Helm values¶
Every value of the openbao-config-operator chart, with its type, default and meaning.
Source of truth
This page is derived from charts/openbao-config-operator/values.yaml (defaults and # --
comments) and charts/openbao-config-operator/values.schema.json (types and constraints).
Those two files are canonical; when they disagree with this page, they win. The chart's
README.md has a shorter table in the same format.
The schema sets additionalProperties: false on the top level and on most sections, so Helm
refuses a misspelt or unknown key at install time. Besides the schema, the templates refuse these
combinations:
| combination | error |
|---|---|
replicas greater than 1 with leaderElection.enabled: false |
replicas > 1 requires leaderElection.enabled=true |
injectedIdentity.projectedToken.path under /var/run/secrets/kubernetes.io/serviceaccount/ |
the automounted token directory is refused |
crds.install: false with crds.keep: false |
an upgrade would delete the CRDs of an earlier release and every custom resource with them |
metrics.certSecret with serviceMonitor.enabled, without serviceMonitor.tlsConfig.ca (or caFile) and serverName, or with insecureSkipVerify: true |
the ServiceMonitor cannot verify the certificate |
String values that land in a manifest (names, image tag, audiences, CIDRs, durations, the token path) must match a pattern without whitespace, so a value cannot add YAML keys to the rendered objects.
Names, CRDs and image¶
| key | type | default | description |
|---|---|---|---|
nameOverride |
string | "" |
override the chart name used in resource names |
fullnameOverride |
string | "" |
override the full resource name prefix |
crds.install |
boolean | true |
install and upgrade the CRDs with the release |
crds.keep |
boolean | true |
annotate the CRDs with helm.sh/resource-policy: keep, so uninstalling the release does not delete every custom resource |
crds.annotations |
map of strings, without controller-gen.kubebuilder.io/version |
{} |
extra annotations on every CRD, for example argocd.argoproj.io/sync-options: ServerSideApply=true. With crds.keep: true, a helm.sh/resource-policy here is replaced by keep; with crds.keep: false it is used as given |
crds.labels |
map of strings, without the chart's own label keys | {} |
extra labels on every CRD. helm.sh/chart, app.kubernetes.io/name, app.kubernetes.io/instance, app.kubernetes.io/version and app.kubernetes.io/managed-by are refused |
pluginKinds.enabled |
boolean | false |
install the CRDs of kinds that need an external OpenBao plugin (cloud auth and secrets, GitHub, Okta, LDAP, Kerberos, RADIUS, Consul, Nomad, Quay, Terraform Cloud, MongoDB Atlas) and pass --enable-plugin-kinds |
image.repository |
string, non-empty | ghcr.io/alexandrust88/openbao-config-operator |
manager image repository |
image.tag |
string | "" |
image tag; empty uses the chart appVersion |
image.digest |
string, sha256: and 64 hex characters, or empty |
"" |
when set, the image is pulled by digest. The released chart sets it |
image.pullPolicy |
Always, IfNotPresent or Never |
IfNotPresent |
|
imagePullSecrets |
list of objects | [] |
image pull secrets, as a list of {name: ...} |
Manager¶
| key | type | default | description |
|---|---|---|---|
replicas |
integer, at least 0 | 2 |
manager replicas; more than one requires leaderElection.enabled |
leaderElection.enabled |
boolean | true |
passes --leader-elect; disable only with replicas: 1 |
watchNamespaces |
list of strings | [] |
namespaces the manager watches (--watch-namespaces); empty means cluster-wide. When set, access to Secrets, ConfigMaps, ServiceAccount tokens, Events and every namespaced managed kind is granted through a Role in each listed namespace instead of a ClusterRole; the release namespace, unless listed, gets only the hash-key Secret |
apiServerAudiences |
list of strings | [] |
passes --apiserver-audiences: audiences the Kubernetes API server accepts besides the built-in ones (https://kubernetes.default.svc, https://kubernetes.default.svc.cluster.local, kubernetes.default.svc, rke2, k3s) and the discovered service account issuer. A tenant OpenBaoConnection is never given a token for them |
health.port |
integer, 1 to 65535 | 8081 |
port of /healthz and /readyz |
pprof.enabled |
boolean | false |
serve Go pprof profiles on 127.0.0.1 inside the pod (no Service); reach them with kubectl port-forward |
pprof.port |
integer, 1 to 65535 | 8082 |
pprof port |
extraArgs |
list of strings | [] |
extra manager arguments. Flags the chart sets from a value (--leader-elect, --metrics-secure, --metrics-bind-address, --metrics-cert-path, --pprof-bind-address, --health-probe-bind-address, --watch-namespaces, --apiserver-audiences, --enable-plugin-kinds) and --hash-key-secret are refused |
extraEnv |
list of objects | [] |
extra manager environment variables; POD_NAMESPACE is refused because the chart sets it |
extraVolumes |
list of objects | [] |
extra volumes, for example a CA bundle ConfigMap trusted through SSL_CERT_DIR in extraEnv; hostPath volumes are refused |
extraVolumeMounts |
list of objects | [] |
extra volume mounts of the manager container |
Manager flags without a chart value, such as --drift-interval-critical,
--drift-interval-standard, --max-concurrent-reconciles,
--max-concurrent-reconciles-high-volume, --write-budget-per-minute, --cluster-id,
--hash-key-secret and the --zap-* logging flags, go through extraArgs.
