GoToSocial on GKE Autopilot

GoToSocial is a lightweight, self-hosted ActivityPub/Fediverse server — a small alternative to Mastodon, written as a single static Go binary. This module deploys GoToSocial 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 GoToSocial 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 GoToSocial costs on RAD, and how that compares

Running GoToSocial on GKE Autopilot pod in your own Google Cloud project costs about $136.37/month in Google Cloud charges, on top of a one-off 110-credit RAD module fee charged once at deploy time. Running a single-user or small-group Fediverse instance doesn't need much compute — GoToSocial is a static Go binary — but a real Postgres database behind it is non-negotiable for federation to work reliably. In a RAD-managed project the module fee drops to 99 credits (10% lower) and the running cost is metered hourly in credits instead of billed directly by Google — see Pause it for free below for the one RAD-managed advantage a self-hosted VPS cannot match — or see the Cloud Run guide for the lower-cost option.

What you pay on RAD

	Own project	RAD-managed project
Module fee (one-off)	110 credits	99 credits (10% lower)
Build time	~3-6 credits per build, either way	~3-6 credits per build, either way
Google Cloud running cost	$136.37/month, billed by Google at list price	~105 credits/day, metered at list price plus RAD's margin

Default running cost in your own project

Resource	US$/month (Google list price, us-central1)
GKE Autopilot pod, 2x vCPU / 4 GiB	$79.35
Cloud SQL for PostgreSQL, 1 vCPU / 3.75 GB, zonal, 10 GB SSD	$51.02
Cloud NAT and networking	$5.00
Cloud Storage (add-ons, backups)	$1.00
GKE cluster management fee	shared across every GKE app in the project; $0 if this is your only cluster, else ~73
Total	$136.37

The GKE cluster management fee is excluded from the total above because it is shared across every GKE app in the project — $0 if this is the only one, otherwise roughly $73/month split across however many you run.

How it compares

• A self-managed comparison sized to this module's own 2 vCPU/4GB: Hetzner CPX22 (2 vCPU/4GB) runs about $24/mo, a DigitalOcean 2 vCPU/4GB Droplet is the same $24/mo, and a GCP Compute Engine e2-standard-2 (2 vCPU/8GB) is about $49/mo.
• Commercial Fediverse hosts exist (masto.host and similar), but they host Mastodon, not GoToSocial — there's no official hosted SaaS for this exact project, so a price comparison against one would be comparing different software.
• The gap is the same as everywhere else on this list: a VPS is cheaper in cash terms, but you are the one rotating the database password, applying security updates, and restoring from backup if the disk fills.

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

This is the headline advantage of a RAD-managed project, and it costs nothing while paused. Deleting a RAD-managed project unlinks its billing first, then asks Google to delete the project outright — Google's own 30-day recoverable soft delete. Unlike deleting one module, this does not tear down Cloud SQL, any VM, or the compute resource one by one: the whole project simply stops, and nothing is charged while it waits, because billing is already unlinked.

Restoring, within 30 days and only by the project's owner, 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. This costs only a handful of credits (under US$1) in total for a typical 2-3-deployment chain.

What this needs: you must own the project (not one RAD only manages billing for), you must restore it yourself within 30 days — after that Google deletes it for good — and restoring is admitted like creating a new project, so your purchased credit balance must still clear the tier's floor (100 credits for the sandbox tier most study/demo use fits). Google says most services are fully working again within 36 hours of a restore.

One real gap: nightly backups are written to a bucket inside the project, and that bucket has Cloud Storage's soft-delete explicitly turned off, 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've customised GoToSocial and want to keep that work, copy a backup out (to Google Drive, or a bucket outside the project) before deleting. For a default install with nothing irreplaceable in it, this does not matter.

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

In your own project, or once 30 days have passed on a RAD-managed one, there is no free pause — deleting removes the resources for good, and bringing GoToSocial back means redeploying from scratch. That costs about 113-116 credits (the 110-credit module fee plus a build).

Deleting only saves money once GoToSocial would otherwise sit unused long enough to clear that redeploy cost against its own running cost — for GoToSocial, that's about 3 days or more. Most of that running cost is usually the database (and, where this module uses one, the shared file-server VM or GKE cluster) — those stop only once nothing else in the project uses them, so deleting this app while something else shares the project saves only its own compute share.

Keep data first: nightly backups go 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.

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 Billing Catalog API list prices; radmodules.dev/pricing for RAD's own fees and daily-credit estimates; Google's project delete/restore documentation for the pause/restore claims.

