Payload CMS on GKE Autopilot

Payload CMS is a TypeScript-native, code-first headless CMS and application framework built directly on Next.js — not a hosted SaaS product, but a library installed into your own Next.js application. Content is modeled through typed "Collections" defined in payload.config.ts, and Payload generates an admin UI plus REST, GraphQL, and Local APIs from that same config. This module deploys a real Payload application on GKE Autopilot on top of the App_GKE foundation, which provisions and manages the shared Google Cloud and Kubernetes infrastructure.

This guide focuses on the cloud services this deployment uses and how to explore and operate them from the Google Cloud Console and the command line. For the mechanics that are common to every GKE application — Workload Identity, ingress, autoscaling, CI/CD, Cloud Armor, IAP, Binary Authorization, VPC Service Controls, backups, and the deployment lifecycle — refer to the App_GKE foundation guide rather than repeating them here.

---

What Payload costs on RAD, and how that compares

For a small team, Payload on RAD's GKE Autopilot module costs about US$107 a month with no per-user licence. GKE suits a Payload application that must stay up, scale across pods, or run beside other Kubernetes workloads. For the lowest cost, the Cloud Run module runs the same Payload for about US$84 a month. Figures are as at 8 October 2026; sources are listed at the end of this section.

If you only need it occasionally — studying for a certification, a demo, a seasonal business — a RAD-managed project can be deleted and restored within 30 days for a few credits, so that US$107 becomes a few dollars a month instead. See Pause it for free, below.

What you pay on RAD

	In a Google Cloud project you own	In a project RAD manages for you
Module fee, once per deployment	175 credits (US$17.50 at the top-up price)	157.5 credits (10% lower)
Build time	About 19 credits per build (RAD's average build takes about 19 minutes)	The same
Google Cloud running cost	Billed by Google to your own billing account (table below)	Metered hourly in credits; RAD publishes 92 credits a day, about 2,760 credits a month (about US$276 at the top-up price, US$221 at the Scale plan's rate)
Who owns what	You own the project and its billing; RAD deploys and updates it	RAD owns the project, with guardrails, quotas and budget alerts; you get console access, and a minimum purchased balance is held in reserve

10 credits cost US$1 on a one-off top-up, and less on a monthly plan. In a project RAD manages, the database, file server and network are shared by every application in the project, so a second application does not add a second database.

Default running cost in your own project (us-central1, Google list prices):

Resource (module default)	Per month
GKE Autopilot pod, 1x vCPU / 1 GiB	US$36.08
Cloud SQL for PostgreSQL, 1 vCPU / 3.75 GB, zonal, 10 GB SSD	US$51.02
NFS/cache file server (small VM)	US$13.43
Cloud NAT and networking	US$5
Cloud Storage (add-ons, backups)	US$1
GKE cluster management fee	shared across every GKE app in the project; $0 if this is your only cluster, else ~73
Total	about US$107, or about US$180 without the free tier

• GKE's free tier covers the management fee of one Autopilot or zonal cluster per billing account, so it applies only if this is your only such cluster. Cloud SQL costs the same in africa-south1 (Johannesburg); the pod and VM cost about 10% more there.
• GKE Autopilot does not scale to zero. At least one Payload pod always runs, so the cost above is the same whether or not anyone is using it; extra pods are added only under load.
• Avoid Filestore for a small deployment. The module uses a small NFS/cache VM by default. Filestore's smallest instance is 1 TiB, which costs about US$164 a month on its own.

How it compares

• Payload Cloud, the project's own hosted offering, does not publish a current public price list, so this compares against a bare server instead.
• A bare VPS — Hetzner CPX22 or a DigitalOcean 2 vCPU/4 GB Droplet, both about US$24/month — undercuts RAD's own-project figure above. Payload only needs 1 vCPU / 1–2 GiB, so a smaller/cheaper VPS tier would likely do; the figures above are the smallest independently verified class.
• The difference is the database and secrets handling: Cloud SQL backups and point-in-time recovery, and Secret Manager, are included on RAD rather than being your own job on a VPS.

