TLS Crypto Provider Fix 🔒​

With both aws-lc-rs and ring linked into rustls 0.23, neither is the process default, so the first outbound TLS handshake panicked. api, chart-discovery, and status-updates now install the aws-lc-rs provider explicitly at startup, so HTTPS to registries, AWS, and OIDC providers works reliably.

Thanks to @ben-schreiber for this contribution!

Bots API: Create and Delete Restored 🤖​

The bots collection exposed only list/get/update in its OpenAPI schema — the POST /bots (createBot) and DELETE /bots/{id} (deleteBot) operations, and the NewBot request body, were missing from the generated schema and therefore from every generated SDK. The bots collection now registers the full set, so NewBot, createBot, and deleteBot are back in the OpenAPI document and the SDKs.

Python SDK 🐍​

There's now a Python SDK, published to PyPI as platz. Like the JavaScript SDK it's generated from the backend's OpenAPI schema and versioned to match the backend. Betas install with the pre-release flag:

pip install --pre platz

Tunable Deployment Credentials 🔑​

k8s-agent previously hard-coded the deployment credentials refresh cadence to a fraction of a fixed token lifetime. Both numbers are now operator-tunable via humantime durations:

• PLATZ_DEPLOYMENT_CREDENTIALS_REFRESH_INTERVAL — how often to refresh per-deployment credentials. Default 20m.
• PLATZ_DEPLOYMENT_CREDENTIALS_TOKEN_DURATION — lifetime of issued deployment credential tokens. Default 1h.

Both surface through the helm chart as per-instance values:

k8sAgent:
  instances:
    - name: default
      deploymentCredentialsRefreshInterval: 30m
      deploymentCredentialsTokenDuration: 2h

The refresh loop also no longer aborts the whole batch on the first deployment that fails to apply a credential — remaining deployments in the same chunk now get their refresh attempt, and the error is propagated after the chunk completes.

Thanks to @ben-schreiber for this contribution!

OCI Registry Discovery 📦​

chart-discovery can poll a generic OCI registry instead of listening for ECR push events. Set provider: oci on a chart-discovery instance and point oci.registryUrl at a Docker Distribution, zot, ghcr.io, or any other registry that speaks the OCI distribution spec.

The poller walks the registry's /v2/_catalog and per-repo tag listings, filters to helm-config-media-type artifacts, pulls each chart with helm pull, and records it the same way the ECR provider does. ECR continues to be the default, and existing deployments need no changes.

New chart values:

chartDiscovery:
  instances:
    - name: default
      provider: oci
      oci:
        registryUrl: "http://registry.platz.svc.cluster.local:5000"
        pollInterval: 5s

The helm_registries table gained a provider column to record which backend a registry came from. The Helm pod's bootstrap script now skips aws ecr get-login-password for non-ECR registries — they're expected to be anonymously readable from the cluster running the pod.

Local Cluster Provider Mode 🏠​

k8s-agent can register a single cluster from a kubeconfig context instead of scanning AWS for EKS clusters. Set provider: local on a k8s-agent instance and optionally pin a localContext; the agent derives a synthetic local:<ctx> provider ID and writes a row through the same code path EKS discovery uses.

This unblocks running Platz outside AWS, and is what powers the single-command local dev stack in platzio/dev.

k8sAgent:
  instances:
    - name: default
      provider: local
      localContext: kind-platz
      disableDeploymentCredentials: true

PLATZ_DISABLE_DEPLOYMENT_CREDENTIALS is readable from the environment in addition to the CLI flag, so the chart values can flip it on without a custom command line.

Configurable Database Connection Pool 🏊​

Every Platz pod that opens the database can have its connection pool tuned via environment variables — DB_POOL_MAX_SIZE (default 50), DB_POOL_MIN_IDLE, DB_POOL_CONNECTION_TIMEOUT_SECS (default 30), DB_POOL_IDLE_TIMEOUT_SECS (default 600), and DB_POOL_MAX_LIFETIME_SECS (default 1800). Database pool exhaustion is now handled as a graceful error rather than a panic.

Surfaced through the helm chart under database.pool and the Terraform module's database_pool variable:

module "platz" {
  source = "github.com/platzio/terraform-aws-platzio?ref=v0.7.0-beta.4/modules/main"
  # ...
  database_pool = {
    max_size                = 200
    connection_timeout_secs = 10
  }
}

Thanks to @joshimi for this contribution!

Build and Release Pipeline ⚡​

The backend Dockerfile was rewritten around cargo-chef with per-architecture musl static builds, and the release workflow now uses per-arch native runners instead of QEMU. End-to-end backend release time dropped from ~50 minutes to ~20. CI is split into separate test, build, and release workflows, with image builds running in parallel with tests and the release gated on tests passing.

The Dockerfile also gained a dev target stage so the same Dockerfile serves local dev and release builds, matching how the helm chart and terraform module already work. The frontend Dockerfile builds dist/ in its own multi-stage, multi-arch build, removing the implicit dependency on a pre-built artifact and matching the backend's release pattern.