1. Overview

GoToSocial runs as a single Go binary workload on GKE Autopilot, deployed directly from the official docker.io/superseriousbusiness/gotosocial image — no custom build. The deployment wires together a focused set of Google Cloud services:

Capability	Google Cloud service	Notes
Compute	GKE Autopilot	Go binary pods on port 8080, 2 vCPU / 4 GiB by default; max_instance_count hard-fixed at 1
Database	Cloud SQL for PostgreSQL 15	Required — fixed at POSTGRES_15; MySQL not supported. Database created with mandatory LC_COLLATE='C' LC_CTYPE='C' collation
Object storage	Cloud Storage	A storage bucket + dedicated HMAC service account, consumed unconditionally via GoToSocial's native S3-compatible client — no GCS FUSE mount
Secrets	Secret Manager	Auto-generated SUPERUSER_PASSWORD, S3 HMAC access/secret key pair; database password. Projected via the Secret Store CSI driver
Ingress	Cloud Load Balancing / Gateway API	LoadBalancer Service by default; static IP reserved by default

Sensible defaults worth knowing up front:

• PostgreSQL 15 with C collation is mandatory. database_type = "POSTGRES_15" is the default, and GoToSocial_GKE's validation.tf rejects any non-Postgres database_type at plan time. The db-init job additionally creates the database with LC_COLLATE='C' LC_CTYPE='C' — GoToSocial refuses to start against any other collation.
• Prebuilt image, not custom. container_image_source = "prebuilt" deploys docker.io/superseriousbusiness/gotosocial directly. No entrypoint wrapper is needed — configuration is entirely through discrete GTS_* env vars the binary reads natively.
• No migrate job. GoToSocial creates and upgrades its own schema automatically on every start; db-init only prepares the C-collation database and role.
• max_instance_count is hard-fixed at 1. GoToSocial's in-process cache has no cross-instance synchronization; upstream does not support multiple instances against the same database/storage. min_instance_count = 0 (scale-to-zero) is safe.
• Cloud SQL is reached through the cloud-sql-proxy sidecar loopback — GTS_DB_TLS_MODE = "disable" is correct on GKE, unlike Cloud Run (see the CloudRun guide's TLS-mode explanation). App_GKE's db_host_env_var_name implementation prefers 127.0.0.1 when the sidecar is present, so GoToSocial_Common's Common-layer default is used unmodified — GoToSocial_GKE's module_env_vars local is empty.
• No GCS FUSE mount. Media/avatar/attachment storage uses GoToSocial's native S3-compatible client pointed at GCS's S3-interop XML endpoint via a dedicated HMAC service account — not a filesystem mount.
• Health probes are TCP, not HTTP. GoToSocial's /readyz//livez endpoints reject any request lacking a User-Agent header with an anti-scraper 418 response — Kubernetes' built-in HTTP prober never sends one. Both startup_probe and liveness_probe are TCP against port 8080.
• The admin account is created best-effort automatically, but not guaranteed. Unlike Cloud Run, GKE's looser initialization-job ordering gives the admin-create job's retry loop a real chance to win the race against the main pod's boot — but it can still lose. See §3.

---

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 GoToSocial workload

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

  kubectl get pods,svc -n "$NAMESPACE" --selector="app.kubernetes.io/name=gotosocial" 2>/dev/null \
    || kubectl get pods,svc -n "$NAMESPACE"
  kubectl logs -n "$NAMESPACE" deploy/<service-name> --tail=100
  kubectl describe pod -n "$NAMESPACE" -l app=<service-name>

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

B. Cloud SQL for PostgreSQL 15

GoToSocial stores all application data (accounts, statuses, follows, media metadata) in a managed Cloud SQL for PostgreSQL 15 instance, created with the mandatory C collation by the db-init job. Pods reach it through the cloud-sql-proxy sidecar on 127.0.0.1; no public IP is exposed.

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

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

The instance name, database, user, and password secret are in the Outputs. See App_GKE for the connection model, backups, and password rotation.

C. Cloud Storage — media, avatars, attachments

A dedicated Cloud Storage bucket (suffix storage) and a service account holding an HMAC key are provisioned automatically, granting the storage SA roles/storage.objectAdmin on the bucket. GoToSocial writes to this bucket unconditionally from first boot — GTS_STORAGE_BACKEND=s3 is not optional.

• Console: Cloud Storage → Buckets.
• CLI:

  gcloud storage buckets list --project "$PROJECT" --filter="name~gotosocial"

See App_GKE for GCS Fuse (not used here) and CMEK options.