Pause it for free: delete a RAD-managed project, restore it when you need it

If Payload runs in a project RAD manages for you, you have a second option that goes well beyond scaling to zero: delete the whole project, and restore it within 30 days for close to nothing. This suits occasional use — studying for a certification, a demo environment, a seasonal business — far better than running Payload continuously.

• How it works. Deleting a RAD-managed project unlinks its billing first, then asks Google to delete the project. Google does not remove the project immediately: it keeps it, recoverable, for 30 days. Because billing is already unlinked, nothing is charged while it waits. Unlike deleting one module, this does not tear down Cloud SQL, any VM or the GKE cluster one by one — the whole project, and everything in it, simply stops.
• Restoring costs a handful of credits, not a rebuild. Within 30 days, the project's owner can restore it. RAD asks Google to undelete the project and reattaches its billing account, then asks you to run Update on each deployment to confirm everything came back. Because nothing was individually destroyed, that Update finds the same resources already there — it is a check, not a rebuild, and an Update never charges the module fee again. That costs a handful of credits (under US$1) for a typical 2–3-deployment chain, against the 180 credits (US$18) a full redeploy costs below.
• So a month of occasional use can cost a few dollars, not US$107. Deploy Payload, use it for a while, delete the project. Restore it next time you want it, confirm with Update, and delete it again when you're done. You pay only for the module fee once, the builds, and whatever time Payload was actually live.
• What this needs. You must own the project (not a bring-your-own one RAD only manages billing for), and you restore it yourself within the 30 days — after that, Google deletes it for good. Restoring is admitted like creating a new project: your purchased credit balance must still clear the tier's floor (100 credits for the sandbox tier most study and demo use fits). Google says most services are fully working again within 36 hours of a restore.
• One real gap: the backup bucket does not survive. Payload's backups are written to a bucket inside the project, and that bucket is not protected by Cloud Storage's soft-delete, so it is very likely gone as soon as you delete the project — even though the project itself is recoverable for 30 days. If you have made changes to Payload you want to keep, copy a backup out (to Google Drive, or a bucket outside the project) before you delete, the same as the redeploy workflow below. For a default installation with nothing irreplaceable in it, this does not matter.

Pay only while you use it, the other way: delete and redeploy

GKE Autopilot does not scale to zero between requests. The option above only applies to a RAD-managed project; in your own project, or once the 30-day window has passed, the way to stop paying is to delete the deployment and deploy it again when you need it.

• What a redeploy costs. The module fee again plus the builds: about 180 credits (US$18), because RAD recreates the project and shared services before Payload. Deleting saves money only once Payload would otherwise sit unused for about 6 days or more, both in your own project (about US$3.55 a day) and in a RAD-managed one (92 credits a day).
• Most of the running cost is usually the database and the shared file/cache VM. They stop only when nothing else in the project uses them, so deleting Payload while something else shares the project saves only Payload's own compute part.
• Keep data first. Backups are written to a bucket inside the deployment and are deleted with it, so copy the latest backup out before deleting if you want to keep it.
• RAD does not recreate a deleted deployment for you; you enter the settings again when you deploy.

Lab sessions and Managed Environments

• Lab sessions, for training. A trainer runs a session for a class. Each participant gets the app in their own Google Cloud project for 15 minutes to 24 hours, within an allowance the trainer sets. Either the trainer funds every place, or each participant pays for their own. Everything is deleted when the session ends and unused credits go back to the trainer.
• Managed Environments, for consultancies. A partner runs the app for a client from a ring-fenced wallet it funds, and settles with the client directly. If the wallet runs low, billing pauses and the data is kept, so nobody receives an unexpected charge. At the end the partner hands the project over and the deployments become the client's own.

Credits can be bought in more than 20 currencies, including XAF, XOF, NGN, GHS, KES and ZAR, by card, bank transfer or mobile money.

