Files
OpenShell/deploy/helm/openshell/README.md.gotmpl
Brandon Squizzato cc780d4e17 feat(helm): add BackendTLSPolicy support (#2728)
* feat(helm): add optional BackendTLSPolicy for e2e TLS

Add grpcRoute.backendTLSPolicy values to optionally create a
BackendTLSPolicy resource that enables end-to-end TLS between the
Gateway proxy and the OpenShell gateway pod. The Gateway proxy
terminates client-facing TLS and re-encrypts when connecting to the
backend, validating the pod's certificate against a user-supplied CA
ConfigMap.

This removes the requirement to set server.disableTls=true when using
HTTPS at the Gateway listener. Supported on OpenShift 4.22+ and other
platforms with BackendTLSPolicy support in the Gateway API
implementation.

Update OpenShift and ingress documentation with e2e TLS instructions.

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

* feat(helm,server): auto-create backend CA ConfigMap in certgen hook

Extend the generate-certs command with --backend-ca-configmap-name and
--backend-ca-source-secret flags. When BackendTLSPolicy is enabled, the
certgen pre-install hook creates the CA ConfigMap automatically:

- pkiInitJob mode (default): uses the CA from the generated PKI bundle.
  Fully automatic on first install.
- cert-manager mode: reads ca.crt from the server TLS Secret. On first
  install the Secret does not exist yet (cert-manager reconciles after
  templates are applied), so the ConfigMap is created on the first helm
  upgrade. Logs a warning on the initial skip.

The caCertificateConfigMapName value now defaults to <fullname>-backend-ca
when empty, so users only need to set backendTLSPolicy.enabled=true.

Update certgen RBAC to include configmaps get/create when the feature is
enabled. Add CLI arg parsing tests for the new flags.

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

* refactor(helm): add server.tls.enableMtls flag for mTLS control

Replace automatic mTLS disabling based on BackendTLSPolicy with an
explicit server.tls.enableMtls flag that defaults to true. The user is
now responsible for setting this to false when using BackendTLSPolicy,
as ingress proxies cannot present client certificates to backends.

Updated:
- values.yaml: Added server.tls.enableMtls (default true)
- gateway-config.yaml: Check enableMtls instead of backendTLSPolicy
- _gateway-workload.tpl: Check enableMtls for client CA mount
- Tests: Updated to use enableMtls flag
- Docs: Added enableMtls=false to BackendTLSPolicy examples
- README: Document new flag and BackendTLSPolicy requirement

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

* docs(helm): clarify cert-manager backend CA ConfigMap workflow

Update documentation to explain the two-step install process required
when using cert-manager with BackendTLSPolicy:

1. helm install - cert-manager issues the server certificate, but the
   certgen hook can't create the backend CA ConfigMap yet (cert-manager
   reconciles after templates are applied)
2. helm upgrade - certgen hook reads the CA from the cert-manager-issued
   certificate and creates the ConfigMap

Previously, the docs said "created on first upgrade" without explaining
why or that the feature won't work until then. The updated docs now:
- Explain the timing issue (cert-manager reconciles after chart install)
- Provide clear steps for the cert-manager workflow
- Note that pkiInitJob (default) creates it immediately on install
- Clarify that users must wait for the Certificate to be Ready before
  running the second upgrade

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

* fix(docs): remove incorrect external hostname requirement for BackendTLSPolicy

BackendTLSPolicy validates the backend certificate against the service FQDN
(e.g., openshell.openshell.svc.cluster.local), not the external hostname.
The external hostname only needs to be on the Gateway listener certificate
for client-facing TLS.

The default certManager.serverDnsNames already includes all required service
FQDN variants, so no configuration is needed for BackendTLSPolicy to work.

Fixed incorrect documentation that claimed:
- "The server certificate SAN list must include the external hostname"
- Users need to "configure certManager.serverDnsNames with the external hostname"

Removed the unnecessary pkiInitJob.serverDnsNames override from the example
and clarified that:
- Gateway listener certificate needs the external hostname (for clients)
- Backend certificate needs the service FQDN (for Gateway proxy)
- The service FQDN is already in the defaults

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

* docs: clarify ACME with LetsEncrypt reference

Change all references from "ACME issuer" to "LetsEncrypt/ACME issuer"
to help users understand that LetsEncrypt is the most common ACME
provider and what ACME means in practice.