D. Secret Manager

GoToSocial's main container reads GTS_STORAGE_S3_ACCESS_KEY and GTS_STORAGE_S3_SECRET_KEY as secret-backed environment variables (projected via the Secret Store CSI driver); SUPERUSER_PASSWORD is only consumed by the admin-create job. The database password is managed separately by the foundation.

• Console: Security → Secret Manager.
• CLI:

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

See App_GKE for the Secret Store CSI integration and rotation, and GoToSocial_Common §2 for why these secrets flow through secret_ids/module_secret_env_vars, not the per-app config object's (dead) secret_environment_variables field.

E. Networking & ingress

service_type = "LoadBalancer" and reserve_static_ip = true are both defaults — keep reserve_static_ip = true: without a reserved static IP, GKE_SERVICE_URL can fall back to an unreachable internal *.svc.cluster.local hostname before the ephemeral LoadBalancer IP is known at the moment Terraform renders the Deployment's env vars (a documented, reproducible fleet-wide race, not specific to GoToSocial).

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

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

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. GoToSocial Application Behaviour

• First-deploy database setup. The db-init job runs scripts/db-init.sh using postgres:15-alpine. It waits for Cloud SQL to accept connections, then idempotently creates the application role and the database with LC_COLLATE='C' LC_CTYPE='C', grants privileges, and signals the cloud-sql-proxy sidecar to shut down (POST http://127.0.0.1:9091/quitquitquit) so the Job completes. Safe to re-run (execute_on_apply = true, max_retries = 3).
• No separate migrate job. GoToSocial migrates its own schema automatically on every server start.
• The admin account is best-effort automatic — but confirm it, don't assume it. GoToSocial has no web-based sign-up flow and no REST endpoint for the very first account. Confirmed live: the CLI panics with NewSignup: instance application not yet created, run the server at least once before creating users unless the main server has already booted successfully once. On GKE, execute_on_apply only controls whether Terraform waits for a job (App_GKE/jobs.tf: wait_for_completion = try(execute_on_apply, true)) — the underlying Job pod is still scheduled immediately, racing the main Deployment's first pod. admin-create.sh retries up to 20 times at 15-second intervals to absorb this race and often wins during the same apply — but it is not guaranteed. Verify the account exists (Task 2 in the lab) and re-trigger manually if it does not:

  kubectl get jobs -n "$NAMESPACE"
  kubectl logs -n "$NAMESPACE" job/<admin-create-job-name>
  kubectl create job --from=job/<admin-create-job-name> <admin-create-job-name>-retry -n "$NAMESPACE"

  Retrieve the generated password:

  SECRET=$(gcloud secrets list --project "$PROJECT" --filter="name~superuser-password" --format="value(name)")
  gcloud secrets versions access latest --secret="$SECRET" --project "$PROJECT"

• Orphaned account row — a real, recurring failure mode, not a one-off. If an admin-create attempt fails partway through (the common case: it raced the server's boot and lost mid-flow, or was interrupted), GoToSocial's NewSignup flow can leave the accounts table row inserted without the matching users row (it inserts the account first, then panics on the user-row step if the instance application doesn't exist yet). A retry then fails confusingly: IsUsernameAvailable reports "already in use" (the orphaned account row exists), but GetAccountByUsernameDomain/ GetUserByAccountID (used by both a further create retry and admin account promote) find nothing, panicking with sql: no rows in result set — a symptom that looks like a completely different bug. Fix: connect to the database directly (e.g. a one-off postgres:15-alpine debug pod on GKE, using the DB credentials from Secret Manager) and run:

  SELECT id, username, domain FROM accounts WHERE username='<username>';
  DELETE FROM account_settings WHERE account_id='<the id above>';
  DELETE FROM account_stats WHERE account_id='<id>';
  DELETE FROM accounts WHERE id='<id>';

  then retry admin-create cleanly. This can recur on any deploy where the first attempt races the boot and loses — see Task 5 of the lab for the full walkthrough.
• Health path — TCP only, and why curl needs a User-Agent. GoToSocial serves real, unauthenticated /readyz (DB SELECT, 500 on failure) and /livez (cheap 200) endpoints, but both reject any request without a User-Agent header with a 418 I'm a teapot response — confirmed live. Kubernetes' HTTP prober never sends one, so both startup_probe and liveness_probe stay TCP against port 8080. Every manual verification command needs an explicit -A/--user-agent flag:

  curl -A "gotosocial-check/1.0" -s "http://${EXTERNAL_IP}/readyz"