Sources (8 October 2026): Google Cloud list prices from the Cloud Billing Catalog API; radmodules.dev/pricing for RAD's own fees and daily-credit estimates; Delete and restore projects for the pause/restore window and what it says about Cloud Storage objects without soft delete.

---

1. Overview

Payload runs as a Node.js (Next.js) pod on GKE Autopilot. There is no official Payload Docker image — this module builds a real, locally-verified starter application from source (a blank create-payload-app template using the PostgreSQL adapter) via Cloud Build. The deployment wires together a focused set of Google Cloud services:

Capability	Google Cloud service	Notes
Compute	GKE Autopilot	Node.js (Next.js standalone) pods, horizontally autoscaled
Build	Cloud Build	Builds the bundled Payload starter app from Payload_Common/scripts/Dockerfile — no prebuilt image exists to pull
Database	Cloud SQL for PostgreSQL 15	Required — Payload's Postgres adapter is used; MySQL/MongoDB are not wired
Object storage	None	No bucket is provisioned; media uploads go to local, ephemeral container disk
Secrets	Secret Manager	Auto-generated PAYLOAD_SECRET; database password
Ingress	Cloud Load Balancing	External LoadBalancer by default, optional custom domain + managed certificate

Sensible defaults worth knowing up front:

• PostgreSQL 15 is mandatory and schema is not created on boot. Booting the built server against a fresh database creates zero tables — the payload-migrate init job applies schema via the payload migrate CLI, using a pre-generated migration file baked into the image.
• container_image_source is fixed at "custom". There is nothing to deploy without a Cloud Build run — the module always builds Payload_Common/scripts/ from source.
• Health probes target /admin, not / or an API route. /admin serves Payload's login/first-user-creation form and returns an unauthenticated 200; Payload's REST/GraphQL routes require auth and are unsuitable probe targets.
• No storage bucket is provisioned. By default uploaded media is written to local container disk and does not survive a pod restart or redeploy. Set stateful_pvc_enabled = true to keep it on a per-pod block PVC mounted at /app/media (where Payload writes uploads, with stateful_fs_group = 1001).
• enable_redis and related Group 21 variables are declared but inert. They are not forwarded to Payload_Common, which has no Redis wiring.
• service_type defaults to LoadBalancer. This deployment's own live verification used ClusterIP only because the target project's IN_USE_ADDRESSES static IP quota was exhausted at deploy time — an operational choice made for that specific deployment, not a module default. Flip back to LoadBalancer (or reserve a static IP) once quota is available.
• The first admin user is created manually. Payload has no non-interactive CLI for this — visiting /admin on an empty users collection shows a signup form.
• Minimum 1 replica is maintained by default (min_instance_count = 1).

---

2. Google Cloud Services & How to Explore Them

All commands assume you have run gcloud container clusters get-credentials <cluster> --region <region> --project <project> and that PROJECT, REGION, and NAMESPACE are set. The namespace and other identifiers are reported in the deployment Outputs.

A. GKE Autopilot — the Payload workload

Payload pods are scheduled on Autopilot, which bills for the CPU/memory the pods actually request.

• Console: Kubernetes Engine → Workloads → select the Payload workload to see pods, revisions, and events. Kubernetes Engine → Services & Ingress shows the external IP.
• CLI:

  kubectl get pods,svc -n "$NAMESPACE"
  kubectl logs -n "$NAMESPACE" deploy/<service-name> --tail=100

See App_GKE for how Autopilot, scaling, and the workload type (Deployment vs StatefulSet) are managed.

B. Cloud Build — building the Payload image

Because no official Payload image exists, every deploy (and every redeploy after a Dockerfile or source change) triggers a Cloud Build run against Payload_Common/scripts/.

• Console: Cloud Build → History.
• CLI:

  gcloud builds list --project "$PROJECT" --limit=10
  gcloud builds log <build-id> --project "$PROJECT"

C. Cloud SQL for PostgreSQL 15

Payload stores all application data (Collections, users, uploaded document metadata) in a managed Cloud SQL for PostgreSQL 15 instance. Pods reach it privately through the Cloud SQL Auth Proxy sidecar over a Unix socket; no public IP is exposed. On first deploy, db-init creates the database and role, then payload-migrate applies the schema.

