Skip to content

spotfire-terrservice

Version: 1.0.0 Type: application AppVersion: 1.22.0

A Helm chart for Spotfire® Enterprise Runtime for R - Server Edition

Homepage: https://github.com/spotfiresoftware/spotfire-cloud-deployment-kit ⧉

Source Code

Requirements

Kubernetes: >=1.24.0-0

Repository Name Version
file://../spotfire-common spotfire-common 1.0.0

Overview

This chart deploys the Spotfire® Enterprise Runtime for R - Server Edition (a/k/a TERR™ service) (Formerly TIBCO Enterprise Runtime for R - Server Edition) service on a Kubernetes ⧉ cluster using the Helm ⧉ package manager.

The TERR service pod includes:

  • A Fluent Bit ⧉ sidecar container for log forwarding.
  • Service annotations for Prometheus ⧉ scrapers. The Prometheus server discovers the service endpoint using these specifications and scrapes metrics from the exporter.
  • Predefined configuration for horizontal pod autoscaling with KEDA ⧉ and Prometheus.

This chart is tested to work with Elasticsearch ⧉, Prometheus ⧉, and KEDA ⧉.

Prerequisites

Usage

Installing

  1. Export the SPOTFIRE_SERVER value to connect to the spotfire-server service: 
    export SPOTFIRE_SERVER=$(kubectl get services --selector=app.kubernetes.io/part-of=spotfire,app.kubernetes.io/name=spotfire-server --output=jsonpath={.items..metadata.name})
  2. Forward the logs to the log-forwarder service: 
    export LOG_FORWARDER=$(kubectl get services --selector=app.kubernetes.io/part-of=spotfire,app.kubernetes.io/name=log-forwarder --output=jsonpath={.items..metadata.name})
  3. Install this chart with the release name my-release and custom values from my-values.yaml: 
    helm install my-release . \
    --set acceptEUA=true \
    --set global.spotfire.image.registry="127.0.0.1:32000" \
    --set global.spotfire.image.pullPolicy="Always" \
    --set nodemanagerConfig.serverBackendAddress="$SPOTFIRE_SERVER" \
    --set logging.logForwarderAddress="$LOG_FORWARDER" \
    -f my-values.yaml

Note: This Spotfire Helm chart requires setting the parameter acceptEUA or the parameter global.spotfire.acceptEUA to the value true. By doing so, you agree that your use of the Spotfire software running in the managed containers will be governed by the terms of the Cloud Software Group, Inc. End User Agreement ⧉.

Note: You must provide your private registry address where the Spotfire container images are stored.

See helm install ⧉ for command documentation.

Configuration

To set Custom configuration properties ⧉, add the name of the property as a key under the configuration section in your Helm values.

Example: 

# The maximum number of TERR engine sessions that are allowed to run concurrently in the TERR service.
engine.session.max: 5

# The number of TERR engines preallocated and available for new sessions in the TERR service queue.
engine.queue.size: 10

Uninstalling

To uninstall/delete the my-release deployment: 

helm uninstall my-release

See helm uninstall ⧉ for command documentation.

Scaling

For scaling the my-release deployment, do a helm upgrade, providing the target number of pod instances in the replicaCount variable. 

helm upgrade --install my-release . --reuse-values --set replicaCount=3

Autoscaling with KEDA

To use KEDA ⧉ for autoscaling, first install KEDA in the Kubernetes cluster. You must also install a Prometheus instance that scrapes metrics from the Spotfire pods.

Example: A values.yaml snippet configuration for enabling autoscaling with KEDA: 

resources:
  limits:
    cpu: 5
kedaAutoscaling:
  enabled: true
  spotfireConfig:
    prometheusServerAddress: http://prometheus-server.monitor.svc.cluster.local
  threshold: 3
  minReplicas: 1
  maxReplicas: 3

The spotfire-terrservice has the following autoscaling defaults:

  • metric: spotfire_service_queue_engines_inUse ( serviceQueueEnginesInUse TERR service counter).
  • query: the sum of spotfire_service_queue_engines_inUse of the TERR service instances for the release name.

