fix(deploy): manage Cloud Resource Manager API enablement in Pulumi (#1213)

## Summary

Follow-up to #1211 to encode the GCP API and IAM dependencies its
node-SA bindings rely on. After #1211 merged, the staging deploy hit two
undocumented prereqs in sequence — Cloud Resource Manager API not
enabled (because `projects.NewIAMMember` calls SetIamPolicy through
CRM), then the Pulumi service account missing
`roles/resourcemanager.projectIamAdmin`. Both fixed manually to unblock;
this PR makes them explicit so they don't bite again.

## Changes

**`ensureRequiredAPIs` adopts five GCP APIs** as Pulumi-managed
`projects.Service` resources, all with `DisableOnDestroy: false` and
`DisableDependentServices: false` so a Pulumi destroy/refactor can never
disable a shared API:

| API | Why |
|-----|-----|
| `cloudresourcemanager` | `projects.NewIAMMember` → SetIamPolicy |
| `compute` | `compute.GetDefaultServiceAccount` invoke + GKE |
| `container` | GKE cluster |
| `logging` | fluentbit-gke ships container logs |
| `monitoring` | managed Prometheus collector ships metrics |

CRM is created explicitly (not in the loop) so callers get a direct
reference for `pulumi.DependsOn`. Storage is intentionally **not**
Pulumi-managed — Pulumi state itself lives in a GCS bucket,
chicken-and-egg.

**`grantNodeServiceAccountRoles`** now `DependsOn` the CRM
`projects.Service` for the four IAM bindings, and uses
`compute.GetDefaultServiceAccount` to derive the SA email (Compute API
was already required for GKE — avoids a CRM-dependent project-number
lookup).

**`deploy/README.md`** — adds `roles/resourcemanager.projectIamAdmin` to
the Pulumi SA's required roles and clarifies which APIs must be enabled
before the first `pulumi up` (storage, cloudresourcemanager, container)
vs. which `ensureRequiredAPIs` adopts.

## Notes

`projects.Service` is idempotent against already-enabled APIs — on
existing projects (staging/prod), Pulumi adopts the live enablement into
state without re-enabling. This was confirmed live: staging deploy went
green after the manual unblocks above, validating the runtime model.

## Test plan

- [x] `go build ./...` clean for `deploy/`
- [x] `golangci-lint run` clean for `deploy/pkg/providers/gcp/...` (one
pre-existing `nilnil` at line 61, unrelated)
- [x] Manual unblock + green staging deploy verified the runtime
behavior this PR encodes
- [ ] On merge: staging deploy adopts the existing API enablements
without churn, then proceeds normally
- [ ] Prod deploy via #1212 (image promotion) inherits the same setup

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Radoslav Dimitrov 2026-04-27 12:32:22 -04:00 • committed by GitHub
commit c6d3096d1b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 97 additions and 8 deletions