• Console: SQL → select the instance for connections, backups, flags, and metrics.
• CLI:

  gcloud sql instances list --project "$PROJECT"
  gcloud sql instances describe <instance-name> --project "$PROJECT"
  gcloud sql connect <instance-name> --user=<db-user> --database=<db-name> --project "$PROJECT"

The instance name, database name, user, and the Secret Manager secret holding the password are all surfaced in the Outputs. See App_GKE for the connection model, automated backups, and password rotation.

D. Secret Manager

One cryptographic secret is generated automatically and stored in Secret Manager: PAYLOAD_SECRET (used to sign Payload's own session/auth tokens). The database password is managed separately by the foundation.

• Console: Security → Secret Manager.
• CLI:

  gcloud secrets list --project "$PROJECT"
  gcloud secrets versions access latest --secret=<secret-name> --project "$PROJECT"

See App_GKE for the Secret Store CSI integration and rotation.

E. Networking & ingress

By default the workload is exposed through an external Cloud Load Balancing IP (service_type = LoadBalancer). A custom domain with a Google-managed certificate can be enabled, and a static IP can be reserved so the address survives redeploys.

• Console: Network services → Load balancing; VPC network → IP addresses.
• CLI:

  kubectl get svc -n "$NAMESPACE"
  gcloud compute addresses list --project "$PROJECT"

If the deployment was set to service_type = "ClusterIP" (for example because static IP quota was exhausted at deploy time), reach the app from inside the cluster instead:

kubectl port-forward -n "$NAMESPACE" svc/<service-name> 18080:3000
curl -s http://localhost:18080/admin -o /dev/null -w '%{http_code}\n'

See App_GKE for custom domains, Cloud CDN, and static IP details.

F. Cloud Logging & Monitoring

Pod stdout/stderr flow to Cloud Logging; GKE and Cloud SQL metrics flow to Cloud Monitoring. Optional uptime checks and alert policies are available.

• Console: Logging → Logs Explorer; Monitoring → Dashboards / Alerting.
• CLI:

  gcloud logging read 'resource.type="k8s_container" AND resource.labels.namespace_name="'"$NAMESPACE"'"' \
    --project "$PROJECT" --limit 50

---

3. Payload Application Behaviour

• First-deploy database setup. db-init (using postgres:15-alpine) connects through the Cloud SQL Auth Proxy and idempotently creates the application role and database.
• Schema migration is a separate, dependent job. payload-migrate (depends_on_jobs = ["db-init"]) runs ./node_modules/.bin/payload migrate from a full /app/cli copy of node_modules + TypeScript source baked into the image — the trimmed Next.js standalone runtime used to serve traffic does not include the Payload CLI or its dependencies. On GKE the Cloud SQL Auth Proxy runs as a native sidecar; the migrate script signals it to stop via http://localhost:9091/quitquitquit after migrations complete.
• PAYLOAD_SECRET should be treated as immutable after first boot. It signs Payload's session/auth tokens; rotating it invalidates all active sessions.
• Health path. Startup and liveness probes target /admin — Payload's admin UI route, which returns an unauthenticated 200 once the Node.js server and database connection are ready. Allow several minutes on first boot for the payload-migrate job to complete before the service is expected to serve real content.
• First admin account. Payload has no CLI command to create the first admin user non-interactively. Visit $SERVICE_URL/admin (or kubectl port-forward if ClusterIP) — with an empty users collection Payload shows a signup form to create the first administrator. This is a manual, one-time operator step.
• Media uploads do not persist by default. No storage bucket is provisioned; uploaded files are written to local container disk and are lost on the next pod restart or redeploy unless stateful_pvc_enabled = true puts /app/media on a block PVC.
• service_type may need a manual flip. If exposed as ClusterIP due to IP quota constraints, the app is reachable only via kubectl port-forward/kubectl exec until it is flipped back to LoadBalancer (or a static IP is reserved) and re-applied.
• Inspect job execution:

  kubectl get jobs -n "$NAMESPACE"
  kubectl logs -n "$NAMESPACE" job/<job-name>

---

4. Configuration Variables

Variables are grouped exactly as they appear on the deployment platform. Only settings specific to or notable for Payload are listed; every other input is inherited from App_GKE with its standard behaviour and defaults.