The counter serviceQueueEnginesInUse provides the total number of engines currently executing. By default, the TERR service has number of cores - 1 available slots, which means that kedaAutoscaling.threshold should be synchronized with resources.limits.cpu. Typically, you want to scale out before all the available capacity is taken. Therefore, the kedaAutoscaling.threshold should be lower than resources.limits.cpu. Note that clients requesting a slot typically wait until a slot is available.

For more information, see Monitoring the Spotfire Service for TERR using JMX ⧉.

Note: You can tune nodemanagerConfig.preStopDrainingTimeoutSeconds and other timeouts (for example, engine.execution.timeout and engine.session.maxtime) so that long-running jobs are not aborted prematurely when an instance is stopped to scale in. See Engine Timeout ⧉ for more details.

For more advanced scenarios, see kedaAutoscaling.advanced ⧉ and kedaAutoscaling.fallback ⧉.

Additionally, you can define your own custom scaling triggers ⧉. Helm template functionality is available: 

kedaAutoscaling:
  triggers: {} # list of triggers to activate scaling of the target resource

Note: For more details on the autoscaling defaults, refer to the file templates/keda-autoscaling.yaml inside the chart.

Upgrading

When you upgrade to a newer Spotfire Server version and newer Spotfire services versions, upgrade the Spotfire Server first, and then upgrade the Spotfire services. See helm upgrade ⧉ for helm command documentation.

Upgrading helm chart version

Please review the release notes ⧉ for any changes, moved, or renamed parameters before upgrading the release.

