Configuration Reference

Values

Key	Type	Default	Description
nameOverride	string	""	Unique identifier of SuperSONIC instance (equal to release name by default)
serverLoadMetric	string	""	Prometheus query used by the KEDA autoscaler and, per healthy replica, by Envoy’s prometheus-based rate limiter. Empty selects the default metric: the number of replicas the current in-flight work needs (see the configuration guide). A custom query is used verbatim.
serverLoadThreshold	float	1.5	Scaling threshold: KEDA scales to ceil(metric / threshold). Set fractions from a values file or –set-json. Remove a leftover serverLoadThreshold: 100 when upgrading.
serverLoadRateInterval	string	"30s"	rate() window of the default metric; keep it at or above 4x the Prometheus scrape interval.
serverAdmissionThreshold	float	3	Threshold of Envoy’s prometheus-based rate limiter (the scaling metric per healthy replica, or a custom serverLoadMetric verbatim). Envoy reads it at startup.
scaleFromZero	object	{"admissionImage":"python:3.14-slim","enabled":false,"holdMinReplicasSeconds":300,"readyTimeoutSeconds":300}	On RepositoryIndex, scale Triton to at least max(1, keda.minReplicaCount) replicas and return the index only after Envoy has a healthy Triton upstream. Requires keda.enabled and envoy.enabled.
scaleFromZero.enabled	bool	false	Enable scale from zero
scaleFromZero.readyTimeoutSeconds	int	300	Seconds to wait for a healthy Triton upstream before rejecting RepositoryIndex
scaleFromZero.holdMinReplicasSeconds	int	300	Seconds to keep KEDA minReplicaCount at the wake target after the last RepositoryIndex is answered (the hold is refreshed once the server is ready, so it is measured from readiness rather than from the wake request). Must be >= readyTimeoutSeconds (validated at render time), otherwise KEDA can scale the waking pod back to zero before RepositoryIndex is answered.
scaleFromZero.admissionImage	string	"python:3.14-slim"	Image for the admission sidecar on the Envoy pod
triton.replicas	int	1	Number of Triton server instances. Unused when scaleFromZero.enabled is true.
triton.image	string	"nvcr.io/nvidia/tritonserver:26.08-py3-min"	Docker image for the Triton server
triton.command	list	["/bin/sh","-c"]	Command and arguments to run in Triton container
triton.args[0]	string	"/opt/tritonserver/bin/tritonserver \\\n--model-repository=/tmp/ \\\n--log-verbose=0 \\\n--exit-timeout-secs=60\n"
triton.resources	object	{"limits":{"cpu":1,"memory":"2G"},"requests":{"cpu":1,"memory":"2G"}}	Resource limits and requests for each Triton instance. You can add necessary GPU request here.
triton.annotations	object	{}	Annotations for Triton pods
triton.nodeSelector	object	{}	Node selector for Triton pods
triton.tolerations	list	[]	Tolerations for Triton pods
triton.affinity	object	{}	Affinity rules for Triton pods - another way to request GPUs
triton.modelRepository	object	{"enabled":false,"mountPath":""}	Model repository configuration
triton.modelRepository.mountPath	string	""	Model repository mount path
triton.service.labels	object	{}
triton.service.annotations	object	{}
triton.service.ports	list	[{"name":"http","port":8000,"protocol":"TCP","targetPort":8000},{"name":"grpc","port":8001,"protocol":"TCP","targetPort":8001},{"name":"metrics","port":8002,"protocol":"TCP","targetPort":8002}]	Ports for communication with Triton servers. Inference sent directly to them bypasses Envoy and the scaling metric.
