Temporal Cloud mTLS Private Key Errors

I'm currently trying to setup temporal cloud with Retool which requires mTLS using the Helm Chart based on this Deploy Retool on Kubernetes with Helm | Retool Docs.

On the temporal cloud docs Manage certificates | Temporal Platform Documentation there is no specific requirement on how the certificate is generated (as you can bring in your own Cert Management Infra - which we're doing).

To facilitate the provisioning and lifecycle of the certificate we are using https://cert-manager.io/ to generate the end entity certificate and transforming the resulting values to what Retool expects (base64 encoded) so it can be used in the env var WORKFLOW_TEMPORAL_TLS_KEY

When I plug in the secret I get the following error (which crashes the workflow-worker pod)

TransportError: tonic::transport::Error(Transport, General("failed to parse private key as RSA, ECDSA, or EdDSA"))
    at NativeConnection.connect (/retool_backend/node_modules/.pnpm/@temporalio+worker@1.11.6_@swc+helpers@0.5.3_metro@0.80.9_encoding@0.1.13_metro-minify-terser@0.80.9_/node_modules/@temporalio/worker/lib/connection.js:58:23)
    at async gF (/retool_backend/bundle/main.js:11499:21822)
    at async INo (/retool_backend/bundle/main.js:11567:11716)
    at async Object.oIb (/retool_backend/bundle/main.js:11567:15085)

Other things I've tried:

  • Changing the private key encoding (PKCS1/PKCS8)
  • Switching private key algorithms (RSA, ECDSA)
  • Plugging the base64 value directly in the env var in the deployment
  • Using the secret directly (get a ParseKeyError)

Unfortunately outside the environment variables, there isnt much documentation on how this is supposed to work.

Helm Values for reference:

# nameOverride:
# fullnameOverride:

config:
  # licenseKey: "EXPIRED-LICENSE-KEY-TRIAL"
  # licenseKeySecretName is the name of the secret where the Retool license key is stored (can be used instead of licenseKey)
  licenseKeySecretName: retool
  # licenseKeySecretKey is the key in the k8s secret, default: license-key
  licenseKeySecretKey: licenseKey
  useInsecureCookies: false
  # Timeout for queries, in ms.
  dbConnectorTimeout: 600000
  auth:
    google:
      clientId: clientId
      # clientSecret:
      # clientSecretSecretName is the name of the secret where the google client secret is stored (can be used instead of clientSecret)
      clientSecretSecretName: retool
      # clientSecretSecretKey is the key in the k8s secret, default: google-client-secret
      clientSecretSecretKey: clientSecret
      domain:
  encryptionKey:
  # encryptionKeySecretName is the name of the secret where the encryption key is stored (can be used instead of encryptionKey)
  encryptionKeySecretName: retool
  # encryptionKeySecretKey is the key in the k8s secret, default: encryption-key
  encryptionKeySecretKey: encryptionKey
  jwtSecret:
  # jwtSecretSecretName is the name of the secret where the jwt secret is stored (can be used instead of jwtSecret)
  jwtSecretSecretName: retool
  # jwtSecretSecretKey is the key in the k8s secret, default: jwt-secret
  jwtSecretSecretKey: jwtSecret

  # IMPORTANT: Incompatible with postgresql subchart
  # Please disable the subchart in order to use a managed or external postgres instance.
  postgresql:
    host: xxxxxx
    port: "5432"
    db: "retool"
    user: retool
    ssl_enabled: 'true'
    # password:
    # ssl_enabled: whether to enable SSL for connecting to the Retool Postgres database, equivalent to the POSTGRES_SSL_ENABLED environment variable. Place here instead of in the env block.
    # passwordSecretName is the name of the secret where the pg password is stored (can be used instead of password)
    passwordSecretName: retool
    # passwordSecretKey is the key in the k8s secret, default: postgresql-password
    passwordSecretKey: dbPassword

image:
  repository: "tryretool/backend"
  # You need to pick a specific tag here, this chart will not make a decision for you
  tag: "3.253.6-stable"
  pullPolicy: "IfNotPresent"

commandline:
  args: []

env: {}

# Optionally specify additional environment variables to be populated from Kubernetes secrets.
# Useful for passing in SCIM_AUTH_TOKEN or other secret environment variables from Kubernetes secrets.
environmentSecrets:
  - name: WORKFLOW_TEMPORAL_TLS_CRT
    secretKeyRef:
      name: temporal-certs
      key: temporal-tls-cert
  - name: WORKFLOW_TEMPORAL_SERVER_ROOT_CA_CRT
    secretKeyRef:
      name: temporal-certs
      key: temporal-root-ca
  # - name: SCIM_AUTH_TOKEN
  #   secretKeyRef:
  #     name: retool-scim-auth-token
  #     key: auth-token
  # - name: GITHUB_APP_PRIVATE_KEY
  #   secretKeyRef:
  #     name: retool-github-app-private-key
  #     key: private-key