Local Dev Lives in platzio/dev 🛠️​

The old docker-compose.yaml, scripts/run-api.sh, scripts/run-db.sh, scripts/run-oidc.sh, and scripts/container-build.sh have been removed from the backend repo. Local development now lives in platzio/dev, which orchestrates the full stack via Tilt against a kind cluster running k8s-agent in local provider mode and chart-discovery against a Docker Distribution registry.

SDKs 📚​

All three SDKs are published at 0.7.0-beta.4, regenerated against this release's OpenAPI schema (so they pick up the restored bots create/delete operations):

• platz-sdk (crates.io) — Rust.
• @platzio/sdk (npm) — TypeScript.
• platz (PyPI) — Python, new in this cycle. PyPI normalizes the version to 0.7.0b4; install betas with pip install --pre platz.

The Rust SDK's HelmRegistry carries the provider field, and the collections that had drifted out of it — bots, bot_tokens, deployment_permissions, env_user_permissions, helm_tag_formats — remain exposed.

Helm Chart ⎈​

Images bumped to v0.7.0-beta.4 (backend and frontend; the helm base image stays at v9). The chart's backend dependency platz-chart-ext moved to 0.7.1 (utoipa 5, juspay_jsonlogic). The chart is published with artifacthub.io/prerelease: true, so ArtifactHub flags it as a pre-release. Every Rust pod (api, chart-discovery, k8s-agent, resource-sync, status-updates) honors a per-pod extraEnv list for injecting arbitrary env vars without forking the chart.

Tunable Deployment Credentials 🔑​

k8s-agent previously hard-coded the deployment credentials refresh cadence to a fraction of a fixed token lifetime. Both numbers are now operator-tunable via humantime durations:

• PLATZ_DEPLOYMENT_CREDENTIALS_REFRESH_INTERVAL — how often to refresh per-deployment credentials. Default 20m.
• PLATZ_DEPLOYMENT_CREDENTIALS_TOKEN_DURATION — lifetime of issued deployment credential tokens. Default 1h.

Both surface through the helm chart as per-instance values:

k8sAgent:
  instances:
    - name: default
      deploymentCredentialsRefreshInterval: 30m
      deploymentCredentialsTokenDuration: 2h

The refresh loop also no longer aborts the whole batch on the first deployment that fails to apply a credential — remaining deployments in the same chunk now get their refresh attempt, and the error is propagated after the chunk completes.

Thanks to @ben-schreiber for this contribution!

OCI Registry Discovery 📦​

chart-discovery can poll a generic OCI registry instead of listening for ECR push events. Set provider: oci on a chart-discovery instance and point oci.registryUrl at a Docker Distribution, zot, ghcr.io, or any other registry that speaks the OCI distribution spec.

The poller walks the registry's /v2/_catalog and per-repo tag listings, filters to helm-config-media-type artifacts, pulls each chart with helm pull, and records it the same way the ECR provider does. ECR continues to be the default, and existing deployments need no changes.

New chart values:

chartDiscovery:
  instances:
    - name: default
      provider: oci
      oci:
        registryUrl: "http://registry.platz.svc.cluster.local:5000"
        pollInterval: 5s

The helm_registries table gained a provider column to record which backend a registry came from. The Helm pod's bootstrap script now skips aws ecr get-login-password for non-ECR registries — they're expected to be anonymously readable from the cluster running the pod.

Local Cluster Provider Mode 🏠​

k8s-agent can register a single cluster from a kubeconfig context instead of scanning AWS for EKS clusters. Set provider: local on a k8s-agent instance and optionally pin a localContext; the agent derives a synthetic local:<ctx> provider ID and writes a row through the same code path EKS discovery uses.

This unblocks running Platz outside AWS, and is what powers the single-command local dev stack in platzio/dev.

k8sAgent:
  instances:
    - name: default
      provider: local
      localContext: kind-platz
      disableDeploymentCredentials: true

PLATZ_DISABLE_DEPLOYMENT_CREDENTIALS is readable from the environment in addition to the CLI flag, so the chart values can flip it on without a custom command line.

Configurable Database Connection Pool 🏊​

Every Platz pod that opens the database can have its connection pool tuned via environment variables — DB_POOL_MAX_SIZE (default 50), DB_POOL_MIN_IDLE, DB_POOL_CONNECTION_TIMEOUT_SECS (default 30), DB_POOL_IDLE_TIMEOUT_SECS (default 600), and DB_POOL_MAX_LIFETIME_SECS (default 1800). Database pool exhaustion is now handled as a graceful error rather than a panic.

Surfaced through the helm chart under database.pool and the Terraform module's database_pool variable:

module "platz" {
  source = "github.com/platzio/terraform-aws-platzio?ref=v0.7.0-beta.2/modules/main"
  # ...
  database_pool = {
    max_size                = 200
    connection_timeout_secs = 10
  }
}

Thanks to @joshimi for this contribution!

Build and Release Pipeline ⚡​

The backend Dockerfile was rewritten around cargo-chef with per-architecture musl static builds, and the release workflow now uses per-arch native runners instead of QEMU. End-to-end backend release time dropped from ~50 minutes to ~20.

The Dockerfile also gained a dev target stage so the same Dockerfile serves local dev and release builds, matching how the helm chart and terraform module already work. The release workflow now runs image builds in parallel with tests and only publishes if tests pass.