Updated:
- docs/kubernetes/managing-certificates.mdx
- docs/kubernetes/openshift.mdx
- deploy/helm/openshell/values.yaml
- deploy/helm/openshell/README.md
- deploy/helm/openshell/ci/values-openshift-route-cert-manager.yaml

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

* docs(openshift): restructure end-to-end TLS options and clarify Gateway hostname

Reorganize the OpenShift production deployment documentation:

1. Changed main section from "Production Deployments" to "Options for
   end-to-end TLS" for better clarity

2. Renamed subsections for consistency and clarity:
   - "End-to-end TLS using Gateway API and BackendTLSPolicy (OpenShift 4.22+)"
   - "End-to-end TLS using pass-through Route (all OpenShift versions)"

3. Clarified that the Gateway hostname is typically a wildcard:
   "typically a wildcard like *.openshell-ingress-gw.example.com"

4. Removed the recommendation to copy the cluster's wildcard certificate
   from openshift-ingress namespace, as this is not a recommended
   security best practice

These changes make it clearer that users have two end-to-end TLS options
and help them understand the typical naming pattern for Gateway hostnames.

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

* feat(helm): eliminate two-stage install for BackendTLSPolicy with cert-manager

When using BackendTLSPolicy with cert-manager, the certgen hook now polls
for up to 90 seconds waiting for cert-manager to issue the TLS certificate
before creating the backend CA ConfigMap. This eliminates the need for a
second `helm upgrade` in most cases.

The hook polls every 2 seconds with progress logging every 10 seconds.
If cert-manager takes longer than 90 seconds, the hook times out gracefully
and logs a warning, preserving the fallback to manual ConfigMap creation
or a second upgrade.

The Job's activeDeadlineSeconds is 120s, so the 90s timeout leaves 30s
margin for ConfigMap creation and hook completion.

Signed-off-by: Brandon Squizzato <bsquizzato@nvidia.com>

* feat(helm): add configurable timeout for certgen hook

Add `pkiInitJob.timeoutSeconds` Helm value (default 120) to control how long
the certgen hook Job can run. When using cert-manager with BackendTLSPolicy,
the hook polls for (timeoutSeconds - 30) seconds to leave margin for ConfigMap
creation and cleanup.

This allows users to increase the timeout for environments where cert-manager
takes longer than 90 seconds to issue certificates, without requiring code
changes.

Example usage:
```yaml
pkiInitJob:
  timeoutSeconds: 180  # Hook polls for 150 seconds
```

Signed-off-by: Brandon Squizzato <bsquizzato@nvidia.com>

* docs(helm): document configurable certgen timeout

Update documentation to mention the pkiInitJob.timeoutSeconds value and
how it affects the cert-manager polling behavior when using BackendTLSPolicy.

Signed-off-by: Brandon Squizzato <bsquizzato@nvidia.com>

* feat(helm): add configurable failure behavior for certgen timeout

Add `pkiInitJob.failOnTimeout` Helm value (default false) to control whether
the certgen hook fails or succeeds when cert-manager does not issue a
certificate within the polling timeout.

When false (default), the hook succeeds with a warning and users can run
`helm upgrade` after cert-manager issues the certificate to create the
backend CA ConfigMap. This provides backwards-compatible behavior.

When true, the hook fails immediately if the timeout is reached, providing
clear feedback that BackendTLSPolicy is non-functional. This is useful for
strict validation requirements where incomplete installs should fail fast.

Example usage:
```yaml
pkiInitJob:
  timeoutSeconds: 180
  failOnTimeout: true  # Fail install if cert-manager takes >150s
```

Signed-off-by: Brandon Squizzato <bsquizzato@nvidia.com>

* feat(helm): change failOnTimeout default to true and add troubleshooting docs

Change `pkiInitJob.failOnTimeout` default from false to true to provide
immediate feedback when cert-manager does not issue certificates within
the polling timeout. This prevents silent failures where BackendTLSPolicy
is non-functional but the install appears to succeed.