# Optionally specify environmental variables. Useful for variables that are not key-value, as env: {} above requires.
# Can also include environment secrets here instead of in environmentSecrets
environmentVariables:
  - name: BASE_DOMAIN
    value: xxxxxx

  #   - name: SCIM_AUTH_TOKEN
  #     valueFrom:
  #       secretKeyRef:
  #         name: retool-scim-auth-token
  #         key: auth-token
  #   - name: GITHUB_APP_PRIVATE_KEY
  #     valueFrom:
  #       secretKeyRef:
  #         name: retool-github-app-private-key
  #         key: private-key
  #   - name: POD_HOST_IP
  #     valueFrom:
  #       fieldRef:
  #         fieldPath: status.hostIP

# Enables support for the legacy external secrets (enabled) and the modern External Secrets Operator (externalSecretsOperator.enabled).
# These are mutually exclusive as both enable reading in environments variables via External Secrets.
externalSecrets:
  # Support for legacy external secrets, note this is deprecated in favour of External Secrets Operator: https://github.com/godaddy/kubernetes-external-secrets
  # This mode only allows a single secret name to be provided.
  enabled: false
  name: retool-config
  # Array of secrets to be use as env variables. (Optional)
  secrets:
    []
    # - name: retool-config
    # - name: retool-db
  # Support for External Secrets Operator: https://github.com/external-secrets/external-secrets
  externalSecretsOperator:
    enabled: false
    # External Secrets Operator Backend Types: https://github.com/external-secrets/external-secrets#supported-backends
    # Default set to AWS Secrets Manager.
    backendType: secretsManager
    # Array of name/path key/value pairs to use for the External Secrets Objects.
    secretRef:
      []
      # - name: retool-config
      #   path: global-retool-config
      # - name: retool-db
      #   path: global-retool-db-config
    # When true, uses kubernetes-client CRDs and not external-secrets CRDs
    # Defaults to true
    useLegacyCR: true

files: {}

deployment:
  labels: {}
  annotations: {}

service:
  type: ClusterIP
  externalPort: 3000
  internalPort: 3000
  # externalIPs:
  # - 192.168.0.1
  #
  ## LoadBalancer IP if service.type is LoadBalancer
  # loadBalancerIP: 10.2.2.2
  annotations: {}
  labels:
    app: "retool"
  ## Limit load balancer source ips to list of CIDRs (where available)
  # loadBalancerSourceRanges: []
  selector: {}
  # portName: service-port

ingress:
  enabled: false
  # For k8s 1.18+
  # ingressClassName:
  labels: {}
  annotations: {}
  # kubernetes.io/ingress.class: nginx
  # kubernetes.io/tls-acme: "true"
  hosts:
  # - host: retool.example.com
  #   paths:
  #     - path: /
  tls:
  # - secretName: retool.example.com
  #   hosts:
  #     - retool.example.com
  # servicePort: service-port
  pathType: ImplementationSpecific
  # For supporting other ingress controllers that require customizing the .backend.service.name and .backend.service.port.name,
  # like AWS ALB extraPaths allows that customization, and it takes precedence in the list of paths of for the host,
  # this is in order to allow a rule like ssl-redirect from port 80-->443 to be first ( otherwise there wouldn't be a redirect )
  #   extraPaths:
  #     - path: /*
  #       backend:
  #         service:
  #           name: ssl-redirect
  #           port:
  #             name: use-annotation
  #       pathType: ImplementationSpecific