View file

@ -65,8 +65,20 @@ Pre-requisites:
1. Create a project: `gcloud projects create mcp-registry-prod`
2. Set the project: `gcloud config set project mcp-registry-prod`
3. Enable required APIs: `gcloud services enable storage.googleapis.com && gcloud services enable container.googleapis.com`
4. Create a service account with necessary permissions, and get the key:
3. Enable the bootstrap APIs needed before Pulumi can run. The remaining APIs
(`cloudresourcemanager`, `compute`, `container`, `logging`, `monitoring`) are
adopted as Pulumi-managed `projects.Service` resources by `ensureRequiredAPIs`
in `pkg/providers/gcp/provider.go`, so they self-heal against drift but still
need to be reachable on the very first deploy:
```bash
gcloud services enable storage.googleapis.com # for the Pulumi state bucket below
gcloud services enable cloudresourcemanager.googleapis.com # for the IAM bindings Pulumi creates
gcloud services enable container.googleapis.com # for the GKE cluster
```
4. Create the Pulumi service account and grant it the roles required to manage
the deploy. `projectIamAdmin` is needed because the deploy creates IAM bindings
on the project (for the default compute SA). The other four are general
resource management:
```bash
gcloud iam service-accounts create pulumi-svc
sleep 10
@ -74,6 +86,7 @@ Pre-requisites:
gcloud projects add-iam-policy-binding mcp-registry-prod --member="serviceAccount:pulumi-svc@mcp-registry-prod.iam.gserviceaccount.com" --role="roles/compute.admin"
gcloud projects add-iam-policy-binding mcp-registry-prod --member="serviceAccount:pulumi-svc@mcp-registry-prod.iam.gserviceaccount.com" --role="roles/storage.admin"
gcloud projects add-iam-policy-binding mcp-registry-prod --member="serviceAccount:pulumi-svc@mcp-registry-prod.iam.gserviceaccount.com" --role="roles/storage.hmacKeyAdmin"
gcloud projects add-iam-policy-binding mcp-registry-prod --member="serviceAccount:pulumi-svc@mcp-registry-prod.iam.gserviceaccount.com" --role="roles/resourcemanager.projectIamAdmin"
gcloud iam service-accounts add-iam-policy-binding $(gcloud projects describe mcp-registry-prod --format="value(projectNumber)")-compute@developer.gserviceaccount.com --member="serviceAccount:pulumi-svc@mcp-registry-prod.iam.gserviceaccount.com" --role="roles/iam.serviceAccountUser"
gcloud iam service-accounts keys create sa-key.json --iam-account=pulumi-svc@mcp-registry-prod.iam.gserviceaccount.com
```