Add comprehensive troubleshooting section to docs/kubernetes/ingress.mdx
documenting the specific error "TLS error: Secret is not supplied by SDS"
that occurs when the backend CA ConfigMap is missing, with step-by-step
resolution instructions.

Updated comments in values.yaml to clearly document the default behavior
and explain when administrators might see connectivity errors if they
override the default to failOnTimeout=false.

BREAKING CHANGE: pkiInitJob.failOnTimeout now defaults to true. Helm
installs will fail if cert-manager takes longer than (timeoutSeconds - 30)
seconds to issue certificates. To restore the old behavior of allowing
installs to succeed with a warning, set `pkiInitJob.failOnTimeout=false`.

Signed-off-by: Brandon Squizzato <bsquizzato@nvidia.com>

* fix(helm): make cert-manager resources pre-install hooks to fix ordering

Make Certificate and Issuer resources run as pre-install/pre-upgrade hooks
with weight -30, before the certgen hook (weight -20). This fixes the
chicken-and-egg problem where the certgen hook was waiting for Secrets
created by Certificates that hadn't been created yet.

**Hook ordering:**
1. Certificate and Issuer resources created (weight -30)
2. cert-manager issues certificates and creates Secrets
3. certgen hook runs (weight -20), finds Secrets, creates ConfigMap
4. Main resources (StatefulSet, Service, etc.) created

Previously, the certgen pre-install hook would run before any resources
were created, poll for a non-existent Secret, timeout, and fail. The
Certificate resources would never get created because Helm waits for
all pre-install hooks to succeed before creating main resources.

This fix allows single-stage installs to work reliably as long as
cert-manager can issue certificates within the polling timeout.

Signed-off-by: Brandon Squizzato <bsquizzato@nvidia.com>

* feat(helm): add validation to prevent enableMtls with BackendTLSPolicy

Add Helm chart validation that fails the install if both
server.tls.enableMtls=true and grpcRoute.backendTLSPolicy.enabled=true
are set, since this is an invalid configuration.

BackendTLSPolicy requires mTLS to be disabled because the Gateway proxy
cannot present client certificates to the backend. This validation provides
immediate, clear feedback at install time rather than allowing the
misconfiguration to be discovered through runtime errors.

Example error message:
```
Error: grpcRoute.backendTLSPolicy requires mTLS to be disabled because
the Gateway proxy cannot present client certificates to the backend;
set server.tls.enableMtls=false
```

Also updated documentation to mention this validation check.

Signed-off-by: Brandon Squizzato <bsquizzato@nvidia.com>

* docs(helm): clarify pkiInitJob.timeoutSeconds polling behavior

Improve documentation to clearly explain that pkiInitJob.timeoutSeconds
controls the Job deadline, but the actual polling timeout is
(timeoutSeconds - 30) to reserve 30 seconds for ConfigMap creation
and cleanup.

Added concrete example: "timeoutSeconds=180 allows 150 seconds of polling"
to make the relationship explicit and avoid confusion where users might
expect the hook to poll for the full timeout value.

Updated both values.yaml inline comments and ingress.mdx documentation
for consistency.

Signed-off-by: Brandon Squizzato <bsquizzato@nvidia.com>

* docs(openshift): remove outdated two-stage install instructions

Update OpenShift documentation to reflect that single-stage installs now
work with cert-manager and BackendTLSPolicy. The Certificate resources
run as pre-install hooks (weight -30) before certgen (weight -20),
allowing the hook to poll for and find the issued certificates.

Removed the outdated two-step process:
1. helm install (cert-manager issues cert, hook logs warning)
2. helm upgrade (hook creates ConfigMap)

Replaced with current single-stage behavior:
- Certificate resources created as pre-install hooks
- certgen hook polls for up to 90 seconds (configurable)
- Single helm install succeeds in most cases
- Fails fast by default if timeout reached

This brings openshift.mdx in line with the already-updated ingress.mdx
documentation.