postgresql:
  # We highly recommend you do NOT use this subchart as is to run Postgres in a container
  # for your production instance of Retool; it is a default. Please use a managed Postgres,
  # or self-host more permanently. Use enabled: false and set in config above to do so.
  enabled: false
  ssl_enabled: false
  auth:
    database: hammerhead_production
    username: postgres
    password: retool
    # IMPORTANT: When setting username to `postgres` a random password will be generated unless the postgresPassword field is uncommented
    # postgresPassword: retool
  service:
    port: 5432
  # Use the offical docker image rather than bitnami/docker
  # since Retool depends on the uuid-ossp extension
  image:
    repository: "postgres"
    tag: "13"
  postgresqlDataDir: "/data/pgdata"
  primary:
    persistence:
      enabled: true
      mountPath: "/data/"

serviceAccount:
  # Specifies whether a service account should be created
  create: true
  # The name of the service account to use.
  # If not set and create is true, a name is generated using the fullname template
  # If set and create is false, the service account must be existing
  name: retool
  annotations: {}

livenessProbe:
  enabled: true
  path: /api/checkHealth
  initialDelaySeconds: 30
  timeoutSeconds: 10
  periodSeconds: 10
  failureThreshold: 3

readinessProbe:
  enabled: true
  path: /api/checkHealth
  initialDelaySeconds: 30
  timeoutSeconds: 10
  periodSeconds: 10
  successThreshold: 5

# To avoid increasing livenessProbe initialDelaySeconds the good practice is to use a startupProbe.
startupProbe:
  enabled: false
  path: /api/checkHealth
  initialDelaySeconds: 60
  timeoutSeconds: 10
  periodSeconds: 10
  successThreshold: 5

extraContainers: []

extraVolumeMounts: []

extraVolumes: []


# These resource specifications apply to the main backend and workflows backend pods.
resources:
  # If you have more than 1 replica, the minimum recommended resources configuration is as follows:
  # - cpu: 2048m
  # - memory: 4096Mi
  # If you only have 1 replica, please double the above numbers.
  limits:
    cpu: 4096m
    memory: 8192Mi
  requests:
    cpu: 2048m
    memory: 4096Mi

priorityClassName: ""

# Affinity for pod assignment
# Ref: https://kubernetes.io/docs/concepts/configuration/assign-pod-node/#affinity-and-anti-affinity
affinity:
  podAntiAffinity:
    preferredDuringSchedulingIgnoredDuringExecution:
      - weight: 100
        podAffinityTerm:
          labelSelector:
            matchExpressions:
              - key: "app.kubernetes.io/name"
                operator: In
                values:
                  - retool
          topologyKey: "kubernetes.io/hostname"

# Tolerations for pod assignment
# Ref: https://kubernetes.io/docs/concepts/configuration/taint-and-toleration/
tolerations:
  - key: amd64
    operator: "Equal"
    value: 'true'
    effect: NoSchedule

# Node labels for pod assignment
# Ref: https://kubernetes.io/docs/user-guide/node-selection/
nodeSelector:
  kubernetes.io/arch: amd64

# Common annotations for all pods.
podAnnotations: {}

revisionHistoryLimit: 3

# Optional pod disruption budget, for ensuring higher availability of the
# Retool application.  Specify either minAvailable or maxUnavailable, as
# either an integer pod count (1) or a string percentage ("50%").
# Ref: https://kubernetes.io/docs/tasks/run-application/configure-pdb/
#
# Example:
# podDisruptionBudget:
#   maxUnavailable: 1

# Common labels for all pods (backend and job runner) for pod assignment
podLabels: {}

# Increasing replica count will deploy a separate pod for backend and jobs
# Example 1: with 1 replicas, you will end up with 1 combined backend and jobs pod (unless jobRunner.enabled is true, see below)
# Example 2: with 2 replicas, you will end up with 2 backends + 1 jobs pod
# Example 3: with 3 replicas, you will end up with 3 backends + 1 jobs pod
replicaCount: 2

jobRunner:
  # explicitly enable this pod if exactly 1 api backend container and
  # 1 jobs runner container is desired. otherwise a replicaCount of 2
  # will already launch a job runner pod
  # enabled: true

  # Annotations for job runner pods
  annotations: {}

  # Labels for job runner pods
  labels: {}

backend:
  # Annotations for backend pods
  annotations: {}

  # Labels for backend pods
  labels: {}

ui:
  # Annotations for ui pods
  annotations: {}

  # Labels for ui pods
  labels: {}

workflows:
  # enabled by default from Chart version 6.0.2 and Retool image 3.6.11 onwards
  # explicitly set other fields as needed