Values

Key Type Default Description
global.spotfire.acceptEUA bool nil Accept the Cloud Software Group, Inc. End User Agreement ⧉ by setting the value to true.
global.spotfire.image.pullPolicy string "IfNotPresent" The global container image pull policy.
global.spotfire.image.pullSecrets list [] The global container image pull secrets.
global.spotfire.image.registry string nil The global container image registry. Used for spotfire/ container images, unless it is overridden.
acceptEUA bool nil Accept the Cloud Software Group, Inc. End User Agreement ⧉ by setting the value to true.
affinity object {}
configuration object {} Add Custom configuration properties ⧉. Keys under configuration should be named the same as the configuration property, for example 'engine.execution.timeout'.
extraContainers list [] Additional sidecar containers to add to the service pod.
extraEnvVars list [] Additional environment variables.
extraEnvVarsCM string "" The name of the ConfigMap containing additional environment variables.
extraEnvVarsSecret string "" The name of the Secret containing extra additional environment variables.
extraInitContainers list [] Additional init containers to add to the service pod.
extraVolumeMounts list [] Extra volumeMounts for the service container. More info: kubectl explain deployment.spec.template.spec.containers.volumeMounts.
extraVolumes list [] Extra volumes for the service container. More info: kubectl explain deployment.spec.template.spec.volumes.
fluentBitSidecar.image.pullPolicy string "IfNotPresent" The image pull policy for the fluent-bit logging sidecar image.
fluentBitSidecar.image.repository string "fluent/fluent-bit" The image repository for fluent-bit logging sidecar.
fluentBitSidecar.image.tag string "3.2.8" The image tag to use for fluent-bit logging sidecar.
fluentBitSidecar.securityContext object {} The securityContext setting for fluent-bit sidecar container. Overrides any securityContext setting on the Pod level.
fullnameOverride string ""
image.pullPolicy string nil The spotfire-server image pull policy. Overrides global.spotfire.image.pullPolicy.
image.pullSecrets list [] Image pull secrets.
image.registry string nil The image registry for spotfire-server. Overrides global.spotfire.image.registry value.
image.repository string "spotfire/spotfire-terrservice" The spotfire-server image repository.
image.tag string "1.22.0-v3.0.0" The container image tag to use.
kedaAutoscaling object {"advanced":{},"cooldownPeriod":300,"enabled":false,"fallback":{},"maxReplicas":4,"minReplicas":1,"pollingInterval":30,"spotfireConfig":{"prometheusServerAddress":"http://prometheus-server.monitor.svc.cluster.local"},"threshold":null,"triggers":[]} KEDA autoscaling configuration. See https://keda.sh/docs/latest/concepts/scaling-deployments for more details.
kedaAutoscaling.cooldownPeriod int 300 The period to wait after the last trigger reported active before scaling the resource back to 0.
kedaAutoscaling.maxReplicas int 4 This setting is passed to the HPA definition that KEDA creates for a given resource and holds the maximum number of replicas of the target resource.
kedaAutoscaling.minReplicas int 1 The minimum number of replicas KEDA scales the resource down to.
kedaAutoscaling.pollingInterval int 30 The interval to check each trigger on.
kedaAutoscaling.spotfireConfig object {"prometheusServerAddress":"http://prometheus-server.monitor.svc.cluster.local"} Spotfire specific settings.
kedaAutoscaling.spotfireConfig.prometheusServerAddress string "http://prometheus-server.monitor.svc.cluster.local" REQUIRED. The URL for the Prometheus server from where metrics are fetched.
livenessProbe.enabled bool true
livenessProbe.failureThreshold int 10
livenessProbe.httpGet.path string "/spotfire/liveness"
livenessProbe.httpGet.port string "registration"
livenessProbe.initialDelaySeconds int 60
livenessProbe.periodSeconds int 3
logging.logForwarderAddress string "" The spotfire-server log-forwarder name. Template.
logging.logLevel string "debug" Set to debug, trace, minimal, or leave empty for info. This applies for both node manager and the service.
nameOverride string ""
nodeSelector object {}
nodemanagerConfig.preStopDrainingTimeoutSeconds int 610 The draining timeout after which the service is forcefully shut down.
nodemanagerConfig.serverBackendAddress string "" The spotfire-server service name. This value is evaluated as a helm template.
podAnnotations."prometheus.io/path" string "/spotfire/metrics"
podAnnotations."prometheus.io/port" string "9080"
podAnnotations."prometheus.io/scrape" string "true"
podSecurityContext object {} The Pod securityContext setting applies to all of the containers inside the Pod.
readinessProbe.enabled bool false
readinessProbe.failureThreshold int 10
readinessProbe.httpGet.path string "/spotfire/readiness"
readinessProbe.httpGet.port string "registration"
readinessProbe.initialDelaySeconds int 60
readinessProbe.periodSeconds int 3
replicaCount int 1
resources object {}
securityContext object {} The securityContext setting for the service container. Overrides any securityContext setting on the Pod level.
service.port int 9501
service.type string "ClusterIP"
serviceAccount.annotations object {}
serviceAccount.create bool false
serviceAccount.name string ""
startupProbe.enabled bool true
startupProbe.failureThreshold int 20
startupProbe.httpGet.path string "/spotfire/started"
startupProbe.httpGet.port string "registration"
startupProbe.initialDelaySeconds int 60
startupProbe.periodSeconds int 3
tolerations list []
volumes.packages.existingClaim string "" When 'persistentVolumeClaim.create' is 'false', then use this value to define an already existing persistent volume claim.
volumes.packages.persistentVolumeClaim.create bool false If 'true', then a 'PersistentVolumeClaim' is created.
volumes.packages.persistentVolumeClaim.resources object {"requests":{"storage":"1Gi"}} Specifies the standard Kubernetes resource requests and/or limits for the customExt volume claims.
volumes.packages.persistentVolumeClaim.storageClassName string "" Specifies the name of the 'StorageClass' to use for the customExt volume-claim.
volumes.packages.persistentVolumeClaim.volumeName string nil Specifies the name of the persistent volume that should be used for the customExt volume-claim.
volumes.packages.subPath string "" The subPath of the volume to be used for the volume mount
volumes.troubleshooting.existingClaim string "" When 'persistentVolumeClaim.create' is 'false', then use this value to define an already existing persistent volume claim.
volumes.troubleshooting.persistentVolumeClaim.create bool false If 'true', then a 'PersistentVolumeClaim' will be created.
volumes.troubleshooting.persistentVolumeClaim.resources object {"requests":{"storage":"2Gi"}} Specifies the standard Kubernetes resource requests and/or limits for the volumes.troubleshooting claims.
volumes.troubleshooting.persistentVolumeClaim.storageClassName string "" Specifies the name of the 'StorageClass' to use for the volumes.troubleshooting-claim.
volumes.troubleshooting.persistentVolumeClaim.volumeName string nil Specifies the name of the persistent volume to use for the volumes.troubleshooting-claim.