Skip to main content

FreshRSS on GKE Autopilot

FreshRSS on GKE Autopilot

FreshRSS is a free, self-hosted, GPL-3.0-licensed RSS and Atom feed aggregator — a lightweight, multi-user "news reader" written in PHP that runs behind Apache and exposes the Google Reader and Fever APIs for mobile clients. This module deploys FreshRSS 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 FreshRSS 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 FreshRSS costs on RAD, and how that compares​

Running FreshRSS in your own Google Cloud project costs about US$110.13/month in Google Cloud charges, on top of a one-time 110-credit RAD module fee (99 credits — 10% lower — in a RAD-managed project). FreshRSS is a free, open-source, self-hosted RSS and Atom feed aggregator. If you deploy into a RAD-managed project, you can pause the whole thing for free and pick it back up later — see "Pause it for free" below.

FreshRSS is also available on Cloud Run — see the Cloud Run guide for the lower-cost option; this guide covers the GKE variant.

What you pay on RAD​

Own projectRAD-managed project
Module fee110 credits (one-time)99 credits (one-time, 10% lower)
Build time~3–6 credits (one-time, per build)~3–6 credits (one-time, per build)
Google Cloud running cost~US$110.13/month, billed to your own Google billing account at list price~99 credits/day (~US$9.90/day), metered hourly at list price plus RAD's margin

Default running cost in your own project

ResourceUS$/month (Google list price, us-central1)
GKE Autopilot pod, 1x vCPU / 2 GiBUS$39.68
Cloud SQL for PostgreSQL, 1 vCPU / 3.75 GB, zonal, 10 GB SSDUS$51.02
NFS/cache file server (small VM)US$13.43
Cloud NAT and networkingUS$5
Cloud Storage (add-ons, backups)US$1
GKE cluster management feeshared across every GKE app in the project; $0 if this is your only cluster, else ~73
TotalUS$110.13

How it compares​

  • There is no officially hosted/cloud version of FreshRSS to compare against, so the honest comparison is against self-managed infrastructure, not a competing SaaS.
  • A bare VPS sized to roughly 1 vCPU / 2 GiB — a Hetzner CPX22 (2 vCPU/4GB, ~US$24/mo), a DigitalOcean 2 vCPU/4GB Droplet (US$24/mo), or GCP's own Compute Engine e2-standard-2 (2 vCPU/8GB, ~US$49/mo) — usually undercuts RAD's sticker price; a VPS a size or two smaller would undercut it further.
  • What that lower price does not include is the admin's own time: patching the OS, rotating database backups, renewing TLS certificates and watching for downtime. RAD's managed Cloud SQL, Secret Manager and monitoring do that for you, which is the actual trade being made, not the sticker price.
  • RAD is not claiming to be cheaper than a bare VPS in cash terms — it usually is not. The case for RAD is what's managed for you.

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

If FreshRSS is running in a RAD-managed project, you can stop paying for it entirely without destroying it. Deleting that project first unlinks its billing, then asks Google to delete the project outright — Google's own 30-day recoverable soft delete. Unlike deleting the FreshRSS module on its own, 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 build-time credits in total — roughly 5–10 credits, under US$1, for a typical 2–3-deployment chain. The 99-credit module fee is not charged again.

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 have customised FreshRSS 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​

If FreshRSS is in your own project (or more than 30 days have passed on a RAD-managed one), the way to stop paying is to delete the deployment and redeploy later when you need it again.

A redeploy costs the module fee again, plus the builds: roughly 110 + ~3–6 credits (~US$11.45) in your own project, or 99 + ~3–6 credits on a RAD-managed one.

Deleting only saves money once FreshRSS would otherwise sit unused long enough to clear that redeploy cost against its own running cost: roughly about 4 days or more idle in your own project (against its ~US$110.13/month Google Cloud bill), or about 2 days or more idle in a RAD-managed project (against its ~99 credits/day).

Most of FreshRSS's running cost is the database and the shared file/cache VM. They stop only when nothing else in the project uses them, so deleting this app while something else shares the project saves only this app's own compute share of the total above.

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 docs for the pause/restore mechanics.

