- For the core and modules tiers, the deployer script reads a NexusOne YAML configuration file, and Terraform picks up every key in it automatically.
- For the tenant tier, you set its variables separately, through Tenant Manager, when you create a tenant.
Core and module variables
You set these variables in the deployer’s YAML configuration file for the core and modules tiers.Component toggles
Component toggles
false, other non-NexusOne pods in your cluster can connect.monitoring_type is embedded.Container registry
Container registry
Database
Database
Directory and Single Sign-On
Directory and Single Sign-On
true, but does
nothing useful until you also set:ad_bind_credentialad_bind_dnad_connection_urlad_groups_dnad_users_dn
false if you don’t use Active Directory or LDAP.OU=Groups,DC=example,DC=com. NexusOne only applies this on the first deployment. To change it
later, update it in Keycloak directly.OU=Users,DC=example,DC=com. NexusOne only applies this on the first deployment. To change it
later, update it in Keycloak directly.false to make them optional.Namespace, storage, and domain
Namespace, storage, and domain
example.com. Each component’s
hostname combines its own prefix, your shared_services_namespace, and this domain. For
example, if your namespace is nx1 and your domain is example.com, Grafana is at
grafana-nx1.example.com.Notification email
Notification email
email_enabled is true, since NexusOne has no working
default for it.Observability
Observability
obs_router_enabled is true.Authorization header, followed by the bigpanda_webhook_token. For example,
Authorization: Bearer <token>.Required when bigpanda_enabled is true.bigpanda_enabled is true and a different certificate authority signs your
BigPanda endpoint.- Required when
monitoring_typeisembeddedandgrafana_content_enabledistrue. - Fails if
bigpanda_webhook_urlorbigpanda_webhook_tokenis empty.
Authorization header to authenticate each alert.Required when bigpanda_enabled is true.central_grafana_enabled is true.nx1_grafana_platform_admin and nx1_grafana_platform_viewer roles
appear in NexusOne. Assign one to each person who needs access, as described in
Assign a role. Without one of these roles,
nobody can sign in.nx1-grafana-sync, a background pod that adds each tenant to the central
Grafana.Required when central_grafana_enabled, obs_router_enabled, and
obs_router_create_cluster_role are all true. Otherwise, the central Grafana shows no tenants.grafana_content_enabled
is false.false to manage that content yourself. Grafana, Mimir, Loki, and
Tempo keep running either way. Required when monitoring_type is embedded.embedded for NexusOne’s own Grafana, Mimir, Loki,
and Tempo, or leave as none to skip it.central_grafana_enabled and central_grafana_sync_enabled are both true.obs_router_enabled deploys.Required when all of the following are true:monitoring_typeisembeddedobs_router_enabledisfalsebecause another deployment on your cluster already runs the router
OpenSearch
OpenSearch
opensearch_security_enabled, since
otherwise this would expose an unauthenticated search API.-
accesskey -
bucket -
endpoint -
pathstyle -
secretkeyDefaults to the value of thestate_bucketvariable when not set. Write theopensearch_snapshot_s3values as a nested mapping in your configuration file, for example:
Secrets and credentials
Secrets and credentials
monitoring_type is embedded. NexusOne
generates a random value when you leave it unset.opensearch_security_enabled. NexusOne generates a
random value when you leave it unset, though a disaster recovery pair should pin the same
value on both sites.SSL/TLS and ingress
SSL/TLS and ingress
ssl_type is venafi.- Set it to
trueif your load balancer terminates TLS, so the ingress gateway adds theX-Forwarded-Proto: httpsheader to each request. - Set it to
falseif the ingress gateway terminates TLS.
ingress_type is envoy.- Set it to
httpif your load balancer terminates TLS, so traffic reaches the ingress gateway as plain HTTP. - Set it to
httpsif the ingress gateway terminates TLS.
ingress_type is envoy.State backend
State backend
Multi-site, sizing, and versions
Multi-site, sizing, and versions
nx1-grafana-sync image tag to deploy. Its version is independent of the NexusOne release.