The frontend Dockerfile builds dist/ in its own multi-stage, multi-arch build, removing the implicit dependency on a pre-built artifact and matching the backend's release pattern.

Local Dev Lives in platzio/dev 🛠️​

The old docker-compose.yaml, scripts/run-api.sh, scripts/run-db.sh, scripts/run-oidc.sh, and scripts/container-build.sh have been removed from the backend repo. Local development now lives in platzio/dev, which orchestrates the full stack via Tilt against a kind cluster running k8s-agent in local provider mode and chart-discovery against a Docker Distribution registry.

SDKs 📚​

platz-sdk (crates.io) bumped to 0.7.0-beta.2. The HelmRegistry struct gained a provider field, and five collections that had drifted out of the SDK — bots, bot_tokens, deployment_permissions, env_user_permissions, helm_tag_formats — are now exposed.

@platzio/sdk (npm) bumped to 0.7.0-beta.2, regenerated from the backend's OpenAPI schema.

Helm Chart ⎈​

Image bumped to v0.7.0-beta.2. The chart is published with artifacthub.io/prerelease: true, so ArtifactHub flags it as a pre-release.

In addition to the settings called out above, every Rust pod (api, chart-discovery, k8s-agent, resource-sync, status-updates) honors a per-pod extraEnv list for injecting arbitrary env vars without forking the chart. The resource-sync deployment correctly honors resourceSync.replicaCount — previously it was reading a non-existent top-level replicaCount.

error scanning image docker.io/bitnami/postgresql:14.5.0-debian-11-r35:
image not found (package platzio:0.6.2)

We've removed the entries that triggered it.

What happened​

Platz chart versions up to and including 0.6.2 declared a dependency on the Bitnami PostgreSQL subchart, which in turn pulled docker.io/bitnami/postgresql:14.5.0-debian-11-r35. In late 2025 Bitnami moved their historical image tags to a separate bitnamilegacy namespace and deleted them from docker.io/bitnami/*. The Helm chart still references the old path, so scanners (and anyone trying a fresh install of an old version) get a 404.

The dependency itself was removed in v0.6.3. The reasoning is in the v0.6.8 release notes and on the Database page: operators get full control over the database, and Platz doesn't ship with a specific database chart.

What changed​

We removed entries <= 0.6.2 from the Helm repository index. The repository now serves 0.6.3 through 0.6.8. That's enough history to roll back across the post-Bitnami removal; everything older was tied to the broken dependency anyway.

The release tarballs themselves are untouched — they still live on GitHub Releases. If you've pinned a specific older version by URL in your own automation, it will continue to resolve.

What you should do​

• If you're on 0.6.3 or later: nothing. You won't see any change.
• If you're still on 0.6.2 or earlier: plan an upgrade to the latest 0.6.x. Provision Postgres yourself (RDS, Cloud SQL, self-hosted — see the Database page), populate the postgres-creds secret, and bump the chart. The schema migration runs automatically on API pod startup.
• If you depended on the bundled Bitnami subchart in production: same as above. The bundled subchart was never intended for serious production use, and it's now structurally broken; migrating off it is the path forward.

OCI Registry Discovery 📦​

chart-discovery can now poll a generic OCI registry instead of listening for ECR push events. Set provider: oci on a chart-discovery instance and point oci.registryUrl at a Docker Distribution, zot, ghcr.io, or any other registry that speaks the OCI distribution spec.

The poller walks the registry's /v2/_catalog and per-repo tag listings, filters to helm-config-media-type artifacts, pulls each chart with helm pull, and records it the same way the ECR provider does. ECR continues to be the default, and existing deployments need no changes.

New chart values:

chartDiscovery:
  instances:
    - name: default
      provider: oci
      oci:
        registryUrl: "http://registry.platz.svc.cluster.local:5000"
        pollInterval: 5s

The helm_registries table gained a provider column to record which backend a registry came from. The Helm pod's bootstrap script now skips aws ecr get-login-password for non-ECR registries — they're expected to be anonymously readable from the cluster running the pod.

Local Cluster Provider Mode 🏠​

k8s-agent can now register a single cluster from a kubeconfig context instead of scanning AWS for EKS clusters. Set provider: local on a k8s-agent instance and optionally pin a localContext; the agent derives a synthetic local:<ctx> provider ID and writes a row through the same code path EKS discovery uses.

This unblocks running Platz outside AWS, and is what powers the single-command local dev stack in platzio/dev.

k8sAgent:
  instances:
    - name: default
      provider: local
      localContext: kind-platz
      disableDeploymentCredentials: true

PLATZ_DISABLE_DEPLOYMENT_CREDENTIALS is now readable from the environment in addition to the CLI flag, so the chart values can flip it on without a custom command line.

Configurable Database Connection Pool 🏊​

Every Platz pod that opens the database can now have its connection pool tuned via environment variables — DB_POOL_MAX_SIZE (default 50), DB_POOL_MIN_IDLE, DB_POOL_CONNECTION_TIMEOUT_SECS (default 30), DB_POOL_IDLE_TIMEOUT_SECS (default 600), and DB_POOL_MAX_LIFETIME_SECS (default 1800). Database pool exhaustion is now handled as a graceful error rather than a panic.