Group 1 — Project & Identity

Variable	Default	Description
project_id	(required)	Target Google Cloud project.
region	us-central1	Region for the workload and regional resources.

Group 2 — Deployment Environment

Variable	Default	Description
tenant_id	demo	Short suffix that makes resource names unique per environment.

Group 3 — Application Identity

Variable	Default	Description
application_name	payload	Base name for resources. Do not change after first deploy.
application_display_name	Payload CMS	Human-readable name shown in the Console. Leftover text from the module's clone source — override to Payload CMS at deploy time; it is cosmetic only.
application_version	latest	Deployment-tracking tag baked into the image via the Cloud Build application_version build arg.

Group 4 — Runtime & Scaling

Variable	Default	Description
deploy_application	true	Set false to provision infrastructure only.
container_image_source	custom	Fixed — there is no prebuilt Payload image to deploy.
min_instance_count / max_instance_count	1 / 3	Minimum replicas kept warm; GKE does not scale to zero.
container_port	3000	Next.js default.
container_resources	{ cpu_limit="1000m", memory_limit="1Gi" }	Payload (Next.js) needs headroom for the standalone server plus the migrate job's TypeScript/CLI footprint; consider raising memory for production.
enable_cloudsql_volume	true	Cloud SQL Auth Proxy sidecar for socket connections.
enable_image_mirroring	true	Mirrors the built image into Artifact Registry.

Group 5 — Environment Variables & Secrets

Variable	Default	Description
environment_variables	{}	Extra non-secret settings. Do not set PAYLOAD_SECRET or DATABASE_URL here — both are computed automatically.
secret_environment_variables	{}	Map of env var → Secret Manager secret name.

Group 6 — GKE Backend & Cluster

Variable	Default	Description
service_type	LoadBalancer	How the Kubernetes Service is exposed. Set to ClusterIP if the project's static IP quota is exhausted; access via kubectl port-forward instead.
workload_type	null (resolves to Deployment)	Payload has no need for per-pod PVCs by default.
session_affinity	None	No sticky-session requirement for this minimal starter app.

Group 11 — Cloud Storage & Artifact Registry

Variable	Default	Description
enable_gcs_storage	false	Declared in variables.tf with a description implying an S3-compatible GCS storage adapter, but not forwarded to Payload_Common — has no effect. Payload_Common's storage_buckets output is always [].
gcs_volumes	[]	Genuinely forwarded — GCS Fuse volume mounts, if you want to wire persistent storage in yourself.

Group 13 — Filesystem (NFS) & Jobs

Variable	Default	Description
initialization_jobs	[]	Leave empty to use the built-in db-init → payload-migrate chain.
enable_nfs	false	Not required — Payload's own data lives in Postgres, not NFS.

Group 14 — Observability & Health

Variable	Default	Description
startup_probe_config	HTTP /admin, 120s delay, 15s period, 40 retries	~12-minute total window for first-boot migrations to complete.
health_check_config	HTTP /admin, 30s delay	Liveness probe.

Group 15 — Database Backend

Variable	Default	Description
database_type	POSTGRES_15	Fixed; Payload requires PostgreSQL.
application_database_name / application_database_user	payload / payload	PostgreSQL database name and application user. Immutable after first deploy.

Group 19 — Custom Domain, Static IP & Networking

Variable	Default	Description
reserve_static_ip	true	Stable external IP across redeploys — useful once service_type = LoadBalancer is restored after any temporary ClusterIP fallback.