triton.readinessProbe	object	{"command":["/bin/sh","-c","curl -sf http://localhost:8000/v2/health/ready > /dev/null && [ ! -f /tmp/shutdown ]"],"failureThreshold":10,"initialDelaySeconds":10,"periodSeconds":10,"reset":false,"successThreshold":1,"timeoutSeconds":15}	Custom readiness probe configuration
triton.readinessProbe.reset	bool	false	If true, will reset settings to k8s defaults (other readinessProbe settings will be ignored)
triton.startupProbe	object	{"failureThreshold":24,"httpGet":{"path":"/v2/health/ready","port":"http"},"initialDelaySeconds":0,"periodSeconds":10,"reset":false,"timeoutSeconds":15}	Custom startup probe configuration
triton.startupProbe.reset	bool	false	If true, will reset settings to k8s defaults (other startupProbe settings will be ignored)
envoy.enabled	bool	true	Enable Envoy Proxy
envoy.replicas	int	1	Number of Envoy Proxy pods in Deployment
envoy.image	string	"envoyproxy/envoy:v1.39.1"	Envoy Proxy Docker image
envoy.args	list	["--config-path","/etc/envoy/envoy.yaml","--log-level","info","--log-path","/dev/stdout"]	Arguments for Envoy
envoy.external_config	object	{"configmap_key":"","configmap_name":"","load_from_configmap":false}	External Envoy configuration settings
envoy.external_config.load_from_configmap	bool	false	If true, load Envoy configuration from an external ConfigMap instead of generating it dynamically
envoy.external_config.configmap_name	string	""	Name of the external ConfigMap containing the Envoy configuration
envoy.external_config.configmap_key	string	""	Key name in the external ConfigMap (defaults to “envoy.yaml”)
envoy.resources	object	{"limits":{"cpu":8,"memory":"4G"},"requests":{"cpu":1,"memory":"2G"}}	Resource requests and limits for Envoy Proxy. Note: an Envoy Proxy with too many connections might run out of CPU
envoy.annotations	object	{}	Annotations for Envoy pods
envoy.nodeSelector	object	{}	Node selector for Envoy pods
envoy.tolerations	list	[]	Tolerations for Envoy pods
envoy.service.type	string	"ClusterIP"	This is the client-facing endpoint. In order to be able to connect to it, either enable ingress, or use type: LoadBalancer.
envoy.service.ports	list	[{"name":"grpc","port":8001,"targetPort":8001},{"name":"admin","port":9901,"targetPort":9901}]	Envoy Service ports
envoy.ingress	object	{"annotations":{},"enabled":false,"hostName":"","ingressClassName":""}	Ingress configuration for Envoy
envoy.grpc_route_timeout	string	"0s"	Timeout for gRPC route in Envoy; disabled by default (0s), preventing Envoy from closing connections too early.
envoy.rate_limiter.listener_level	object	{"enabled":false,"fill_interval":"12s","max_tokens":5,"tokens_per_fill":1}	This rate limiter explicitly controls the number of client connections to the Envoy Proxy.
envoy.rate_limiter.listener_level.enabled	bool	false	Enable rate limiter
envoy.rate_limiter.listener_level.max_tokens	int	5	Maximum number of simultaneous connections to the Envoy Proxy. Each new connection takes a “token” from the “bucket” which initially contains max_tokens tokens.
envoy.rate_limiter.listener_level.tokens_per_fill	int	1	tokens_per_fill tokens are added to the “bucket” every fill_interval, allowing new connections to be established.
envoy.rate_limiter.listener_level.fill_interval	string	"12s"	For example, adding a new token every 12 seconds allows 5 new connections every minute.
envoy.rate_limiter.prometheus_based	object	{"enabled":false,"luaConfig":"cfg/envoy-filter.lua"}	This rate limiter rejects new RepositoryIndex requests when the scaling metric per healthy Triton replica exceeds serverAdmissionThreshold.
