pod configured with ClusterScopeIndex, the operator cannot access correct port

[enumz]

When configure different ports for different pods, it seems operator try access clickhouse instance only through chPort which in the operator configuration.

For example: chi-examples/15-hostNetwork-02-simple-port-distribution.yaml

and chPort is 13000

E0414 16:56:04.655493 6 connection.go:83] connect():FAILED Ping(http://:@***:13000/). Err: dial tcp ****:13000 : connect: connection refused

E0414 16:56:04.655530 6 connection.go:170] Exec():FAILED connect(http://:@***:13000/) for SQL: CREATE DATABASE IF NOT EXISTS "default" Engine = Ordinary

[roshanths]

Hi @enumz, .spec.chPort considered as deprecated right now. Just to confirm can you see your change in the clickhouse-operator configuration file, because all these parameters are comming from this file: https://github.com/Altinity/clickhouse-operator/blob/7ec06b79eb70430d546f9a085c92a5339c8f3e4a/config/config.yaml#L85

[enumz]

Hi @roshanths , I think the http port should be acquired at runtime, not in the config file. Because CHI support different shards or different replicas use different ports:

apiVersion: "clickhouse.altinity.com/v1"
kind: "ClickHouseInstallation"
metadata:
  name: "demo"
spec:
  configuration:
    clusters:
      - name: "demo"
        layout:
          replicasCount: 3
          shards:
            - templates:
                hostTemplate: shard-00-template
            - templates:
                hostTemplate: shard-01-template

  templates:
    hostTemplates:
      - name: shard-00-template
        spec:
          tcpPort: 9000
          httpPort: 10000
          interserverHTTPPort: 11000
      - name: shard-01-template
        spec:
          tcpPort: 9001
          httpPort: 10001
          interserverHTTPPort: 11001

  templates:
    hostTemplates:
      - name: port-distribution
        portDistribution:
          - type: ClusterScopeIndex
        spec:
          tcpPort: 10000
          httpPort: 11000
          interserverHTTPPort: 12000

[sunsingerus]

Cluster has to have the same port (as specified in operator configuration) on all nodes in order schema management functions to be executed by the operator