Tape on GKE Autopilot⌗
tape provision gcp --target gke --store bigtable --apply
tape deploy gcp --target gke --image-tag 0.2
# wrote: deploy/gcp/release/values.generated.yaml
helm upgrade --install tape tape/deploy/gcp/k8s/chart/tape \
--namespace tape --create-namespace \
-f deploy/gcp/release/values.generated.yaml
The chart layout:
tape/deploy/gcp/k8s/chart/tape/
Chart.yaml
values.yaml
templates/
_helpers.tpl
serviceaccounts.yaml # WI-bound to tape-server/-reactor GSAs
server.yaml # Deployment + Service + HPA + PDB
reactors.yaml # one Deployment per enabled reactor
networkpolicy.yaml
Workload Identity⌗
Set workloadIdentity.serverGsa and workloadIdentity.reactorGsa to the GSA
emails created by the IAM Terraform module. The chart wires the KSAs with the
iam.gke.io/gcp-service-account annotation; you still need to grant
roles/iam.workloadIdentityUser on each GSA to the KSA member
(serviceAccount:PROJECT.svc.id.goog[tape/tape-server] and .../tape-reactor).
Kustomize overlays⌗
For the self-managed manifest in tape/deploy/k8s/tape.yaml, store-specific
overlays live in tape/deploy/gcp/k8s/overlays/{alloydb,bigtable,spanner}/.
They're useful if you'd rather not pull in Helm; Helm is the recommended path.
See also⌗
- Cloud Run (recommended) — the simpler topology.
- IAM cheat sheet — the SAs + roles you'll bind to KSAs via Workload Identity.
- Stores — picking the right backend.