envoy.rate_limiter.prometheus_based.enabled	bool	false	Enable rate limiter
envoy.rate_limiter.prometheus_based.luaConfig	string	"cfg/envoy-filter.lua"	Envoy Lua filter (prometheus rate limiter and scaleFromZero)
envoy.loadBalancerPolicy	string	"LEAST_REQUEST"	Envoy load balancer policy. Options: ROUND_ROBIN, LEAST_REQUEST, RING_HASH, RANDOM, MAGLEV
envoy.auth.enabled	bool	false	Enable authentication in Envoy proxy
envoy.auth.jwt_issuer	string	""
envoy.auth.jwt_remote_jwks_uri	string	""
envoy.auth.audiences	list	[]
envoy.auth.url	string	""
envoy.auth.port	int	443
keda.enabled	bool	false	Enable autoscaling (requires Prometheus). KEDA scales Triton on serverLoadMetric against serverLoadThreshold.
keda.metricType	string	""	How the HPA interprets the metric. Empty selects AverageValue (fleet-wide) for the default metric and Value (per replica) for a custom serverLoadMetric.
keda.minReplicaCount	int	1	Minimum and maximum number of Triton servers. Set minReplicaCount to 0 to release all resources when idle (requires scaleFromZero.enabled). With scaleFromZero, a RepositoryIndex request scales Triton to max(1, minReplicaCount), and upgrades keep the live ScaledObject minReplicaCount.
keda.maxReplicaCount	int	2
keda.pollingInterval	int	10	Seconds between KEDA activity checks, which govern scaling to and from zero (scaling between 1 and maxReplicaCount follows the HPA sync period). 0 falls back to 10.
keda.cooldownPeriod	int	300	Seconds the metric must stay at or below activationThreshold before KEDA scales to zero. An empty query result counts as no load.
keda.activationThreshold	float	0	Metric value at or below which the trigger is inactive for scaling to zero; about 0.1 ignores probe traffic through Envoy.
keda.zeroIdleReplicas	bool	false	If set to true, KEDA sets idleReplicaCount to 0. This cannot scale from 0 back to 1 when the load metric comes from Triton. Use scaleFromZero.enabled with keda.minReplicaCount: 0 instead.
keda.scaleUp.stabilizationWindowSeconds	int	0	HPA scale-up stabilization window in seconds (0 reacts to the latest evaluation)
keda.scaleUp.periodSeconds	int	15	Scale-up rate-limit period in seconds
keda.scaleUp.stepsize	int	2	Replicas added per period
keda.scaleDown.stabilizationWindowSeconds	int	90	HPA scale-down stabilization window in seconds
keda.scaleDown.periodSeconds	int	15	Scale-down rate-limit period in seconds
keda.scaleDown.stepsize	int	2	Replicas removed per period
prometheus.external.enabled	bool	false	Enable external Prometheus instance. If true, Prometheus parameters outside of prometheus.external will be ignored.
prometheus.external.url	string	""	External Prometheus server url
prometheus.external.port	int	443	External Prometheus server port number
prometheus.external.scheme	string	"https"	Specify whether external Prometheus endpoint is exposed as http or https
prometheus.enabled	bool	false	Enable or disable custom Prometheus deployment
prometheus.server.extraFlags[0]	string	"web.enable-remote-write-receiver"
prometheus.server.useExistingClusterRoleName	string	"supersonic-prometheus-role"
prometheus.server.releaseNamespace	bool	true
prometheus.server.persistentVolume.enabled	bool	false
prometheus.server.resources.requests.cpu	string	"500m"
prometheus.server.resources.requests.memory	string	"512Mi"
prometheus.server.resources.limits.cpu	int	1
prometheus.server.resources.limits.memory	string	"1Gi"
prometheus.server.retention	string	"15d"
prometheus.server.global.scrape_interval	string	"5s"
prometheus.server.global.evaluation_interval	string	"5s"
prometheus.server.service.enabled	bool	true
prometheus.server.service.servicePort	int	9090
prometheus.server.configMapOverrideName	string	"prometheus-config"