Required when central_grafana_sync_enabled is true.cluster_name on each site, such as primary and standby.Tenant variables
You configure tenant modules tier variables separately from core and module variables. Instead of a YAML configuration file, you set them through Tenant Manager’s web app or API when you create a tenant.AI and LLM
AI and LLM
llm_anthropic_base_url when
llm_provider = "anthropic".llm_provider is ollama.llm_provider is openai, and must end in /v1, for example
https://inference.baseten.co/v1. Leave it empty for ollama and anthropic.llm_provider is
ollama or anthropic. Trino’s AI functions aren’t configured when llm_provider is openai.none, ollama, anthropic, or openai. Setting ollama also deploys
Ollama in a tenant. Setting openai requires llm_endpoint.Column encryption
Column encryption
vault_enabled is true.Component toggles
Component toggles
spark-connect-<client>.<domain>.This variable follows these rules:- Required when
spark_versionis4. - On OpenShift,
ocp_set_certmust betrueand your IngressController must allow HTTP/2.
true, the
tenant reuses that namespace’s Keycloak realm and observability stack instead of creating its
own.Container registry
Container registry
Decisions
Decisions
decisions, with one schema for each decision
workspace.Set to false to stop Trino and Spark from calling decisions. In the portal, you can still
create, simulate, approve, and publish decisions.decisions_enabled is true.Directory and Single Sign-On
Directory and Single Sign-On
true, but does
nothing useful until you also set:ad_bind_credentialad_bind_dnad_connection_urlad_groups_dnad_users_dn
false if you don’t use Active Directory or LDAP.OU=Groups,DC=example,DC=com. NexusOne only applies this on the first deployment. To change it
later, update it in Keycloak directly.OU=Users,DC=example,DC=com. NexusOne only applies this on the first deployment. To change it
later, update it in Keycloak directly.false if a tenant can’t reach the NexusOne API. Kyuubi then accepts only Keycloak tokens.false to make them optional.Filled in by Tenant Manager
Filled in by Tenant Manager
demo and your domain is example.com, that tenant’s Airflow is at airflow-demo.example.com.Karpenter
Karpenter
false
to keep them on your default worker nodes.{ "karpenter.sh/nodepool" = "workers" }.Required when karpenter_enabled is true.effect: One ofNoSchedule,PreferNoSchedule, orNoExecute. Defaults toNoSchedule.key: Key of the taint to tolerate.operator: EitherEqualorExists. Defaults toEqual.value: Value of the taint to match.
effect = "NoSchedule", key = "worker", and operator = "Exists".Required when karpenter_enabled is true.Metastore
Metastore
Notification email
Notification email
- Airflow
- Trino
- Superset
- JupyterHub
- Kyuubi
- Ranger
- The NexusOne API
- Portal
- Keycloak
email_enabled is true, since NexusOne has no working
default for it.Object storage
Object storage
Observability
Observability
Authorization header, followed by the bigpanda_webhook_token. For example,
Authorization: Bearer <token>.Required when bigpanda_enabled is true.bigpanda_enabled is true and a different certificate authority signs your
BigPanda endpoint.- Required when
monitoring_typeisembeddedandgrafana_content_enabledistrue. - Ignored when
tenant_shares_namespaceistrue. - Fails if
bigpanda_webhook_urlorbigpanda_webhook_tokenis empty.
Authorization header to authenticate each
alert.Required when bigpanda_enabled is true.bigpanda_enabled is true.grafana_content_enabled
is false.false to manage that content yourself. Grafana, Mimir,
Loki, and Tempo keep running either way.embedded for a tenant’s own
Grafana, Mimir, Loki, and Tempo, or leave as none to skip it.monitoring_type is embedded and the router runs in a different namespace.monitoring_type is embedded.alloy.monitoring.svc.cluster.local.Required when monitoring_type is embedded and you run your own Alloy collector.tracing_enabled on a running tenant, then ensure you restart the notebook servers
and Kyuubi engines that are already running.Required when monitoring_type is embedded.OpenSearch
OpenSearch
reuse_opensearch is false.Ranger and catalog naming
Ranger and catalog naming
ranger_repo_mode is group. Every tenant in the group shares the
nx1-unifiedsql-grp-<group> service in Ranger.Tenants in a group must also use the shared metastore and Ranger schema. Leave
the metastore_dbname variable empty, and keep the ranger_db_schema variable set to ranger.Use between 2 and 31 lowercase letters, numbers, or hyphens, starting with a letter or number.
