, RBAC storage,
OIDC/SAML protocol wiring) that core never imports -- only calls over
HTTP. Found and fixed a real cross-tenant vulnerability in dashboards
(no tenant_id filtering at all) while writing the threat model doc.

Two things are explicitly NOT done, documented rather than hidden:
tenant isolation for log data itself (/query still shares one ClickHouse
connection and Tantivy index across every tenant -- RBAC controls who
can query, not what a query can see), and human SSO login (protocol
wiring exists, no HTTP handler calls it yet). See
docs/security/threat-model.md and docs/phase-4-runbook.md.

Also adds deploy/ (Go Operator + Helm chart, validated offline only --
no cluster was reachable in this environment).

2026-08-13 22:16:59 -07:00

3.1 KiB

Raw Blame History

deploy/operator

A small controller-runtime Operator managing one CRD: Tenant (sentry.io/v1alpha1). See internal/controller/tenant_controller.go's doc comment for exactly what it reconciles and -- just as importantly -- what it deliberately doesn't (no ClickHouse calls, no Tantivy filesystem access, no enterprise/internal/rbacstore wiring; those are enterprise/internal/tenantprovision, still unbuilt).

Not kubebuilder-scaffolded

No kubebuilder/controller-gen binary was available in this environment, so this package is hand-written rather than generated:

api/v1alpha1/zz_generated.deepcopy.go -- normally controller-gen object output; hand-written here, covered by api/v1alpha1/api_test.go's round-trip tests (mutate a copy, assert the original is untouched -- exactly the class of bug a hand-written DeepCopy is prone to).
config/crd/sentry.io_tenants.yaml -- normally controller-gen crd output from the +kubebuilder:validation:* markers on api/v1alpha1/tenant_types.go; hand-written here and only as strong as keeping the two in sync by hand. Validated by strict-unmarshaling it into the real k8s.io/apiextensions-apiserver Go type (see /deploy/README.md's verification section) -- catches YAML/structural mistakes, not a drift between the CRD's field descriptions and the Go doc comments.
+kubebuilder:rbac markers on internal/controller/tenant_controller.go are present as documentation/intent (matching kubebuilder convention) but were never run through controller-gen rbac -- the actual ClusterRole is hand-written in /deploy/helm/sentry/templates/tenant-operator.yaml, kept in sync with those markers by hand, same caveat as the CRD above.

Layout

api/v1alpha1/          Tenant, TenantSpec, TenantStatus -- the CRD's Go types
internal/controller/    TenantReconciler -- see its doc comment
cmd/tenant-operator/     main.go -- manager setup, matches every other
                         service's cmd/<name>/main.go convention in this repo
config/crd/               hand-written CRD YAML (see above)

Building & testing

go build ./...
go vet ./...
go test ./...

Tests use sigs.k8s.io/controller-runtime/pkg/client/fake, not envtest -- envtest needs a real kube-apiserver/etcd binary pair (setup-envtest) not available in this environment. The fake client exercises real reconcile logic (object CRUD, owner references, status writes) but not anything a real apiserver does for you (admission, garbage collection, watch-triggered re-reconciliation) -- see internal/controller/tenant_controller_test.go's doc comment.

docker build -f Dockerfile -t sentry-tenant-operator .   # context is deploy/operator/, not the repo root

Not verified in this session -- see /deploy/README.md.

Trying it against a real cluster

kubectl apply -f config/crd/sentry.io_tenants.yaml
kubectl apply -f - <<'EOF'
apiVersion: sentry.io/v1alpha1
kind: Tenant
metadata:
  name: acme
spec:
  displayName: "Acme Corp"
EOF
kubectl get tenant acme -o yaml   # status.phase should reach Active
kubectl get secret sentry-tenant-acme-clickhouse -o yaml

3.1 KiB Raw Blame History