View file

@ -6,8 +6,8 @@ import (
"strings"
"github.com/pulumi/pulumi-gcp/sdk/v8/go/gcp"
"github.com/pulumi/pulumi-gcp/sdk/v8/go/gcp/compute"
"github.com/pulumi/pulumi-gcp/sdk/v8/go/gcp/container"
"github.com/pulumi/pulumi-gcp/sdk/v8/go/gcp/organizations"
"github.com/pulumi/pulumi-gcp/sdk/v8/go/gcp/projects"
"github.com/pulumi/pulumi-gcp/sdk/v8/go/gcp/storage"
"github.com/pulumi/pulumi-kubernetes/sdk/v4/go/kubernetes"
@ -61,10 +61,75 @@ func createGCPProvider(ctx *pulumi.Context, name string) (*gcp.Provider, error)
return nil, nil
}
// ensureRequiredAPIs adopts the GCP APIs the deploy depends on as Pulumi-managed
// services. Encoding them as Pulumi resources makes the dependency explicit and
// protects against drift (an org policy reset or accidental disable).
//
// Note: bootstrap APIs (storage, cloudresourcemanager, container — see
// deploy/README.md) must already be enabled before this runs, because the
// resources that need them are created earlier in the stack. For those, this
// function adopts them into Pulumi state after the fact. For the rest (logging,
// monitoring), the adoption happens before any consumer runs.
//
// DisableOnDestroy and DisableDependentServices are both false — a Pulumi destroy
// or refactor must never disable a shared API that other resources (and humans)
// rely on. Enable-only.
//
// Returns the Cloud Resource Manager API resource specifically, since the node-SA
// IAM bindings need to DependsOn it (SetIamPolicy is gated on CRM).
func ensureRequiredAPIs(ctx *pulumi.Context, projectID string, resourceOpts []pulumi.ResourceOption) (*projects.Service, error) {
// Service Usage API (which projects.NewService itself uses) is enabled by default
// on GCP projects, so we don't need to manage it here.
// CRM is created explicitly (not in the loop below) because callers need a
// direct reference to it for DependsOn — projects.NewIAMMember calls
// SetIamPolicy under the hood, which is gated on CRM.
crm, err := projects.NewService(ctx, "crm-api", &projects.ServiceArgs{
Project: pulumi.String(projectID),
Service: pulumi.String("cloudresourcemanager.googleapis.com"),
DisableOnDestroy: pulumi.Bool(false),
DisableDependentServices: pulumi.Bool(false),
}, resourceOpts...)
if err != nil {
return nil, fmt.Errorf("failed to ensure cloudresourcemanager.googleapis.com is enabled: %w", err)
}
otherAPIs := []struct {
resourceName string
serviceName string
}{
// Required for compute.GetDefaultServiceAccount and the GKE cluster.
{"compute-api", "compute.googleapis.com"},
// Required for the GKE cluster.
{"container-api", "container.googleapis.com"},
// Required for fluentbit-gke to ship container logs.
{"logging-api", "logging.googleapis.com"},
// Required for the managed Prometheus collector to ship metrics.
{"monitoring-api", "monitoring.googleapis.com"},
}
for _, api := range otherAPIs {
_, err := projects.NewService(ctx, api.resourceName, &projects.ServiceArgs{
Project: pulumi.String(projectID),
Service: pulumi.String(api.serviceName),
DisableOnDestroy: pulumi.Bool(false),
DisableDependentServices: pulumi.Bool(false),
}, resourceOpts...)
if err != nil {
return nil, fmt.Errorf("failed to ensure %s is enabled: %w", api.serviceName, err)
}
}
return crm, nil
}
// grantNodeServiceAccountRoles grants the default compute service account the standard
// GKE node roles required for log shipping (fluentbit-gke) and metrics scraping
// (managed Prometheus). New GCP projects no longer auto-grant Editor to the default
// compute SA, so these have to be set explicitly or every pod log/metric is dropped.
//
// The IAM bindings depend on Cloud Resource Manager API being enabled (managed by
// ensureRequiredAPIs above) — without it, projects.NewIAMMember fails because
// SetIamPolicy lives behind that API.
func grantNodeServiceAccountRoles(ctx *pulumi.Context, projectID string, gcpProvider *gcp.Provider) error {
invokeOpts := []pulumi.InvokeOption{}
resourceOpts := []pulumi.ResourceOption{}
@ -73,14 +138,22 @@ func grantNodeServiceAccountRoles(ctx *pulumi.Context, projectID string, gcpProv
resourceOpts = append(resourceOpts, pulumi.Provider(gcpProvider))
}
project, err := organizations.LookupProject(ctx, &organizations.LookupProjectArgs{
ProjectId: pulumi.StringRef(projectID),
}, invokeOpts...)
crmAPI, err := ensureRequiredAPIs(ctx, projectID, resourceOpts)
if err != nil {
return fmt.Errorf("failed to look up project number for node SA bindings: %w", err)
return err
}
nodeSA := fmt.Sprintf("serviceAccount:%s-compute@developer.gserviceaccount.com", project.Number)
// Get the default compute SA email directly from the Compute API instead of
// constructing it from the project number — avoids needing CRM API just for
// the project lookup, and matches whatever GCP actually returns.
defaultSA, err := compute.GetDefaultServiceAccount(ctx, &compute.GetDefaultServiceAccountArgs{
Project: pulumi.StringRef(projectID),
}, invokeOpts...)
if err != nil {
return fmt.Errorf("failed to look up default compute service account: %w", err)
}
nodeSA := "serviceAccount:" + defaultSA.Email
roles := []string{
"roles/logging.logWriter",
@ -89,13 +162,16 @@ func grantNodeServiceAccountRoles(ctx *pulumi.Context, projectID string, gcpProv
"roles/stackdriver.resourceMetadata.writer",
}
bindingOpts := make([]pulumi.ResourceOption, 0, len(resourceOpts)+1)
bindingOpts = append(bindingOpts, resourceOpts...)
bindingOpts = append(bindingOpts, pulumi.DependsOn([]pulumi.Resource{crmAPI}))
for _, role := range roles {
name := fmt.Sprintf("node-sa-%s", strings.ReplaceAll(strings.TrimPrefix(role, "roles/"), ".", "-"))
_, err := projects.NewIAMMember(ctx, name, &projects.IAMMemberArgs{
Project: pulumi.String(projectID),
Role: pulumi.String(role),
Member: pulumi.String(nodeSA),
}, resourceOpts...)
}, bindingOpts...)
if err != nil {
return fmt.Errorf("failed to grant %s to node SA: %w", role, err)
}