Injected identity¶
| key | type | default | description |
|---|---|---|---|
injectedIdentity.projectedToken.enabled |
boolean | true |
mount a projected ServiceAccount token with its own audience for ClusterOpenBaoConnection injectedIdentity logins. The operator never sends the automounted token, so a connection with injectedIdentity needs this or another projected token |
injectedIdentity.projectedToken.audience |
string, non-empty | openbao-config-operator |
token audience; the OpenBao kubernetes or jwt role must accept it |
injectedIdentity.projectedToken.expirationSeconds |
integer, at least 600 | 3600 |
token lifetime in seconds; the kubelet rotates the file before it expires |
injectedIdentity.projectedToken.path |
string, absolute path of at least two segments | /var/run/secrets/openbao/token |
token file in the manager container; its directory must not hold other mounts. It is also the CRD's default tokenPath; if you change it, set the connection's tokenPath to the same file |
ServiceAccount and RBAC¶
| key | type | default | description |
|---|---|---|---|
serviceAccount.create |
boolean | true |
create the ServiceAccount |
serviceAccount.name |
string | "" |
ServiceAccount name; defaults to the full name; required when create is false; default is refused, so the manager rights never land on the namespace's default ServiceAccount |
serviceAccount.annotations |
map of strings | {} |
|
rbac.create |
boolean | true |
create the manager ClusterRole, Roles and bindings |
rbac.userRoles |
boolean | true |
create the user-facing <release>-tenant-editor and <release>-viewer ClusterRoles |
rbac.tenantEditor.excludeKinds |
list of plurals | [] |
kinds to leave out of <release>-tenant-editor, for example openbaoendpoints to keep the raw-path kind for platform admins (ADR 0006). The viewer never includes OpenBaoServer or ClusterOpenBaoConnection |
rbac.aggregate.admin |
boolean | true |
aggregate the tenant-editor role into the built-in admin role |
rbac.aggregate.edit |
boolean | true |
aggregate the tenant-editor role into the built-in edit role |
rbac.aggregate.view |
boolean | true |
aggregate the viewer role into the built-in view role |
Metrics and monitoring¶
| key | type | default | description |
|---|---|---|---|
metrics.enabled |
boolean | true |
serve metrics over HTTPS, protected by Kubernetes authentication and authorization, and create the metrics Service |
metrics.port |
integer, 1 to 65535 | 8443 |
metrics port |
metrics.certSecret |
string | "" |
Secret (tls.crt, tls.key) with the metrics serving certificate; empty uses a self-signed certificate |
metrics.service.type |
ClusterIP |
ClusterIP |
only ClusterIP: the metrics endpoint is for in-cluster scrapers |
metrics.service.annotations |
map of strings | {} |
|
serviceMonitor.enabled |
boolean | false |
create a prometheus-operator ServiceMonitor |
serviceMonitor.labels |
map of strings | {} |
labels for your Prometheus selector |
serviceMonitor.interval |
string | 30s |
scrape interval |
serviceMonitor.scrapeTimeout |
string | 10s |
scrape timeout |
serviceMonitor.tlsConfig |
object | {insecureSkipVerify: true} |
TLS config of the scrape endpoint. insecureSkipVerify fits only the default self-signed certificate; with metrics.certSecret the chart requires ca (or caFile) and serverName and refuses insecureSkipVerify |
serviceMonitor.relabelings |
list of objects | [] |
|
serviceMonitor.metricRelabelings |
list of objects | [] |
|
prometheusRule.enabled |
boolean | false |
create a prometheus-operator PrometheusRule with the alerts in runbooks |
prometheusRule.labels |
map of strings | {} |
labels of the PrometheusRule object |
prometheusRule.selector |
string | "" |
series selector of the operator's metrics; empty means job="<fullname>-metrics" in the release namespace |
prometheusRule.alertLabels |
map of strings | {} |
labels added to every alert, for example for routing |
prometheusRule.disabledAlerts |
list of strings | [] |
alert names to leave out |
prometheusRule.reconcileErrorRatio |
number, at most 1 | 0.1 |
ratio of failed reconciles per controller above which OpenBaoConfigOperatorReconcileErrors fires |
prometheusRule.openbaoErrorRatio |
number, at most 1 | 0.05 |
ratio of OpenBao 5xx and transport errors per connection above which OpenBaoConfigOperatorOpenBaoServerErrors fires |
prometheusRule.steadyStateWritesFor |
string, a number followed by s, m or h |
30m |
how long writes must continue before OpenBaoConfigOperatorSteadyStateWrites fires |
prometheusRule.runbookBaseURL |
string, http:// or https:// URL |
https://alexandrust88.github.io/openbao-config-operator/operate/runbooks/ |
base of each alert's runbook_url; charts of v0.3.0 and earlier default to the GitHub page docs/operations/runbooks.md |
Observability explains the metrics and the alerts.