• Storage IAM propagation. GoToSocial panics on boot if it cannot reach its S3 storage backend. The storage SA's roles/storage.objectAdmin grant is wired against the Foundation's own storage_buckets output (not a whole-module depends_on, which would deadlock) — but a fresh first deploy can still see the very first pod boot race the IAM grant's ~1–2 minute propagation delay, producing a brief Access Denied crash-loop that self-resolves.
• Inspect the init jobs and running config:

  kubectl get jobs -n "$NAMESPACE"
  kubectl logs -n "$NAMESPACE" job/<db-init-job-name>
  kubectl exec -n "$NAMESPACE" deploy/<service-name> -- env | grep -E 'GTS_HOST|GTS_STORAGE|GTS_DB'

---

4. Configuration Variables

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

Group 3 — Application Identity

Variable	Default	Description
application_name	gotosocial	Base name for resources. Must be lowercase. Do not change after first deploy.
application_display_name	GoToSocial	Human-readable name shown in the Console.
application_description	GoToSocial — a lightweight, self-hosted ActivityPub/Fediverse server, on GKE Autopilot	Workload description field.
application_version	latest	Docker Hub image tag.
host	gotosocial.local	GTS_HOST — the public domain. Baked into every ActivityPub URI at creation time, immutable after first boot. Set your real domain before production.
account_domain	""	GTS_ACCOUNT_DOMAIN — optional vanity handle domain, separate from host. Defaults to host when empty. Same immutability risk.

Group 4 — Runtime & Scaling

Variable	Default	Description
container_image_source	prebuilt	Deploys the official Docker Hub image directly — no custom build needed.
min_instance_count	0	Scale-to-zero is safe — GoToSocial's single-instance constraint is about concurrency, not warmth.
max_instance_count	1	Hard architectural ceiling — GoToSocial's in-process cache has no cross-instance synchronization. Do not raise.
container_port	8080	GoToSocial's native GTS_PORT default. Must match the probe ports, or the pod never becomes Ready.
container_resources	{ cpu_limit = "2000m", memory_limit = "4Gi" }	Per-pod CPU/memory.
enable_cloudsql_volume	true	Runs the cloud-sql-proxy sidecar — GoToSocial connects over its 127.0.0.1 loopback.

Group 5 — Environment Variables & Secrets

Standard App_GKE behaviour — see App_GKE.

Group 6 — GKE Backend & Cluster

Variable	Default	Description
service_type	LoadBalancer	Public-facing app — leave LoadBalancer.
workload_type	null (resolves to Deployment)	GoToSocial does not need a StatefulSet PVC — media lives in GCS via the S3 client.
session_affinity	ClientIP	Sticky routing so a client reaches the same pod.

Group 10 — Observability & Health

Variable	Default	Description
startup_probe	TCP, /readyz (informational path only), initial_delay_seconds=15, failure_threshold=10	The only probe type that works against GoToSocial's User-Agent-gated health endpoints.
liveness_probe	TCP, /livez, initial_delay_seconds=30, failure_threshold=3	Same reasoning as startup_probe.
startup_probe_config / health_check_config	HTTP /, various	Foundation-level structured probes; superseded by startup_probe/liveness_probe above for this module.

Group 13 — Filesystem (NFS)

Variable	Default	Description
enable_nfs	false	Provisions Filestore. Not used by GoToSocial — media storage is via the native S3 client, not a mount.

Group 14 — Cloud Storage

Variable	Default	Description
storage_buckets	one generic data bucket (overridden)	main.tf supplies the actual storage bucket via GoToSocial_Common's output, consumed unconditionally from first boot — not opt-in like some other apps' S3 buckets.
gcs_volumes	[]	No GCS Fuse volumes mounted by default.

Group 15 — Redis

Variable	Default	Description
enable_redis	false	GoToSocial has no Redis dependency at all — its cache is in-process. Leave false.

Group 16 — Database Backend

Variable	Default	Description
database_type	POSTGRES_15	Plan-validated — validation.tf rejects anything but PostgreSQL 13/14/15 or NONE.
application_database_name	gotosocial	The database actually created (with C collation) and injected as GTS_DB_DATABASE.
application_database_user	gotosocial	The role actually created and injected as GTS_DB_USER; password auto-generated in Secret Manager.
db_host_env_var_name / db_port_env_var_name / db_user_env_var_name / db_name_env_var_name / db_password_env_var_name	GTS_DB_ADDRESS / GTS_DB_PORT / GTS_DB_USER / GTS_DB_DATABASE / GTS_DB_PASSWORD	Set by main.tf, not left at their generic-empty variable defaults — the mechanism that lets the GoToSocial binary read the Foundation's DB connection info.

