Alauda Data Services Analytical Database E1
  • English

    • Instance details

    This page explains how to read a live instance: how topology maps onto Kubernetes objects, what the status fields mean, and which endpoints exist.

    TOC

    TopologyObject namesEndpointsStatus fieldsHealth

    Topology

    kubectl -n <namespace> get chi e1-demo -o yaml

    Topology is declared under spec.configuration.clusters[].layout and derived as follows:

    • shardsCount × replicasCount = number of hosts.
    • Each host is one StatefulSet with replicas: 1, holding one pod.
    • Both counts default to 1, and each is raised automatically if you explicitly enumerate more shards or replicas than the count states.

    So a shardsCount: 3, replicasCount: 2 cluster is 6 StatefulSets, 6 pods, 6 PVC sets — not 3 StatefulSets of 2 pods each.

    Pod management is OrderedReady and the update strategy is RollingUpdate.

    Object names

    ObjectPattern
    Instance Serviceclickhouse-{chi}
    Cluster Servicecluster-{chi}-{cluster}
    Shard Serviceshard-{chi}-{cluster}-{shard}
    StatefulSet / host Servicechi-{chi}-{cluster}-{shard}-{replica}
    Podchi-{chi}-{cluster}-{shard}-{replica}-0
    Common config ConfigMapchi-{chi}-common-configd
    Common users ConfigMapchi-{chi}-common-usersd
    Host config ConfigMapchi-{chi}-deploy-confd-{cluster}-{shard}-{replica}
    NOTE

    The cluster and shard Services are named by the patterns above but are not created by default. The operator returns no object for either unless you set a clusterServiceTemplate or shardServiceTemplate. Out of the box you get exactly two kinds of Service: one per instance, and one per host.

    Endpoints

    PortNamePurpose
    8123httpHTTP interface, and what the operator itself uses for queries.
    9000tcpNative protocol, used by clickhouse-client.
    9009interserverReplica-to-replica traffic. Not for clients.

    status.endpoint carries the instance-wide address, and status.fqdns lists per-host addresses. See Access the instance for how to reach them and what the default Service type is.

    Status fields

    kubectl -n <namespace> get chi e1-demo -o jsonpath='{.status}' | jq
    FieldMeaning
    statusInProgress, Completed, or Terminating.
    clusters, shards, replicas, hostsCounts of the reconciled topology.
    pods, pod-ips, fqdnsPer-host pod names, pod IPs, and hostnames.
    endpointAddress of the instance-wide Service.
    action, actionsMost recent reconcile action, and a bounded history.
    error, errorsMost recent error, and a bounded history.
    taskID, taskIDsStarted, taskIDsCompletedUpdate tracking, matching spec.taskID.
    updated, added, deleted, deleteObjects touched by the last reconcile.
    chop-version, chop-commit, chop-date, chop-ipWhich operator build and pod reconciled this instance.
    normalized, normalizedCompletedThe spec after defaulting and template merging — useful for seeing what the operator actually acted on.

    status.normalized is the most useful field when a setting appears not to have taken effect: it shows the merged result of your spec, any templates it used, and the operator's defaults.

    Health

    status.status: Completed means the operator finished its reconcile pass. On release 4.4 that pass will not complete until every host's StatefulSet is ready, and the status is demoted back to InProgress if a host degrades afterwards.

    Confirm health against the workload itself:

    kubectl -n <namespace> get statefulset -l clickhouse.altinity.com/chi=e1-demo \
  -o custom-columns=NAME:.metadata.name,DESIRED:.spec.replicas,READY:.status.readyReplicas

    Every row must show READY equal to DESIRED. That is the operator's own definition of a ready host.