1. Overview​

FreshRSS runs as a PHP/Apache web workload. The deployment wires together a focused set of Google Cloud services:

CapabilityGoogle Cloud serviceNotes
ComputeGKE AutopilotPHP/Apache pods on port 80, 1 vCPU / 2 GiB by default
DatabaseCloud SQL for PostgreSQL 15Required — the entrypoint installs with --db-type pgsql
Persistent storageNFS (Filestore / self-managed) or block PVCData dir at /var/www/FreshRSS/data; holds config, per-user state, feed cache. No GCS bucket
CacheRedis (optional)Off by default; FreshRSS does not require it
SecretsSecret ManagerAuto-generated FRESHRSS_ADMIN_PASSWORD; database password
IngressCloud Load BalancingExternal LoadBalancer, optional custom domain + managed certificate

Sensible defaults worth knowing up front:

  • PostgreSQL 15 is the supported engine. The container entrypoint hardcodes --db-type pgsql and the db-init job is Postgres-only; the schema is created by FreshRSS's own installer on first boot.
  • On GKE the Cloud SQL Auth Proxy runs as a sidecar bound to 127.0.0.1:5432, so FreshRSS dials the loopback address (DB_HOST = 127.0.0.1) — the entrypoint treats loopback as a plaintext TCP host (the proxy terminates TLS).
  • NFS is enabled by default (enable_nfs = true) and mounted at /var/www/FreshRSS/data. FreshRSS writes its generated config, per-user state, cached articles, and favicons there — without a persistent volume this state is lost on pod restart. An NFS-backed Deployment uses the Recreate strategy so two pods never write the same volume during a rollout.
  • FRESHRSS_ADMIN_PASSWORD is generated automatically and stored in Secret Manager. It seeds the default admin account (and its API password) on first install.
  • Session affinity is ClientIP by default. Keeps a client's requests on the same pod.
  • Minimum 1 replica is maintained (min_instance_count = 1, max_instance_count = 1); GKE does not support scale-to-zero, so the in-container feed-refresh cron always runs.
  • A static external IP is reserved by default (reserve_static_ip = true) and a custom-domain Ingress is provisioned by default (enable_custom_domain = true) — populate application_domains to serve real hostnames.
  • The container listens on port 80 (Apache), not 8080.

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 FreshRSS workload​

FreshRSS pods are scheduled on Autopilot, which bills for the CPU/memory the pods actually request. Because the data dir is NFS-backed, the Deployment uses the Recreate update strategy rather than a rolling update.

  • Console: Kubernetes Engine → Workloads → select the FreshRSS workload to see pods 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
    kubectl describe deploy -n "$NAMESPACE" <service-name>

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

B. Cloud SQL for PostgreSQL 15​

FreshRSS stores all application data (feeds, subscriptions, articles, categories, users) in a managed Cloud SQL for PostgreSQL 15 instance. Pods reach it privately through the Cloud SQL Auth Proxy sidecar over the 127.0.0.1:5432 loopback; no public IP is exposed. On first deploy the db-init Job creates the application database and user, and FreshRSS's own installer creates the schema.

  • Console: SQL → select the instance for connections, backups, flags, 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, user, and password secret are all surfaced in the Outputs. For the connection model, backups, and password rotation, see App_GKE.

C. Persistent storage (NFS / PVC)​

FreshRSS's data directory (/var/www/FreshRSS/data) is backed by an NFS volume (enable_nfs = true), which holds the generated config, per-user state, cached articles, and favicons. This module declares no GCS bucket. A block PVC (StatefulSet) is an alternative persistence mode via the Group 7 StatefulSet variables.

  • Console: Filestore → Instances (managed NFS); Kubernetes Engine → Storage (PVCs).
  • CLI:
    gcloud filestore instances list --project "$PROJECT"
    kubectl get pvc -n "$NAMESPACE"
    kubectl exec -n "$NAMESPACE" deploy/<service-name> -- ls /var/www/FreshRSS/data

See App_GKE for the NFS server model, block PVCs, and GCS Fuse.

D. Redis (optional cache)​

Redis is disabled by default (enable_redis = false) and FreshRSS does not require it. It is exposed as a forwarded option for parity with sibling PHP modules; leave it off unless you have a specific reason to enable it.

  • Console: Memorystore → Redis (if using a managed instance).
  • CLI:
    redis-cli -h <redis-host> ping

E. Secret Manager​

One application secret is generated automatically: FRESHRSS_ADMIN_PASSWORD, which seeds the default admin account and its API password on first install. The database password is managed separately by the foundation.

  • Console: Security → Secret Manager.
  • CLI:
    gcloud secrets list --project "$PROJECT" --filter="name~freshrss"
    gcloud secrets versions access latest --secret=<secret-name> --project "$PROJECT"

The database password secret name is in the Outputs. See App_GKE for the Secret Store CSI integration and rotation.

F. Networking & ingress​

By default the workload is exposed through an external Cloud Load Balancing IP with a reserved static IP, and a custom-domain Ingress is provisioned. Add hostnames via application_domains for a Google-managed certificate.