Signed-off-by: Brandon Squizzato <bsquizzato@nvidia.com>

* fix(helm): make pkiInitJob.timeoutSeconds the actual polling duration

The timeout value now represents the actual polling time that users
experience when waiting for cert-manager to issue certificates.
The Job activeDeadlineSeconds is set to (timeoutSeconds + 30) to
allow buffer time for ConfigMap creation and cleanup.

Previously, the hook polled for (timeoutSeconds - 30) seconds, which
was confusing when users set timeoutSeconds=180 and only got 150
seconds of actual polling.

Updated documentation in values.yaml, ingress.mdx, and openshift.mdx
to reflect the clearer behavior.

Signed-off-by: Brandon Squizzato <bsquizza@nvidia.com>

* docs(helm): update values.yaml and README with correct polling duration

Updated the caCertificateConfigMapName description to reflect that the
hook polls for exactly pkiInitJob.timeoutSeconds seconds, not
(timeoutSeconds - 30) seconds.

Regenerated README.md with helm-docs.

Signed-off-by: Brandon Squizzato <bsquizza@nvidia.com>

* docs(kubernetes): add OIDC configuration to helm install and CLI examples

Updated all helm install and openshell gateway add examples in ingress.mdx
and openshift.mdx to include OIDC issuer and audience configuration.

Examples now use concrete placeholder values:
- OIDC issuer: https://keycloak.example.com/realms/openshell
- OIDC audience: openshell-cli
- Hostname: gateway.example.com
- ClusterIssuer: letsencrypt-prod

This makes it clearer how to configure OIDC authentication, which is
required when using BackendTLSPolicy or HTTPS termination since the
Gateway proxy cannot present client certificates to the backend.

Signed-off-by: Brandon Squizzato <bsquizza@nvidia.com>

* docs(kubernetes): explicitly list OIDC client ID in gateway add examples

Added --oidc-client-id openshell-cli to all openshell gateway add
commands in ingress.mdx and openshift.mdx, making the default client
ID explicit in the examples even though it's the CLI default.

This improves clarity and helps users understand the complete OIDC
configuration needed for gateway registration.

Signed-off-by: Brandon Squizzato <bsquizza@nvidia.com>

* fix(helm): address PR review feedback for BackendTLSPolicy

- Read backend CA from the authoritative server Secret instead of the
  in-memory PKI bundle so enabling BackendTLSPolicy on an existing
  release uses the CA that actually signed the server certificate.
- Reconcile the backend CA ConfigMap on every hook run (compare and
  update) instead of skipping when it already exists, so CA rotations
  propagate automatically.
- Remove hook annotations from cert-manager Issuer/Certificate resources
  so they remain regular release objects managed by Helm lifecycle. Split
  the cert-manager backend CA ConfigMap creation into a separate
  post-install/post-upgrade hook Job that polls after cert-manager
  Certificate resources are applied.
- Update architecture/gateway.md, docs/reference/gateway-config.mdx,
  debug-openshell-cluster skill, and helm-dev-environment skill with
  BackendTLSPolicy, backend CA ConfigMap, enableMtls, and timeout
  documentation.

Signed-off-by: Brandon Squizzato <bsquizza@nvidia.com>
Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

* fix(docs): resolve markdown lint errors in helm README and kubernetes docs

Escape inline HTML angle brackets in README.md template placeholders,
remove trailing spaces, and add blank lines around fenced code blocks
in numbered lists.

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

* Update docs

* fix(helm): escape inline HTML in values.yaml descriptions and sync mise lockfile

Wrap `<fullname>` and `<namespace>` template placeholders in backticks
so markdownlint does not flag them as inline HTML (MD033). Regenerate
mise.lock to match current mise.toml after rebase onto main.

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

* Run 'mise lock'

* fix: align mise.lock with CI mise version output

The lockfile was regenerated locally with mise 2026.8.10 which resolves
uv Linux artifacts to gnu variants and adds provenance_verified fields,
but CI uses v2026.4.25 which produces musl variants without those fields.

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

* fix(docs): correct cert-manager hook ordering and clientCaSecretName comment

Update ingress.mdx and openshift.mdx to describe Certificate resources
as regular release objects with a post-install/post-upgrade Job, matching
the current implementation and architecture/gateway.md.