  # enabled: true

  # ADVANCED: The temporal cluster can be scaled separately in the subchart (charts/retool-temporal-services-helm/values.yaml)
  # If your needs require scaling temporal, reach out to us for guidance -- it is likely the bottleneck is DB or worker replicaCount
  worker:
    # A replicaCount of 1 will launch 6 pods -- 1 workflow backend, 1 workflow worker, and 4 pods that make up the executor temporal cluster
    # Scaling this number will increase the number of workflow workers, e.g. a replicaCount of 4
    # will launch 9 pods -- 1 workflow backend, 4 workflow workers, and 4 for temporal cluster
    replicaCount: 1

  backend:
    # A replicaCount of 1 will launch 6 pods -- 1 workflow backend, 1 workflow worker, and 4 pods that make up the executor temporal cluster
    # Scaling this number will increase the number of workflow backends, e.g. a replicaCount of 4
    # will launch 9 pods -- 4 workflow backend, 1 workflow workers, and 4 for temporal cluster
    replicaCount: 1

  # Timeout for queries, in ms. This will set the timeout for workflows-related pods only
  # If this value is not set but config.dbConnectorTimeout is, we will set workflows pod timeouts
  # to .Values.config.dbConnectorTimeout
  # dbConnectorTimeout: 120000

  # Annotations for workflows worker and workflow backend pods
  annotations: {}

  # Labels for workflows worker and workflow backend pods
  labels: {}

  # IMPORTANT: Incompatible with retool-temporal-services-helm subchart
  # This allows configuring a Retool Workflows deployment that uses your own Temporal cluster
  # instead of deploying a new one. Set enabled to true and add the config variables.
  temporal:
    # set enabled to true if using a pre-existing temporal cluster
    enabled: true
    # discoverable service name for the temporal frontend service
    host: xxxx.yyyy.tmprl.cloud
    # discoverable service port for the temporal frontend service
    port: 7233
    # temporal namespace to use for Temporal Workflows related to Retool
    namespace: namespace
    # whether to use TLS/SSL when connecting to Temporal
    sslEnabled: true
    # base64 encoded string of TLS/SSL client certificate to use for mTLS
    # sslCert:
    # base64 encoded string of TLS/SSL client certificate secret key to use for mTLS
    # sslKey:
    # recommended alternative to sslKey. Name and key of k8s secret containing base64 encoded sslKey
    sslKeySecretName: temporal-certs
    # sslKeySecretKey

  # Config for workflows worker pods. Node heap size limits can be overridden here
  # otelCollector can be set to an OpenTelemetry Collector in your k8s cluster. This will configure Temporal metrics collection which
  # provides observability into Workflows worker performance, particularly useful in high QPS use-cases
  # environmentVariables will only be set on the workflows worker
  # only change the CONCURRENT_*_LIMIT values if you have higher load usecases and deploy
  # code_executor. Otherwise, the worker may OOM if the Workflows blocks use too much memory.
  config: {}
  # config: {
  #   nodeOptions: --max_old_space_size=1024
  #   otelCollector: {
  #     enabled: true
  #     endpoint: http://$(HOST_IP):4317
  #   }
  #   environmentVariables: []
  #   - name: WORKFLOW_TEMPORAL_CONCURRENT_TASKS_LIMIT
  #     value: "100"
  #   - name: WORKFLOW_TEMPORAL_CONCURRENT_ACTIVITIES_LIMIT
  #     value: "100"
  # }

  # Resources for the workflow worker only - these are sane inputs that bias towards stability
  # Can adjust but may see OOM errors if memory too low for heavy workflow load
  # To make adjustments to workflows backend, use top level resources key.
  resources:
    limits:
      cpu: 2000m
      memory: 8192Mi
    requests:
      cpu: 1000m
      memory: 2048Mi

dbconnector:
  java:
    # Disable this to disable Retool's Java dbconnector.
    enabled: true

multiplayer:
  # Enable this to use Retool's experimental multiplayer editing feature.
  # This feature is not ready for production use; please check with the Retool team before enablement.
  enabled: false

  replicaCount: 1

  # Resources for multiplayer pods
  resources:
    requests:
      cpu: "200m"
      memory: "256Mi"
    limits:
      memory: "2048Mi"

  # Set environment variables for multiplayer pods, e.g. defining which origin to use
  # environmentVariables:
  #   - name: MAIN_DOMAIN
  #     value: retool.foo.com

  # Annotations for multiplayer pods
  annotations: {}

  # Labels for multiplayer pods
  labels: {}

codeExecutor:
  # enabled by default from Chart version 6.0.15 and Retool image 3.20.15 onwards
  # explicitly set other fields as needed