Don’t include the namespace of any tenant in the group name, because NexusOne would then treat the
Ranger service as that tenant’s own.tenant: Nobody. The tenant keeps its own policies, in thenx1-unifiedsql-<namespace>service in Ranger.shared: Every other tenant that’s also set toshared. They all use thenx1-unifiedsqlservice in Ranger.group: Only the tenants that use the same value for theranger_repo_groupvariable. They all use thenx1-unifiedsql-grp-<group>service in Ranger. Setranger_repo_groupwhen you usegroup, and only then.
ranger_repo_mode variable unset, then NexusOne falls back to the deprecated single_ranger_repo variable.
A single_ranger_repo value of true becomes shared, and false becomes tenant.If you update the ranger_repo_mode value on a running tenant, then the tenant moves to the Ranger
service that matches the new value. Also, the policies you created in the old Ranger service don’t carry over
to the new one.false, the agent only extends a tenant’s default access. When true, the main site’s
agent narrows each tenant’s access to its own tmp/<tenant>/ and spark/<tenant>/ paths.Only set it to true after both sites and every tenant sharing the repo run ranger-sync 0.10.0
or later. Refer to the Deployed images page for the ranger-sync version in
each release.ranger_repo_mode instead. Setting this to true is the same as setting
ranger_repo_mode to shared.Whether a tenant shares its Ranger service, and the policies under it, with other tenants,
instead of having its own. Shows up as the nx1-unifiedsql service in Ranger, instead of
nx1-unifiedsql-<namespace>.S3 gateway
S3 gateway
s3_gateway_enabled already deploys. Leave it empty if you only need that
default instance.Each entry needs a unique name and a backend of s3, hdfs, or azure, plus a block named
after that backend with these fields:-
s3:endpoint,access_key,secret_key,region,path_style,meta_bucket,ssl_skip -
hdfs:namenodes,subpath,user,data_transfer_protection,kerberos,kerberos_spn,kerberos_user,kerberos_realm,hadoop_conf_secret,krb5_conf_secret,keytab_secret -
azure:accountandurl. Addaccess_keyorsas_tokento authenticate, or leave both out to useurlalone For example:
Secrets and credentials
Secrets and credentials
SSL/TLS and ingress
SSL/TLS and ingress
domain in their URLs.ssl_type is venafi.- Set it to
trueif your load balancer terminates TLS, so the ingress gateway adds theX-Forwarded-Proto: httpsheader to each request. - Set it to
falseif the ingress gateway terminates TLS.
ingress_type is envoy.- Set it to
httpif your load balancer terminates TLS, so traffic reaches the ingress gateway as plain HTTP. - Set it to
httpsif the ingress gateway terminates TLS.
ingress_type is envoy.Multi-site, sizing, timeouts, and versions
Multi-site, sizing, timeouts, and versions
cluster_name on each site, such as primary and standby.jupyter_tag, unless you need to
pin the hub separately.small, medium, large, or
xlarge.3.12. NexusOne can’t deploy unless you saved the
model with this Python version.-
3: Spark 3.5.6 on Scala 2.12 -
4: Spark 4.2.0 on Scala 2.13 Airflow, JupyterHub, and Kyuubi each match the Spark version automatically, so the whole tenant stays on one consistent Scala version.
dynamic: Each user gets their own volumestatic: Each user gets a folder on one shared volume
name: Unique identifier for the clusterprofile: Workload type, each tuned differently:appetlbianalytics
min_workers: Optional minimum worker countmax_workers: Optional maximum worker countautoscaling: Optional, whether to autoscale workers
Additional resources
- For what these variables actually deploy, refer to Deployment architecture.
- For how to choose a deployment model, refer to Deployment overview.