  • Console: Network services → Load balancing; VPC network → IP addresses.
  • CLI:
    kubectl get ingress,svc -n "$NAMESPACE"
    gcloud compute addresses list --project "$PROJECT"

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

G. 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. FreshRSS Application Behaviour​

  • First-deploy database setup. The db-init Job runs using postgres:15-alpine. It connects through the Cloud SQL Auth Proxy and idempotently creates the application database and user and grants privileges. The job is safe to re-run.
  • First-boot install. The container's platform-entrypoint.sh resolves the DB host (127.0.0.1 proxy loopback on GKE), then drives FreshRSS's own cli/do-install.php (creates data/config.php and the schema) and cli/create-user.php (creates the admin account from FRESHRSS_ADMIN_PASSWORD), then chains the upstream entrypoint. The install is idempotent — skipped once data/config.php exists on the persistent volume.
  • Feed refresh cron. The upstream image starts an in-container cron (CRON_MIN = */15) that actualizes subscribed feeds every 15 minutes. Because GKE keeps at least one replica running, the cron always fires.
  • Admin credential. The default login is admin with the generated FRESHRSS_ADMIN_PASSWORD; the same value is set as the API password used by Google Reader / Fever API mobile clients. Change it in the FreshRSS UI after first login — rotating the Secret Manager value alone will not re-set an already-installed account.
  • Rollout strategy. With NFS enabled the Deployment uses Recreate (not RollingUpdate) so two pods never write the shared data dir simultaneously — a rolling update on this stateful app would deadlock on the file/DB locks.
  • Health path. The startup probe is a TCP check on port 80; the liveness probe is an HTTP GET on /i/, which returns a literal 200 and renders from the database (/ answers with a 302 redirect, which a load-balancer health check rejects). FreshRSS also serves an unauthenticated /status JSON endpoint suitable for uptime checks. Allow a generous first-boot window while the installer creates the schema.
  • Set the external URL after the IP is known. Confirm the LoadBalancer IP / Ingress host and set BASE_URL (via environment_variables) or application_domains so FreshRSS's self-referencing links resolve correctly:
    kubectl get svc,ingress -n "$NAMESPACE"

4. Configuration Variables​

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

Group 3 — Application Identity​

VariableDefaultDescription
application_namefreshrssBase name for resources. Do not change after first deploy.
application_versionlatestFreshRSS image tag; latest is pinned to a known-good tag (1.26.3) at build time. Pin explicitly in production.

All other inputs follow standard App_GKE behaviour.

Group 4 — Runtime & Scaling​

VariableDefaultDescription
deploy_applicationtrueSet false to provision infrastructure only.
container_image_sourcecustomFreshRSS ships a thin custom build; use prebuilt only with an external container_image.
cpu_limit1000mCPU per pod (1 vCPU).
memory_limit2GiMemory per pod; keep ≥ 512Mi.
min_instance_count1Minimum replicas; GKE requires ≥ 1. Keeps FreshRSS and its refresh cron running.
max_instance_count1Keep at 1 — a single pod owns the refresh cron and file-based state.
container_port80FreshRSS/Apache listens on port 80.
enable_cloudsql_volumetrueCloud SQL Auth Proxy sidecar for Postgres connections.
timeout_seconds300Maximum request duration (0–3600 seconds).

All other inputs follow standard App_GKE behaviour.

Group 6 — GKE Backend & Cluster​

VariableDefaultDescription
service_typeLoadBalancerHow the Kubernetes Service is exposed.
workload_typenullResolves to Deployment; NFS-backed, so it uses the Recreate strategy. Set StatefulSet only with a block PVC.
session_affinityClientIPSticky routing so a client stays on one pod.
network_tags["nfsserver"]nfsserver is required when enable_nfs = true.

All other inputs follow standard App_GKE behaviour.

Group 7 — StatefulSet​

VariableDefaultDescription
stateful_pvc_enablednullEnable a per-pod block PVC as an alternative to NFS for the data dir.
stateful_pvc_size10GiPer-pod PVC storage size.
stateful_pvc_mount_path/var/www/FreshRSS/dataContainer mount path for the PVC; must match the NFS mount path so the data dir persists either way.

All other inputs follow standard App_GKE behaviour.

Group 10 — Observability & Health​

VariableDefaultDescription
startup_probeTCP / 30s delay, threshold 20Startup probe; the high threshold allows first-boot install time.
liveness_probeHTTP /i/ 300s delayLiveness probe; /status is an alternative unauthenticated JSON endpoint.
uptime_check_config{enabled=false, path="/"}Cloud Monitoring uptime check; disabled by default.
alert_policies[]Optional metric alert policies.

All other inputs follow standard App_GKE behaviour.

Group 13 — Filesystem (NFS)​

VariableDefaultDescription
enable_nfstrueMounts a persistent NFS volume for the FreshRSS data directory. Keep enabled — required to persist config and per-user state.
nfs_mount_path/var/www/FreshRSS/dataWhere the NFS volume is mounted inside the container.

All other inputs follow standard App_GKE behaviour.

Group 16 — Database Backend​

VariableDefaultDescription
database_typePOSTGRES_15PostgreSQL engine version. FreshRSS installs with --db-type pgsql.
application_database_namefreshrssPostgreSQL database name. Immutable after first deploy.
application_database_userfreshrssApplication database user. Immutable after first deploy.
database_password_length32Generated password length (16–64).