Surfaced through the helm chart under database.pool and the Terraform module's database_pool variable:

module "platz" {
  source = "github.com/platzio/terraform-aws-platzio?ref=v0.7.0-beta.1/modules/main"
  # ...
  database_pool = {
    max_size                = 200
    connection_timeout_secs = 10
  }
}

Thanks to @joshimi for this contribution!

Build and Release Pipeline ⚡​

The backend Dockerfile was rewritten around cargo-chef with per-architecture musl static builds, and the release workflow now uses per-arch native runners instead of QEMU. End-to-end backend release time dropped from ~50 minutes to ~20.

The Dockerfile also gained a dev target stage so the same Dockerfile serves local dev and release builds, matching how the helm chart and terraform module already work.

Local Dev Lives in platzio/dev 🛠️​

The old docker-compose.yaml, scripts/run-api.sh, scripts/run-db.sh, scripts/run-oidc.sh, and scripts/container-build.sh have been removed from the backend repo. Local development now lives in platzio/dev, which orchestrates the full stack via Tilt against a kind cluster running k8s-agent in local provider mode and chart-discovery against a Docker Distribution registry.

SDKs 📚​

platz-sdk (crates.io) bumped to 0.7.0-beta.1. The HelmRegistry struct gained a provider field, and five collections that had drifted out of the SDK — bots, bot_tokens, deployment_permissions, env_user_permissions, helm_tag_formats — are now exposed.

@platzio/sdk (npm) bumped to 0.7.0-beta.1, regenerated from the backend's OpenAPI schema.

Helm Chart ⎈​

Image bumped to v0.7.0-beta.1. The chart is published with artifacthub.io/prerelease: true, so ArtifactHub flags it as a pre-release.

In addition to the settings called out above, every Rust pod (api, chart-discovery, k8s-agent, resource-sync, status-updates) now honors a per-pod extraEnv list for injecting arbitrary env vars without forking the chart. The resource-sync deployment also now correctly honors resourceSync.replicaCount — previously it was reading a non-existent top-level replicaCount.

Versions 0.6.3 thru 0.6.7 are unstable intermediate releases and should not be used.

Alpine Base Image​

The backend image is now based on Alpine to reduce the number of unnecessary dependencies and image size.

Removed Postgres Chart Dependency​

Following Bitnami's deprecation notice the Postgres dependency in old charts can't be used anymore.

Please install Postgres separately from Platz.

Bot Users 🤖​

This new type of user can be created by admins, and have admins manage their API tokens.

Bot users can be used for automations, and have their own identity displayed in deployment task history.

ID-based Deployment Kinds​

Deployment kinds are now stored in their own collection, instead of being a plain string.

This would allow for implementing renames of kinds in the future.

⚠️ This change will break frontend paths, as paths now include the kind ID (a UUID) instead of the plain kind name.

Removed API v1​

API v1 lacked authentication, and was only enabled with an opt-in for past users of Platz.

/api/v1 is now removed completely.

Added env_id filter for /api/v2/deployments​

Thanks @ayalash for this contribution!

Users can now be deactivated​

Thanks @ayalash for this contribution!

Env deletion​

Thanks @ayalash for this contribution!

New showIf options for UI schema inputs​

Thanks @pyros82 for this contribution!

Pagination for Helm registry page and deployment task history​

Dark mode: 🕸️ + 🌚 = 🎉​

Thanks to Bootstrap's support for dark mode, Platz now supports it too!

The current behavior defaults to the system setting.

Fixes for missing metrics values​

Thanks @vmalloc for this contribution!

Missing metric values were left in their original value instead of being returned as zeros.

This is now fixed by setting all possible metrics' values to zero before setting new values.

New Platz annotations in Chart.yaml​

When parsing chart metadata, Platz now searches for a Chart.yaml file. If Chart.yaml is not found, Platz returns an error saying the directory is not of a chart.

Chart.yaml is used to get the chart version, which has be a valid Semver, plus some optional annotations:

• platz.io/git/commit: A full Git commit hash
• platz.io/git/branch: Name of the current Git branch
• platz.io/git/repo: URL for the Git repo
• platz.io/git/provider: A lowercase string representing the Git hosting provider

When provided, this metadata added in the chart annotations takes precedence over Tag Formats, which will be deprecated in Platz 0.6 (see below.)

Here's an example Chart.yaml:

apiVersion: v2
name: shop-chart
description: A Helm chart for Pizza-Platz Shops
type: application
version: 0.1.0
appVersion: "0.1.0"

annotations:
  platz.io/git/commit: a8c1da83308d93b111c14f0c79e0b9acf7f01686
  platz.io/git/branch: main
  platz.io/git/repo: https://github.com/pizza-platz/shop
  platz.io/git/provider: github

The repo and provider annotations are not being used at the moment, but they will allow for generating links into the Git repo web interface, for example, to show diffs of potential upgrades.

It's recommended to start adding these annotations so that charts are annotated far enough in history before Tag Formats are deprecated.