prometheus.server.ingress	object	{"annotations":{},"enabled":false,"hostName":"","hosts":[],"ingressClassName":"","tls":[{"hosts":[]}]}	Ingress configuration for Prometheus
prometheus.server.ingress.hostName	string	""	If this parameter is set in values.yaml, the hosts and tls parameters can be omitted, as they will be set by the installer plugin
prometheus.serviceAccounts.server.create	bool	false
prometheus.serviceAccounts.server.name	string	"supersonic-prometheus-sa"
prometheus.rbac.create	bool	false
prometheus.alertmanager.enabled	bool	false
prometheus.pushgateway.enabled	bool	false
prometheus.kube-state-metrics.enabled	bool	false
prometheus.prometheus-node-exporter.enabled	bool	false
prometheus.prometheus-pushgateway.enabled	bool	false
prometheus.configmapReload.prometheus.enabled	bool	false
grafana.enabled	bool	false	Enable Grafana
grafana.adminUser	string	"admin"
grafana.adminPassword	string	"admin"
grafana.persistence.enabled	bool	false
grafana.rbac.create	bool	false
grafana.serviceAccount.create	bool	false
grafana.datasources	object	{"datasources.yaml":{"apiVersion":1,"datasources":[{"access":"proxy","isDefault":true,"jsonData":{"timeInterval":"5s","tlsSkipVerify":true},"name":"prometheus","type":"prometheus","url":"http://supersonic-prometheus-server:9090"}]}}	Grafana datasources configuration
grafana.dashboardProviders.”dashboardproviders.yaml”.apiVersion	int	1
grafana.dashboardProviders.”dashboardproviders.yaml”.providers[0].name	string	"default"
grafana.dashboardProviders.”dashboardproviders.yaml”.providers[0].orgId	int	1
grafana.dashboardProviders.”dashboardproviders.yaml”.providers[0].folder	string	""
grafana.dashboardProviders.”dashboardproviders.yaml”.providers[0].type	string	"file"
grafana.dashboardProviders.”dashboardproviders.yaml”.providers[0].disableDeletion	bool	false
grafana.dashboardProviders.”dashboardproviders.yaml”.providers[0].editable	bool	true
grafana.dashboardProviders.”dashboardproviders.yaml”.providers[0].options.path	string	"/var/lib/grafana/dashboards/default"
grafana.dashboardsConfigMaps.default	string	"supersonic-grafana-default-dashboard"
grafana.”grafana.ini”.auth.disable_login_form	bool	true
grafana.”grafana.ini”.”auth.anonymous”.enabled	bool	true
grafana.”grafana.ini”.”auth.anonymous”.org_role	string	"Admin"
grafana.”grafana.ini”.dashboards.default_home_dashboard_path	string	"/var/lib/grafana/dashboards/default/default.json"
grafana.”grafana.ini”.server.root_url	string	""
grafana.resources.limits.cpu	int	1
grafana.resources.limits.memory	string	"1Gi"
grafana.resources.requests.cpu	string	"100m"
grafana.resources.requests.memory	string	"128Mi"
grafana.ingress	object	{"annotations":{},"enabled":false,"hostName":"","hosts":[],"ingressClassName":"","path":"/","pathType":"ImplementationSpecific","tls":[{"hosts":[]}]}	Ingress configuration for Grafana
grafana.ingress.hostName	string	""	If this parameter is set in values.yaml, the hosts and tls parameters can be omitted, as they will be set by the installer plugin
metricsCollector.enabled	bool	false	Enable metrics collector
metricsCollector.image	string	"python:3.14-slim"	Image for the metrics collector
metricsCollector.host	string	"0.0.0.0"	Host and port for metrics collector
metricsCollector.port	int	8003
metricsCollector.resources	object	{"limits":{"cpu":1,"memory":"1Gi"},"requests":{"cpu":"500m","memory":"512Mi"}}	Resource limits and requests for metrics collector
metricsCollector.service	object	{"port":8003,"type":"ClusterIP"}	Add service
metricsCollector.ingress	object	{"annotations":{},"enabled":false,"hostName":"","ingressClassName":""}	Ingress configuration for metrics collector