Group 19 — Custom Domain, Static IP & Networking

Variable	Default	Description
enable_custom_domain	true	Provisions a Gateway by default. GoToSocial is told HTTPS is in use (GTS_PROTOCOL=https, Secure cookies) only when application_domains is also set; on a bare IP it runs as http.
application_domains	[]	If empty, a nip.io-style hostname based on the reserved static IP is used.
reserve_static_ip	true	Keep true — see §2E for the internal-DNS race this avoids.

All other inputs follow standard App_GKE behaviour.

---

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 (reserved by default).
service_url	URL to reach GoToSocial.
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 sidecar) / port.
storage_buckets	Created Cloud Storage buckets (the storage media bucket).
network_name / network_exists / regions	VPC network, presence, available regions.
container_image / container_registry	Deployed image and Artifact Registry repo.
monitoring_enabled / monitoring_notification_channels	Monitoring status and channels.
initialization_jobs / db_import_job	Names of the setup (db-init, admin-create) and (optional) import jobs.
deployment_id / tenant_id / resource_prefix	Naming identifiers.
project_id / project_number	Project identifiers.
cicd_enabled / cicd_configuration	CI/CD status and details (repo, trigger, registry).
artifact_registry_repository / cloudbuild_trigger_name / cloudbuild_trigger_id	Registry and build trigger.
kubernetes_ready	Whether the cluster/workload is ready.
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).

Inherited plan-time validation. This module passes its configuration through the App_GKE foundation engine, which validates values and combinations at plan time. GoToSocial_GKE's own validation.tf additionally blocks min_instance_count > max_instance_count, enable_redis = true with neither redis_host nor enable_nfs, database_type away from PostgreSQL, enable_iap = true without both OAuth credentials, and enable_cloudsql_volume = true with database_type = "NONE".

Setting	Sensible value	Risk	Consequence if wrong
host (GTS_HOST)	Set your real domain before first deploy	Critical	Baked into every ActivityPub actor/object URI at creation time; changing it after real accounts/posts exist breaks federation for everything created under the old value.
max_instance_count	1 (do not raise)	Critical	GoToSocial's in-process cache has no cross-instance synchronization; upstream does not support multiple instances against the same database/storage.
database_type	POSTGRES_15	Critical	Plan-time validated by GoToSocial_GKE's own validation.tf — MySQL/SQL Server are rejected before apply.
container_port / probe ports	8080 everywhere	Critical	A mismatch makes the probe hit a dead port and the pod never becomes Ready even though the app is healthy.
Storage IAM wiring (google_storage_bucket_iam_member)	Leave as shipped (references module.app_gke.storage_buckets["storage"])	Critical	GoToSocial panics on boot without S3 access. A depends_on = [module.app_gke] alternative would deadlock the Deployment against its own IAM prerequisite.
admin-create recovery	Follow the orphaned-account-row SQL fix if a retry panics with "no rows"	High	A partially-failed first attempt can leave an orphaned accounts row with no matching users row; naive retries fail confusingly without the cleanup SQL.
Health probes (startup_probe/liveness_probe)	Leave type = "TCP"	High	GoToSocial's /readyz//livez reject any request without a User-Agent header (418); switching to type = "HTTP" makes the probe fail forever, since Kubernetes' prober never sends one.
admin-create job outcome	Verify, don't assume	High	GKE's looser job ordering often lets admin-create win its race against the pod's boot automatically, but not always — confirm the account exists before treating the deploy as fully operational.
reserve_static_ip	true (default)	Medium	Without it, GKE_SERVICE_URL can fall back to an unreachable internal *.svc.cluster.local hostname before the ephemeral LoadBalancer IP is known — a documented fleet-wide race.
Manual curl/health checks	Always pass -A "<agent>"	Medium	Bare curl (and most default HTTP clients/monitors) get 418 I'm a teapot from GoToSocial's anti-scraper User-Agent gate, even on "unauthenticated" endpoints.
quota_memory_requests / _limits	binary units (4Gi, 8192Mi)	Critical	Bare integers are treated as bytes and block all pod scheduling in the namespace.
enable_nfs	false (the default)	Low	GoToSocial does not use the NFS mount in its default configuration, and Filestore is billed whether or not the app writes to it.
enable_redis	false (default)	Low	GoToSocial has no Redis dependency; leaving this true has no functional effect.

---

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. GoToSocial-specific application configuration shared with the Cloud Run variant (secrets, the db-init/admin-create jobs, and the storage service account) is described in GoToSocial_Common (module source: modules/GoToSocial_Common).

Related guides

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