Chart Playground 🎡​

This new feature should make it easier to create and edit chart extensions: UI schema and features.

The main goal of the Chart Playground is to validate the syntax of chart extensions and show the resulting inputs and features visually before building and testing an actual version of a chart.

Also, the Chart Playground always runs in a context of a specific env. The reasoning is that many inputs may depend on data from the env, for example other deployments, secrets and deployment resources.

An env's Chart Playground can be accessed from the Env menu in Platz's navbar.

Multiple chart-discovery Workers​

Similarly to k8s-agent, chart-discovery runs with a role that has credentials to access a certain account with ECR repositories.

In cases where multiple AWS accounts or roles are required, you can now run multiple chart-discovery workers.

Each worker monitors its own SQS queue with ECR notifications. Note that although running multiple workers against the same queue, this was not tested nor recommended.

Refer to Platz's Terraform module for how to configure multiple chart-discovery workers.

OpenAPI Schemas + SDKs​

OpenAPI schemas are now generated on backend build and saved for each release.

For example, see the release page for this release: https://github.com/platzio/backend/releases/tag/v0.5.0.

This OpenAPI schema is used to auto-generate SDKs, the first of which is Platz's SDK for Javascript published to NPM.

Platz's frontend was adapted to use the SDK types and will be migrated to use all actions in a future release.

The Rust SDK is still maintained manually and still be auto-generated in a future release.

Kubernetes Clusters Ignored By Default​

Thanks @ayalash for this contribution!

New Kubernetes clusters are now ignored by default and have to be enabled manually in each env.

This is done to prevent new clusters from becoming automatically visible to users without being configured properly first.

Diesel 2.0​

The backend was upgraded to using Diesel 2.0.

Deprecations​

Tag Formats​

Tag formats were an attempt to add version, branch and commit metadata into Git tags and Helm chart versions.

This approach turned out to be troublesome since tags were sometimes very long, hard to read, and limited to 64 characters in Helm versions.

Existing parsed data is already saved in Platz's database and should not be affected.

👉 As of v0.5, parsing of Tag Formats has to be manually enabled in chart-discovery. This is to prevent multiple chart-discovery workers from deleting metadata when starting and to encourage this feature to be disabled.

👉 Since the new chart annotations provide more metadata, Tag Formats will be removed as of v0.6.

API v1​

API v1 was, well, the first API of Platz. It did not have authentication as it was used in a closed network. To protect users installing Platz using the default configuration, API v1 is disabled by default by not exposing it in Platz's ingress.

API v2 supports all routers of v1, with the addition of authentication and pagination.

👉 API v1 will be removed as of Platz v0.6.

Thanks @ayalash for this contribution!

The most exciting feature for this release: User API tokens 🎉.

This allows users to create API tokens and use them in API calls, and in the soon to be released CLI.

Prometheus metrics​

Thanks @ayalash for this contribution!

Another major feature is Prometheus metrics exposed by the backend.

The backend now exports two metrics:

• platz_deployment_task_status_counter counts how many tasks are in each status, with a task_status label.
• platz_deployment_status_counter counts deployment statuses by name and cluster using the labels deployment_kind, deployment_status and cluster_name.

Added retries for helm task pod creation/deletion​

Helm task pods sometimes failed to create due to network errors.

To resolve this, creation and deletion of Helm pods are now retried several times before marking the task as failed.

Added SIGTERM and SIGINT handlers​

Thanks @vmalloc for this contribution!

Sort API v2 of deployment tasks by creation​

Thanks @ayalash for this contribution!

Deployment tasks are now sorted by created_at when returned by the API. This helps API users to query for recent tasks instead of loading the entire history for certain deployments.

Switched from rusoto to official aws-sdk​

The backend now uses the official Rust AWS SDK instead of rusoto.

Fixed some charts missing in initial frontend load​

Thanks @vmalloc for this contribution!

Some charts were missing when the frontend was initially loaded charts using in_use=true query.

This caused the frontend to miss potential upgrade charts and not show the upgrade icon until all charts were loaded for each deployment kind.

Add 'active_only' and 'created_from' filtering for deployment tasks​

Thanks @ayalash for this contribution!

New filters are now available for /api/v2/deployment-task:

• active_only=true will return only active tasks.
• created_from takes an ISO8601 date-time and returns tasks that were created after this date.

API for getting the backend version​

Thanks @ayalash for this contribution!

A new API /api/v2/self returns version information about the backend.

Ignore ECR events with an empty image_tag​

Some ECR images might be pushed with an empty tag.

Since this is allowed, Platz simply ignores these images and would continue inspecting Helm chart artifacts, while ignoring anything that's pushed without a tag.

Upgraded base image v4 (Helm 3.11.0 and latest Ubuntu)​

The base image for running Helm tasks and the backend has been updated to the latest Ubuntu and Helm 3.11.0.

This matches the current ubuntu-latest in GitHub actions.

Fixed k8s-agent crash when cluster discovery fails​

Cluster discovery might fail sometimes due to network issues, which crashed k8s-agent.

k8s-agent will now retry cluster discovery on the next attempt and not crash when this happens.