  # enabled: true

  replicaCount: 1

  # Annotations for code executor pods
  annotations: {}

  # Labels for code executor pods
  labels: {}

  # Config for code executor. Node heap size limits can be overridden here
  config: {}
  # config: {
  #   nodeOptions: --max_old_space_size=1024
  # }
  volumes: {}
  volumeMounts: {}

  # Config affinity and anti-affinity rules for the code executor pods
  affinity: {}

  # Resources for the code executor. Most common issues will be seen with CPU usage as this will
  # most likely be CPU bound. Adjust the CPU if latency increases under load.
  resources:
    limits:
      cpu: 2000m
      memory: 2048Mi
    requests:
      cpu: 1000m
      memory: 1024Mi

  # code executor uses nsjail to sandbox code execution. nsjail requires privileged container access.
  # If your deployment does not support privileged access, you can set `privileged` to false to not
  # use nsjail. Without nsjail, all code is run without sandboxing within your deployment.
  securityContext:
    privileged: true

  image:
    repository: tryretool/code-executor-service
    # defaults to image.tag if >= 3.20.15, otherwise defaults to 1.1.0; explicitly set to override.
    # tag:
    pullPolicy: IfNotPresent

retool-temporal-services-helm:
  # Enable to spin up a new Temporal Cluster alongside Retool
  enabled: false
  server:
    # Defines image to be used for temporal server
    image:
      repository: tryretool/one-offs
      tag: retool-temporal-1.1.5
      pullPolicy: IfNotPresent
    # this configures grpc_health_probe (https://github.com/grpc-ecosystem/grpc-health-probe)
    # for healthchecks instead of native k8s.
    # Set this to true if deploying in a k8s cluster on version <1.24
    useLegacyHealthProbe: false
    config:
      # the below values specify the database for temporal internals and workflow state
      # both can point to the same db, and even the same as retool main above, although throughput
      # will be limited. We strongly suggest using two total DBs: one for retool-main and one
      # for default and visibility below
      persistence:
        default:
          sql:
            # host:
            # port:
            # the dbname used for temporal
            # database: temporal
            # user:
            # password:
            # existingSecret is the name of the secret where password is stored
            # existingSecret:
            # secretKey is the key in the k8s secret
            # secretKey:
            # options for SSL connections to database
            # tls:
            #   enabled: true
            #   ัaFile:
            #   certFile:
            #   keyFile:
            #   enableHostVerification: false
            #   serverName:
        visibility:
          sql:
            # host:
            # port:
            # the dbname used for temporal visibility
            # database: temporal_visibility
            # user:
            # password:
            # existingSecret is the name of the secret where password is stored
            # existingSecret:
            # secretKey is the key in the k8s secret
            # secretKey:
            # options for SSL connections to database
            # tls:
            #   enabled: true
            #   ัaFile:
            #   certFile:
            #   keyFile:
            #   enableHostVerification: false
            #   serverName:

      # use-cases with very high throughput demands (>10k workflow blocks/sec) can modify
      # below value to be higher, such as 512 or 1024
      numHistoryShards: 128

    # define resources for each temporal service -- these are sane starting points that allow
    # for scaling to ~3 workflow workers without hitting bottlenecks
    resources:
      limits:
        cpu: 500m
        memory: 1024Mi
      requests:
        cpu: 100m
        memory: 128Mi
    # example of setting service-specific resources, here we increase memory limit for history server
    history:
      resources:
        limits:
          cpu: 500m
          memory: 2Gi
        requests:
          cpu: 100m
          memory: 128Mi

  # Set environment variables for temporal pods, e.g. skipping DB creation step at startup
  # environmentVariables:
  #   - name: SKIP_DB_CREATE
  #     value: "true"