Fix values.yaml clientCaSecretName comment to state that "" disables
client certificate verification, matching the helper and access-control
docs.

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>

---------

Signed-off-by: Brandon Squizzato <bsquizza@redhat.com>
Signed-off-by: Brandon Squizzato <bsquizzato@nvidia.com>
Signed-off-by: Brandon Squizzato <bsquizza@nvidia.com>
2026-09-14 17:39:15 +00:00

176 lines
7.5 KiB
Go Template

# OpenShell Helm Chart
<!--
This file is generated by helm-docs.
Edit README.md.gotmpl and values.yaml, then run `mise run helm:docs`.
-->
> **Experimental** - the Kubernetes deployment path is under active development. Expect rough edges and breaking changes.
This chart deploys the OpenShell gateway into a Kubernetes cluster. It is published as an OCI artifact to GHCR at `oci://ghcr.io/nvidia/openshell/helm-chart`.
By default, this chart also creates the namespace-scoped resources needed by
sandboxes. For a shared-gateway deployment, install it with
`workspaceResources.enabled=false`, then install the
`deploy/helm/openshell-workspace` chart in every pre-provisioned workspace
namespace. The gateway and workspace releases can then be upgraded and removed
independently. Use Kubernetes `operator` workspace mode when one gateway serves
multiple pre-provisioned workspace namespaces.
## Prerequisites
The Kubernetes Agent Sandbox CRDs and controller must be installed on the cluster before deploying OpenShell. Install them with:
```shell
kubectl apply -f https://github.com/kubernetes-sigs/agent-sandbox/releases/latest/download/manifest.yaml
```
The chart does not install this cluster-scoped dependency. By default, it
fails before creating gateway resources when the cluster serves neither
supported Sandbox API (`agents.x-k8s.io/v1beta1` or
`agents.x-k8s.io/v1alpha1`). Disable the check with
`agentSandbox.preflight.enabled=false` for offline `helm template` rendering,
where Helm cannot discover cluster APIs.
## Install on Kubernetes
```shell
helm install openshell oci://ghcr.io/nvidia/openshell/helm-chart --version <version>
```
## Install on OpenShift
See the full [OpenShift install guide](https://docs.nvidia.com/openshell/latest/kubernetes/openshift) for details. Quick start:
```shell
# Precreate the openshell namespace so we can create the SCC cluster role
oc create ns openshell
# Sandboxes are deployed into the openshell namespace and use the openshell-sandbox service account
oc adm policy add-scc-to-user privileged -z openshell-sandbox -n openshell
# Deploy openshell with overrides to allow SCC assignment of fsGroup and runAsUser for the gateway
helm install openshell oci://ghcr.io/nvidia/openshell/helm-chart --version <version> -n openshell \
--set server.disableTls=true \
--set podSecurityContext.fsGroup=null \
--set securityContext.runAsUser=null
```
On OpenShift 4.22+, end-to-end TLS is supported via `BackendTLSPolicy`. See the
[OpenShift install guide](https://docs.nvidia.com/openshell/latest/kubernetes/openshift#end-to-end-tls-openshift-422) for details.
## Available versions
| Tag | Source | Notes |
| --- | --- | --- |
| `<semver>` (e.g. `0.6.0`) | Tagged GitHub release | Tracks the matching gateway and supervisor image versions. Recommended for production. |
| `0.0.0-dev` | Latest commit on `main` | Floating tag, overwritten on every push. `appVersion` is `dev`, so images resolve to the `:dev` tag. |
| `0.0.0-dev.<commit-sha>` | A specific commit on `main` | Per-commit pin. Chart version and `appVersion` both use the full 40-character commit SHA, which matches the image tag pushed by CI. |
The `dev` tags are intended for testing changes ahead of a release. Production deployments should pin to a tagged release.
## Configuration
See [`values.yaml`](values.yaml) for source defaults. Selected overlays:
- [`ci/values-gateway.yaml`](ci/values-gateway.yaml) - gateway-only configuration
- [`ci/values-cert-manager.yaml`](ci/values-cert-manager.yaml) - cert-manager integration
- [`ci/values-keycloak.yaml`](ci/values-keycloak.yaml) - Keycloak OIDC integration