Added an opt-in backup cronjob​

The Helm chart and Terraform modules now include an option for installing Platz with a built-in backup cronjob.

The cronjob uses the existing database credentials and runs every hour, saving a full database dump to an S3 bucket.

A new Terraform module creates the S3 bucket and all necessary IAM resources to grant the backup job its necessary credentials.

Multi-arch support (AMD64 and ARM64)​

All Platz images are now built for both linux/amd64 and linux/arm64 targets.

Task failure/success output fix​

Thanks to @vmalloc for contributing this fix

When running tasks, a failure prefix was added when tasks succeeded and not when they failed.

Avoid refreshing credentials of disabled deployments​

Disabled credentials are uninstalled from their clusters, so trying to refresh their credentials caused unnecessary errors as their namespaces didn't exist.

k8s-agent will not exit on task failure anymore​

k8s-agent exited with an error when a task failed. This happened because errors were propagated from the task runner loop instead of stopping after storing the error in the database.

Release builds now have debug info and show full backtraces​

Thanks to @vmalloc for contributing this fix

Backtraces in release mode ommitted many important function calls. Release builds now include full debug information.

Fixed some charts missing when using ?in_use=true in API​

Thanks to @vmalloc for contributing this fix

Helm charts returned by calling /api/v2/helm-charts?in_use=true didn't include charts with a parsed_branch value of null.

This might have happened when parsing a Helm chart version failed, in which case the branch is stored as null. The branch will now be stored as an empty string since this is equivalent to the null case (a Git branch name can't be empty), fixing the result of in use Helm charts.

Added context to errors around Helm chart pulling​

In some task failure cases it was hard to understand the cause of the failure. To help with that, a context was added to anyhow failures to show which stage failed while running Helm (creating pod, waiting for pod output, etc.)

Handle disconnections from control plane when running task pods​

Thanks to @vmalloc for contributing this fix

Some tasks failed running because of disconnections from the control plane while waiting for Helm pods to complete running.

In some cases a task completed successfully but was marked as failed because Platz couldn't get the pod status correctly.

This was fixed by using the kube crate watcher instead of polling the pod status directly.

Fixed chart-discovery OOMKilled​

When chart-discovery starts it scrubs all existing Helm charts to verify their parsed tag is in sync with the current settings.

In large deployments this caused an OOMKilled event right when chart-discovery starts up since it loaded all Helm charts to memory.

This was fixed by scrubbing Helm charts in smaller chunks.

Fixed frontend crash when a chart had no features​

Charts with no features caused a crash, this was fixed.

Fixed to show registry icons when no icon is defined in chart features​

When adding chart icons as a chart-ext feature, the previous behavior of showing the icon by the registry setting was accidentally removed.

This is now fixed to prefer the icon defined in chart-ext, then fall back to the registry setting.

Fix undefined error in findCollectionDependencies​

Thanks to @vmalloc for contributing this fix

Charts with no values_ui chart-ext caused the frontend to crash in findCollectionDependencies.

This was fixed, along with adding logs to catch similar errors in the future.

Platz has a new logo and mascot designed by Rotem Yaari! 🎉

Deployment icons as chart extension​

Deployment icons can now be set in the chart extension features:

apiVersion: platz.io/v1beta2
kind: Features
spec:
  display:
    icon:
      font_awesome: spider-web

Note: Setting Helm registry icons using the frontend will be removed in the future.

Ingress classes​

Clusters now have a setting for ingress class name.

When set, the ingress class name is injected as ingress.className in the chart values.

Ingress classes allow setting an explicit ingress controller to use for Platz deployments, or migrating to a new ingress controller one a per-env basis.

Fixed: Custom resources with no hooks​

Custom resources with no lifecycle hooks were left in the Creating status since no deployment was changing their status.

This was fixed to mark them as Ready when no lifecycle hooks need to be called.

chart-ext is now an independent crate​

The chart-ext crate has been moved out of the backend and is now published at https://crates.io/crates/platz-chart-ext.

This is in preparation for more integrations around chart extensions, such as a CLI and VS Code extension.

Fixed: Global admin env permissions​

Thanks to @vmalloc for reporting this issue

Global admins were unable to see or change envs in the frontend if their permissions were removed, or didn't exist to begin with.

This was fixed so that global admins can see and admin any env in Platz.

Fixed: Show deployment resource errors in resource pages​

Thanks to @ayalash for reporting this issue

Custom resource errors were visible in the resource list but no in the custom resource page.

Custom resource status and any errors are now displayed in the resource page as well.

The APIs for Helm registries and charts can now filter by these properties:

• /api/v2/helm-charts with ?in_use=true returns all Helm charts currently in use and their potential upgrades.
• /api/v2/helm-reegistries contains the kind property for each registry. Registries can also be filtered by their kind.

Frontend Improvements​

The frontend uses the API improvements mentioned above to initially load much faster: Instead of loading all Helm charts, the frontend loads only charts in the use and upgrade candidates.

When navigating to a page that requires all charts from a registry/kind the frontend fetches those specifically. For example, this happens when creating/editing a deployment.

platz Section​

When deployment a chart, the platz section injected into the chart's values now contains the full cluster object as it would be returned from the API.