  # Setting this value to true will spin up the temporal web UI, admintools,
  # along with prometheus and grafana clusters for debugging performance
  # TODO: define this in _helpers.tpl
  # visibilityDebugMode: false
  # DEBUGGING: below pods can be used for debugging and watching metrics
  # launches prometheus pods and listeners on temporal services and worker
  prometheus:
    enabled: false
  # launches a grafana pod with helpful workflows executor metrics and graphs
  grafana:
    enabled: false
  # launches the temporal web UI, which allows you to see which workflows are currently running
  web:
    enabled: false
  # launches the temporal admintools pod, which allows you to manage your cluster, e.g. terminate workflows
  admintools:
    enabled: false

persistentVolumeClaim:
  # set to true to use pvc
  enabled: false
  # set to the name of the existing pvc
  existingClaim: ""
  annotations: {}
  accessModes:
    - ReadWriteOnce
  size: "15Gi"
  ## If defined, storageClassName: <storageClass>
  ## If set to "-", storageClassName: "", which disables dynamic provisioning
  ## If undefined (the default) or set to null, no storageClassName spec is
  ##   set, choosing the default provisioner.  (gp2 on AWS, standard on
  ##   GKE, AWS & OpenStack)
  ##
  storageClass: ""
  mountPath: /retool_backend/pv-data

# default security context
securityContext:
  enabled: false
  allowPrivilegeEscalation: false
  runAsUser: 1000
  fsGroup: 2000

extraConfigMapMounts: []

initContainers: {}

extraManifests: []
# extraManifests:
#  - apiVersion: cloud.google.com/v1beta1
#    kind: BackendConfig
#    metadata:
#      name: "{{ .Release.Name }}-testing"
#    spec:
#      securityPolicy:
#        name: "my-gcp-cloud-armor-policy"

# Support for AWS Security groups for pods
# Ref: https://docs.aws.amazon.com/eks/latest/userguide/security-groups-for-pods.html
securityGroupPolicy:
  enabled: false
  groupIds: []

hostAliases: {}
# Support hostAliases on the backend deployment
# hostAliases:
#   - ip: 1.1.1.1
#     hostnames:
#     - test.com
#     - anothertest.com

telemetry:
  # When enabled, will collect metrics and logs from the other pods in the
  # chart. These will be forwarded to Retool for proactive monitoring and bug
  # identification when `telemetry.sendToRetool.enabled = true`, and can also
  # optionally be sent to a user-managed destination via
  # `telemetry.customVectorConfig`.
  enabled: false

  sendToRetool:
    # Only relevant when `telemetry.enabled = true`. When enabled, telemetry
    # from pods in this chart will be forwarded to Retool for proactive
    # monitoring and bug identification.
    enabled: true

    # Should not be changed except for chart testing.
    address: "https://telemetry.retool.com:443"

  image:
    repository: "tryretool/telemetry"

    # Default to same as top level `image.tag`.
    tag: null

    pullPolicy: "IfNotPresent"

  resources:
    requests:
      cpu: 128m
      memory: 128Mi
    limits:
      cpu: 256m
      memory: 256Mi

  # When present, any vector config here gets added to the telemetry pod (via a
  # created configmap) and added to the internal [vector](https://vector.dev/)
  # instance which runs inside. This can enable adding additional user-specified
  # sinks for collected telemetry data.
  #
  # See [vector sinks
  # documentation](https://vector.dev/docs/reference/configuration/sinks/) for
  # more details.
  customVectorConfig: {}

  # When present, any grafana-agent config here gets added to the telemetry pod
  # (via a created configmap) and added to the internal
  # [grafana-agent](https://grafana.com/docs/agent/latest/) instance which runs
  # inside. This can enable adding additional user-specified sources of extra
  # telemetry data.
  #
  # The internal grafana-agent runs in Flow mode, so the config here must use
  # river syntax. See [grafana-agent Flow mode
  # documentation](https://grafana.com/docs/agent/latest/flow/) for more
  # details.
  customGrafanaAgentConfig:

  serviceAccount:
    create: true
    name:
    automountToken: true

  rbac:
    create: true

  extraEnv: []
  extraAnnotations: {}
  extraPodSpec: {}
  extraVolumes: []
  extraVolumeMounts: []
  extraPorts: []
  extraContainerSpec: {}

I managed to solve this on my own though through a combination of things. So posting this here in case someone runs into the same problem.

The most likely culprit is the key size. I was using a large key size (RSA 8192/ECDSA 521). But dropping it down the RSA 2048 ended up working.

I also changed the signature algorithm from SHA512 to SHA256 but I suspect the key size being the greatest contributing factor.

There is no documentation mentioned on the specification of the private key that I could find. And temporal cloud has the following requirements which should've already been met:

  • The signing algorithm must be either RSA or ECDSA and must include SHA-256 or stronger message authentication. SHA-1 and MD5 cannot be used.

Thanks for providing an update, @achan-alaffia!