All other inputs follow standard App_GKE behaviour.

Group 19 — Custom Domain, Static IP & Networking​

VariableDefaultDescription
enable_custom_domaintrueProvision Ingress for custom hostnames + managed certificate.
application_domains[]Hostnames to serve; set to obtain a managed certificate.
reserve_static_iptrueStable external IP across redeploys.

All other inputs follow standard App_GKE behaviour.

Group 15 — Redis Cache​

VariableDefaultDescription
enable_redisfalseOff by default; FreshRSS does not require Redis.
redis_host / redis_port"" / 6379Redis endpoint if enabled.

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.

OutputDescription
service_nameKubernetes Service name.
namespaceNamespace the workload runs in.
service_cluster_ipIn-cluster ClusterIP.
stage_service_cluster_ipsMap of ClusterIPs for stage-specific services.
service_external_ipExternal LoadBalancer IP (when a static IP is reserved).
service_urlURL to reach FreshRSS.
database_instance_nameCloud SQL instance name.
database_nameApplication database name.
database_userApplication database user.
database_password_secretSecret Manager secret holding the DB password.
database_host / database_portDB endpoint (127.0.0.1 via the Auth Proxy) / port.
storage_bucketsCreated Cloud Storage buckets (FreshRSS declares none).
network_name / network_exists / regionsVPC network, presence, available regions.
container_image / container_registryDeployed image and Artifact Registry repo.
monitoring_enabled / monitoring_notification_channelsMonitoring status and channels.
initialization_jobs / db_import_jobNames of the setup (db-init) and (optional) import jobs.
deployment_id / tenant_id / resource_prefixNaming identifiers.
project_id / project_numberProject identifiers.
cicd_enabled / cicd_configurationCI/CD status and details (repo, trigger, registry).
github_repository_url / github_repository_owner / github_repository_nameCI/CD GitHub details.
artifact_registry_repository / cloudbuild_trigger_name / cloudbuild_trigger_idRegistry and build trigger.
kubernetes_readyWhether the cluster/workload is ready.
vpc_sc_enabled / vpc_sc_perimeter_name / vpc_sc_dry_run_modeVPC-SC status.
audit_logging_enabled / artifact_registry_cmek_enabledAudit 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 — a Deployment workload with stateful_pvc_enabled = true, IAP with no authorized identities, memory quota values without binary unit suffixes, an out-of-range redis_port/backup_retention_days. Invalid configuration fails the plan with a clear, named error before any resource is created, so most mistakes below are caught up front rather than at apply or runtime.

SettingSensible valueRiskConsequence if wrong
enable_nfs (or a block PVC)trueCriticalWithout a persistent volume the FreshRSS data dir is ephemeral — config.php, per-user state, and cache are wiped on pod restart, forcing a re-install.
nfs_mount_path/var/www/FreshRSS/dataCriticalMounting elsewhere leaves the data dir ephemeral (same effect as no NFS).
application_database_name / application_database_userSet onceCriticalImmutable after first deploy; renaming recreates the DB/user and orphans all feed data.
database_typePOSTGRES_15CriticalFreshRSS installs with --db-type pgsql; a non-Postgres engine breaks the installer and db-init.
container_port80HighFreshRSS/Apache listens on 80; a wrong port fails the startup probe and pods never become Ready.
enable_cloudsql_volumetrueHighThe Auth Proxy sidecar is required for PostgreSQL connectivity over the 127.0.0.1 loopback.
max_instance_count1HighRunning more than one pod duplicates the in-container refresh cron and splits file-based session/cache state; a rolling update on the shared NFS dir deadlocks.
session_affinityClientIPHighWithout stickiness, a client's requests scatter across pods, disrupting logged-in sessions.
min_instance_count1HighGKE requires min ≥ 1; the validation guard rejects invalid values. Keeping 1 keeps FreshRSS and its refresh cron running.
quota_memory_requests / _limitsbinary units (4Gi, 8192Mi)CriticalBare integers are bytes and block all pod scheduling in the namespace.
enable_iaponly for private deploysHighIAP blocks all unauthenticated requests, including mobile clients using the Google Reader / Fever API.
FRESHRSS_ADMIN_PASSWORD (auto-generated)Change in the UI after first loginMediumRotating the secret alone does not re-set an already-installed account; the first password remains valid until changed in-app.
application_domainsSet with enable_custom_domainMediumenable_custom_domain = true with no hostnames provisions an Ingress that serves no managed certificate.
memory_limit2GiMediumValues below 512Mi risk OOM under heavy feed refresh.

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. FreshRSS-specific application configuration shared with the Cloud Run variant is described in FreshRSS_Common.

Need RAD to do something it does not do yet? Request it on the roadmap, or vote on what is already there.