This is useful when cluster properties such as its domain are needed during deployment.

The platz-creds secret injected into each deployment namespace now contains two new properties:

• server_url with Platz's URL
• expires_at for when the credentials expire and should be refreshed, in ISO 8601 format

These changes are also applied in the Rust SDK version 0.4.3.

To use the injected secret, run a deployment with the platz-creds secret mounted like so:

apiVersion: apps/v1
kind: Deployment
spec:
  template:
    spec:
      containers:
        - name: ...
          volumeMounts:
            - name: platz-creds
              mountPath: "/var/run/secrets/platz"
              readOnly: true
      volumes:
        - name: platz-creds
          secret:
            secretName: platz-creds
            optional: false

The --admin_email option was renamed to --admin-email.

The previous option was broken when used through the Helm chart or Terraform modules.

Chart ingress.hosts renamed to ingress.rules​

This change is to allow rules to be specified similarly to how an Ingress resource looks like. Each rule host is now optional, allowing for a catch-all ingress.

This only changes the chart of Platz itself, the ingress injected while installing charts stays the same to match the commonly used schema created by default with helm create.

When migrating from previous versions, convert an ingress that looks like this:

ingress:
  create: true
  className: nginx
  hosts:
    - host: platz.pizza.io
      paths:
        - path: /
          pathType: Prefix
  tls:
    - hosts:
        - platz.pizza.io
      secretName: pizza-platz-tls

to:

ingress:
  create: true
  className: nginx
  rules: # <--
    - host: platz.pizza.io
      paths:
        - path: /
          pathType: Prefix
  tls:
    - hosts:
        - platz.pizza.io
      secretName: pizza-platz-tls

Use PLATZ_OWN_URL for OIDC callback​

For generating the OIDC callback URL, the backend used to use proxy headers, which didn't work in some cases.

The backend now uses the PLATZ_OWN_URL environment variable. This environment variable is generated automatically in Platz's Helm chart when setting an Ingress according to the host and tls settings.

ownUrlOverride in Helm chart​

The Helm chart now contains a top-level variable ownUrlOverride for overriding the automatic value created for PLATZ_OWN_URL.

This is useful when Platz is installed behind a reverse proxy that also serves TLS, which generates an http:// URL instead of https://.

Fixed site admin to always be an env admin​

Site admins should be able to have permissions for everything, however in many places they can't perform operations without adding permissions for themselves. This is useful for testing how permissions look like for other users, and of course to avoid mistakes.

However, admins should be able to manage users for all envs to avoid being locked out, so site admins are now also env admins without adding this permission explicitly.

Site admins can still be removed from envs if they don't want to see those envs in the env switcher.

Fixed chart-discovery crash when chart download fails​

Thanks to @vmalloc for reporting this issue!

The chart-discovery worker could crash in cases where it couldn't download a chart from an OCI repo.

Instead of crashing, the worker now stores the chart with an indicative error.

Fixed wrong SQS queue prefix in chart-discovery Terraform module​

Thanks to @vmalloc for reporting this issue!

The chart-discovery Terraform module used k8s_namespace as a prefix for the SQS queue it creates.

This is now fixed to use name_prefix.

Frontend changes​

• "Create Deployment" button moved into an "Actions" menu
• When an admin doesn't have permissions to any env, show a link to the admin section.
• Add a link from the env admin page to env user management.

Backend image size​

The backend Docker image size was reduced from ~2GB to ~500MB by compiling Platz in release mode.

K8s resources are tracked by a separate task per cluster in k8s-agent.

When configuring multiple clusters, resources associated with deployments might sometime disappear and re-appear.

The reason for this was that every time a cluster tracker restarted, it tried to clean old k8s resources, but cleaned up "old" resources from all clusters.

This was fixed by associating k8s resources with their cluster_id, as they were only associated with a deployment_id up until now.

The cluster_id property is exposed via API as well.

Added: create_namespace Variable to Terraform Module​

Thanks to @DuckInCyber for reporting this issue!

When installing Platz using Terraform in an existing namespace, you can specify create_namespace=false in the main Terraform module.

Added: name_prefix Variable to Terraform Modules​

Thanks to @DuckInCyber for reporting this issue!

Names such as IAM roles were generated in the k8s-agent-role and chart-discovery modules using the k8s_namespace variable.

However, this didn't allow for defining multiple roles using the same namespace when trying to install Platz on different clusters using the same namespace.

name_prefix is now used for this prefix.

Since name_prefix and k8s_namespace both default to platz, this has no effect on existing deployments.

The api.oidc section in Platz's Helm chart values was moved to auth.oidc.

New: Admin Emails​

Thanks a lot to @vmalloc for reporting this issue!

This allows for specifying email addresses of users that will automatically get an admin role when logging into Platz.

The main use for this is in new installations.

New users that are successfully identified against the OIDC provider are created as admins if their email is in the admins list.

After installing a new Platz installation and logging-in, the admin emails can be safely removed.

If the admin emails are left as they are and a user is demoted to a regular user role, they will never become admins using this mechanism. Only another admin can promote them to become admins again.