Pod settings¶
| key | type | default | description |
|---|---|---|---|
resources |
object | requests cpu: 50m, memory: 128Mi; limit memory: 512Mi |
manager container resources |
podSecurityContext |
object | runAsNonRoot: true, runAsUser: 65532, runAsGroup: 65532, seccompProfile.type: RuntimeDefault |
pod security context; runAsNonRoot: false is refused. Set runAsUser and runAsGroup to null where the platform assigns UIDs (OpenShift) |
securityContext |
object | allowPrivilegeEscalation: false, privileged: false, readOnlyRootFilesystem: true, runAsNonRoot: true, capabilities.drop: [ALL], seccompProfile.type: RuntimeDefault |
container security context. The schema requires allowPrivilegeEscalation: false, readOnlyRootFilesystem: true and runAsNonRoot: true, and refuses privileged: true and added capabilities |
livenessProbe |
object | initialDelaySeconds: 15, periodSeconds: 20 |
merged into the httpGet liveness probe |
readinessProbe |
object | initialDelaySeconds: 5, periodSeconds: 10 |
merged into the httpGet readiness probe |
podDisruptionBudget.enabled |
boolean | true |
create a PodDisruptionBudget when replicas is greater than 1 |
podDisruptionBudget.minAvailable |
integer or string | 1 |
|
podDisruptionBudget.maxUnavailable |
integer or string | "" |
used instead of minAvailable when set |
topologySpreadConstraints |
list of objects | maxSkew: 1, topologyKey: kubernetes.io/hostname, whenUnsatisfiable: ScheduleAnyway |
a missing labelSelector defaults to the pod selector labels |
priorityClassName |
string, DNS name | "" |
pod priority class |
nodeSelector |
map of strings | {} |
|
tolerations |
list of objects | [] |
|
affinity |
object | {} |
|
terminationGracePeriodSeconds |
integer, at least 0 | 10 |
|
revisionHistoryLimit |
integer, at least 0 | 10 |
|
podAnnotations |
map of strings | {} |
|
podLabels |
map of strings | {} |
|
annotations |
map of strings | {} |
annotations of the Deployment |
Network policy¶
| key | type | default | description |
|---|---|---|---|
networkPolicy.enabled |
boolean | true |
restrict ingress to the metrics port and egress to DNS, the API server and OpenBao |
networkPolicy.flavor |
kubernetes or cilium |
kubernetes |
kubernetes renders a NetworkPolicy; cilium renders a CiliumNetworkPolicy that reaches the API server through the kube-apiserver entity, which an ipBlock does not match on Cilium. metrics.from, dns.to, apiServer.cidrs and openbao.to apply to the kubernetes flavour only |
networkPolicy.metrics.monitoringNamespace |
string, a namespace name or empty | monitoring |
namespace of the metrics scraper; with an empty from, the release namespace and this namespace may scrape |
networkPolicy.metrics.from |
list of objects | [] |
NetworkPolicy peers allowed to scrape metrics, replacing the release and monitoring namespaces |
networkPolicy.dns.to |
list of objects | kube-dns pods in kube-system |
DNS peers |
networkPolicy.dns.nodeLocalDNS.enabled |
boolean | false |
also allow DNS to NodeLocal DNSCache |
networkPolicy.dns.nodeLocalDNS.address |
string, IPv4 | 169.254.20.10 |
NodeLocal DNSCache address |
networkPolicy.apiServer.cidrs |
list | [] |
API server CIDRs; empty allows any destination on the API server ports |
networkPolicy.apiServer.ports |
list | [443, 6443] |
API server ports |
networkPolicy.openbao.to |
list of objects | [] |
OpenBao peers; empty allows any destination on the OpenBao ports |
networkPolicy.openbao.fqdns |
list | [] |
OpenBao host names for the cilium flavour (toFQDNs, needs the Cilium DNS proxy); empty allows any destination on the OpenBao ports |
networkPolicy.openbao.ports |
list | [8200] |
OpenBao ports |
networkPolicy.extraEgress |
list of objects | [] |
extra egress rules, in the syntax of the selected flavour |
The hardening guide explains why
empty apiServer.cidrs and openbao.to leave egress open by destination.