- [`ci/values-high-availability.yaml`](ci/values-high-availability.yaml) - CI overlay for multi-replica external PostgreSQL testing
- [`ci/values-spire.yaml`](ci/values-spire.yaml) - SPIFFE/SPIRE provider token grants
- [`ci/values-spire-stack.yaml`](ci/values-spire-stack.yaml) - SPIRE hardened chart values for local development
### Database backend
By default, OpenShell uses SQLite and runs the gateway as a StatefulSet so the
database is backed by a per-pod PVC:
```yaml
server:
dbUrl: "sqlite:/var/openshell/openshell.db"
```
#### External PostgreSQL
Use external PostgreSQL when the gateway should connect to a database managed
outside this chart. The OpenShell chart does not deploy a database; install
PostgreSQL separately using the chart, operator, or managed service that fits
your environment, then pass the connection URI through a Secret.
Create a Secret containing the PostgreSQL connection URI if one does not
already exist:
```bash
kubectl create secret generic my-pg-credentials -n openshell \
--from-literal=uri="postgresql://user:pass@host:5432/dbname"
```
Then install the chart pointing at that Secret:
```bash
helm install openshell oci://ghcr.io/nvidia/openshell/helm-chart --version <version> \
-n openshell \
--set workload.kind=deployment \
--set server.externalDbSecret=my-pg-credentials
```
Use `workload.kind=deployment` for external database-backed multi-replica
gateways. `workload.kind=statefulset` is still available for single-replica
SQLite installs and for operators who explicitly need StatefulSet identity or
storage semantics.
### Credential storage
By default, the chart uses the gateway's encrypted database credential storage.
The gateway writes encrypted provider credential envelopes to the OpenShell
database. The chart creates a retained Kubernetes Secret with the shared
key-encryption key and injects that key into every gateway pod, so the same
default works for single-replica and external database-backed HA deployments.
Use `kubernetes-secrets` or `vault` instead when credentials should live in a
cluster or external secret backend. Enabling one external credential driver
disables the default credential-storage key-encryption key Secret and env injection.
#### OpenShift
Append these flags to any of the PostgreSQL commands above for OpenShift:
```
--set server.disableTls=true \
--set podSecurityContext.fsGroup=null \
--set securityContext.runAsUser=null
```
## Secret bootstrap
By default, a pre-install/pre-upgrade hook Job runs `openshell-gateway generate-certs`
to create the gateway's server/client mTLS Secrets and sandbox JWT signing Secret.
The Job uses the gateway image itself, so air-gapped environments only need to
mirror that one image (no separate openssl/alpine sidecar).
When `certManager.enabled=true`, cert-manager owns the TLS Secrets and the chart
runs the same hook in JWT-only mode because cert-manager does not create the
sandbox JWT signing Secret. This precedence applies even if
`pkiInitJob.enabled` remains true. Set `pkiInitJob.enabled=false` only when an
external non-cert-manager TLS source manages TLS and you pre-create the sandbox
JWT signing Secret.
## SPIFFE/SPIRE provider token grants
Set `server.providerTokenGrants.spiffe.enabled=true` to let the gateway and
sandbox supervisors use SPIFFE JWT-SVIDs for dynamic provider token grants. The
chart keeps supervisor-to-gateway authentication on gateway-minted sandbox JWTs,
mounts the SPIFFE CSI socket into the gateway pod, exports
`OPENSHELL_GATEWAY_SPIFFE_WORKLOAD_API_SOCKET`, and passes the socket path to
the Kubernetes driver so sandbox pods can mount the same socket.
For local development, uncomment the SPIRE Helm releases in `skaffold.yaml` and
add `ci/values-spire.yaml` to the OpenShell release values files.
The gateway verifies supervisor JWT-SVIDs with JWT bundles fetched from the
SPIFFE Workload API, so this path does not require access to the SPIRE OIDC
discovery endpoint or its TLS CA.
{{ template "chart.valuesSection" . }}
{{ template "helm-docs.versionFooter" . }}