Admin Emails in Helm Chart​

To specify admin emails in the Helm chart, pass the email addresses to auth.adminEmails:

# values.yaml

auth:
  adminEmails:
    - alice@example.com
    - bob@example.com

Admin Emails in Terraform​

The Terraform main module gets an admin_emails variable as a list(str) type.

module "platz" {
  ...
  admin_emails = [
    "alice@example.com",
    "bob@example.com",
  ]
}

New: Helm Tag Format​

Helm tags are parsed when a new chart is pushed to extract information such as the Git branch.

This allows the Platz frontend to understand when a new chart is a candidate for upgrade by deducing that a push to the same branch is probably a newer version.

The new feature allows site admins to add new tag formats, which also means parsing Helm tags has moved to the backend. Information about the parsed tag is available through the /api/v2/helm-charts API.

Tag Format Pattern​

Platz expects tag formats to be Perl-style regular expressions, with the following named groups:

• version
• revision
• branch
• commit

version is used as the name of the version installed for a deployment. revision is usually used to signal minor changes for a specific version, so it's displayed as version-revision even if the tag contains them separately.

When the branch group contains a non-empty result, Platz uses it to display a branch name, where master and main get a special primary-color formatting.

When commit contains a non-empty result, it's displayed along the parsed version name.

Built-In Formats​

The following two tag formats are added automatically when upgrading to this release:

This format matched what you'd normally get from git describe --tags, with the branch name appended to the end of the result:

^(chart-)?v?(?P<version>\d+\.\d+\.\d+)((-(?P<revision>\d+))?(-g(?P<commit>[0-9a-zA-Z]+))?-(?P<branch>[-\w]+))?$

This format matches what a lot of common GitHub Actions plugins generate when auto-tagging a release:

^(chart-)?v?(?P<version>\d+\.\d+\.\d+)-(?P<branch>[-_A-Za-z0-9]+)\.(?P<revision>\d+)$

The chart- optional prefix exists for Platz users of version 0.2 and below, where Helm charts might have been hosted in the same repo as the Docker images, forcing users to push charts with a custom prefix.

When Tags Are Parsed​

Platz scans all Helm charts when the platz-chart-discovery worker starts, and whenever a new tag format is added or an older one is deleted.

This ensures tags are parsed using all formats, from the newest format to the oldest, which allows users to add new formats that may also be matched by older ones, then deleting the old formats if they produce the wrong result.

Also, helm charts are not updated if parsing produces the same result -- being the same tag format as well -- to avoid pushing a lot of updates to all users connected via a websocket.

Note that "same result" also includes the tag format ID itself, as every Helm chart row contains a reference to the tag format that have matched it. If you add a new tag format that matches all charts, it means all charts will be updated, which shouldn't happen very often.

Helm Chart Changes​

New: Support for LetsEncrypt with cert-manager​

If you have cert-manager installed, the chart can create an Issuer and Certificate resources for you:

# In values.yaml
certManager:
  certificate:
    create: false
  issuer:
    create: false
    name: ""
    email: ""
    kind: Issuer
    group: cert-manager.io

When certManager.issuer.create is true, an Issuer resource will be created using LetsEncrypt's production server.

certManager.issuer.name is optional and uses the chart name if not provided.

certManager.issuer.email is required and will be used to register the issuer.

When certManager.certificate.create is true, a Certificate resource will be created using the secretName passed into ingress.tls[0].secretName.

It's also possible to pass certManager.issuer.name without certManager.issuer.create. This would use an existing Issuer. If you set certManager.issuer.kind to ClusterIssuer an existing ClusterIssuer would be used.

Fixed: Resource Definitions​

This version adds a resources section for each container.

If you're running on a very small node, such as Minikube, you can override resources to be empty for experimenting with Platz. It's recommended to leave the resources as they are in production.

Terraform Modules​

New: ingress Variable​

The previous Helm chart had two variables for defining an ingress: domain and tls_secret_name.

If domain was defined, and ingress was created. If tls_secret_name was defined, a tls section was added to the ingress using the domain variable:

# v0.3.4
module "platz" {
  domain          = "platz.example.com"
  tls_secret_name = "tls-secret"
  ...
}

The module now contains an ingress variable grouping everything related to the ingress, which includes cert-manager support and className:

# v0.3.5
module "platz" {
  ingress = {
    host       = "platz.example.com"
    class_name = "nginx"
    tls = {
      secret_name        = "tls-cert"
      create_certificate = true
      create_issuer      = true
      issuer_email       = "acme@example.com"
    }
  }
  ...
}

Changed: chart-discovery Outputs​

The output names of the chart-discovery module have been changed to match the inputs expected by the main module.

Before the change the main module block looked like this:

# v0.3.4
module "platz" {
  ...
  chart_discovery = {
    iam_role_arn = module.platz_chart_discovery.iam_role_arn
    queue_name   = module.platz_chart_discovery.sqs_queue_name
    queue_region = module.platz_chart_discovery.sqs_queue_region
  }
}

The sqs_queue_name and sqs_queue_region outputs have been named to queue_name and queue_region correspondingly, allowing the module resource to be used as an input directly:

# v0.3.5
module "platz" {
  ...
  chart_discovery = module.platz_chart_discovery
}