Group 21 — Cloud Armor & Redis Cache

Variable	Default	Description
enable_redis / redis_host / redis_port / redis_auth	true / "" / 6379 / ""	Inert. Declared in variables.tf (with a description claiming Payload v0.4+ requires Redis) but never forwarded to Payload_Common, which has no Redis wiring at all. Setting these has no effect.

---

5. Outputs

These values are returned on a successful deployment and are the quickest way to locate and explore the running resources.

Output	Description
service_name	Kubernetes Service name.
namespace	Namespace the workload runs in.
service_cluster_ip	In-cluster ClusterIP.
service_external_ip	External LoadBalancer IP (when a static IP is reserved or service_type = LoadBalancer).
service_url	URL to reach Payload.
database_instance_name	Cloud SQL instance name.
database_name / database_user	Application database name / user.
database_password_secret	Secret Manager secret holding the DB password.
database_host / database_port	DB endpoint (127.0.0.1 via the Auth Proxy) / port.
storage_buckets	Always empty — no bucket is provisioned.
container_image / container_registry	Built image and Artifact Registry repo.
initialization_jobs / db_import_job	Names of the setup and (optional) import jobs.
deployment_id / tenant_id / resource_prefix	Naming identifiers.
project_id / project_number	Project identifiers.
kubernetes_ready	Whether the cluster/workload is ready.
monitoring_enabled / monitoring_notification_channels	Monitoring status and channels.
cicd_enabled / cicd_configuration	CI/CD status and details.
vpc_sc_enabled / vpc_sc_perimeter_name / vpc_sc_dry_run_mode	VPC-SC status.
audit_logging_enabled / artifact_registry_cmek_enabled	Audit logging and CMEK status.

---

6. Configuration Pitfalls & Sensible Defaults

Risk: Critical (data loss / outage / security) — High (service degraded) — Medium (cost or partial degradation) — Low (minor).

Setting	Sensible value	Risk	Consequence if wrong
PAYLOAD_SECRET (auto-generated)	Never rotate after first boot	Critical	Rotating it invalidates every active session, forcing all users to log back in.
application_database_name / application_database_user	Set once	Critical	Immutable after first deploy; renaming recreates the DB/user and destroys all data.
startup_probe_config / payload-migrate timing	Allow the full ~12-minute window	High	If the probe window is shortened below the time payload-migrate needs, the pod can be marked unhealthy before schema migration finishes, since the two run concurrently rather than the probe waiting on the job.
Media/upload persistence	Add a real storage adapter before production use	High	With no storage bucket wired, all uploaded media lives on local container disk and is lost on every pod restart or redeploy.
service_type	LoadBalancer (default)	High	If left at ClusterIP (e.g. after a quota-driven fallback), the app has no external reachability until flipped back.
enable_gcs_storage	Do not rely on this toggle	Medium	Declared but not forwarded to Payload_Common — enabling it does not provision or wire any storage.
enable_redis / redis_*	Do not rely on these toggles	Medium	Declared but not forwarded to Payload_Common, which has no Redis wiring — setting them has no effect.
First admin creation	Complete promptly after deploy	Medium	Until the first admin is created via the /admin signup form, the instance has no authenticated user at all.
container_image_source	Leave at custom	Low	There is no prebuilt Payload image; setting prebuilt without a valid container_image breaks the deploy.

---

For the foundation behaviour referenced throughout — IAM and Workload Identity, autoscaling, ingress and certificates, CI/CD, Cloud Armor, IAP, Binary Authorization, VPC-SC, backups, and image mirroring — see App_GKE. Payload-specific application configuration shared with the Cloud Run variant is described in Payload_Common.

Related guides

• Hands-on lab: Payload CMS on GKE Autopilot — deploy it step by step, with the console screens and commands at each stage.
• Payload CMS on Google Cloud Run — the same application on Cloud Run, for when you need the other deployment target.
• Payload Common — Shared Application Configuration — the configuration shared by both deployment targets.