无法从Scala连接Kubernetes中运行的ScyllaDB集群求助

我在Kubernetes中部署了ScyllaDB集群，尝试用Scala的Phantom库连接时遇到连接失败的问题，错误提示All host(s) tried for query failed (tried: /10.98.139.252 (com.datastax.driver.core.exceptions.TransportException: [/10.98.139.252] Cannot connect))，替换Pod IP、Service IP、Service名称都无效。

我的部署与代码情况 #

Kubernetes部署步骤 #

按照官方文档执行以下命令部署集群：

$ minikube start --cpus=6
$ eval $(minikube docker-env)
$ git clone https://github.com/scylladb/scylla-operator.git
$ cd scylla-operator
$ kubectl apply -f examples/common/cert-manager.yaml
$ kubectl wait --for condition=established crd/certificates.cert-manager.io crd/issuers.cert-manager.io
$ kubectl -n cert-manager rollout status deployment.apps/cert-manager-webhook
$ kubectl apply -f deploy/operator.yaml
$ kubectl wait --for condition=established crd/scyllaclusters.scylla.scylladb.com
$ kubectl -n scylla-operator rollout status deployment.apps/scylla-operator
$ kubectl create -f examples/generic/cluster.yaml

集群状态验证结果：

$ kubectl -n scylla get pods
NAME                                    READY   STATUS    RESTARTS   AGE
simple-cluster-us-east-1-us-east-1a-0   2/2     Running   0          90m
simple-cluster-us-east-1-us-east-1a-1   2/2     Running   0          80m
simple-cluster-us-east-1-us-east-1a-2   0/2     Pending   0          77m

$ kubectl -n scylla get svc
NAME                                    TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)                                                                                                          AGE
simple-cluster-client                   ClusterIP   10.98.139.252   <none>        7000/TCP,7001/TCP,9042/TCP,9142/TCP,19042/TCP,19142/TCP,7199/TCP,10001/TCP,9180/TCP,5090/TCP,9100/TCP,9160/TCP   90m
simple-cluster-us-east-1-us-east-1a-0   ClusterIP   10.98.37.120    <none>        7000/TCP,7001/TCP,9042/TCP,9142/TCP,19042/TCP,19142/TCP,7199/TCP,10001/TCP,9180/TCP,5090/TCP,9100/TCP,9160/TCP   90m
simple-cluster-us-east-1-us-east-1a-1   ClusterIP   10.105.171.34   <none>        7000/TCP,7001/TCP,9042/TCP,9142/TCP,19042/TCP,19142/TCP,7199/TCP,10001/TCP,9180/TCP,5090/TCP,9100/TCP,9160/TCP   90m
simple-cluster-us-east-1-us-east-1a-2   ClusterIP   10.110.32.166   <none>        7000/TCP,7001/TCP,9042/TCP,9142/TCP,19042/TCP,19142/TCP,7199/TCP,10001/TCP,9180/TCP,5090/TCP,9100/TCP,9160/TCP   90m

$ kubectl exec -n scylla -it simple-cluster-us-east-1-us-east-1a-0 -- nodetool status
Defaulted container "scylla" out of: scylla, scylla-manager-agent, sidecar-injection (init)
Datacenter: us-east-1
=====================
Status=Up/Down
|/ State=Normal/Leaving/Joining/Moving
--  Address        Load       Tokens       Owns    Host ID                               Rack
UN  10.98.37.120   108.99 KB  256          ?       a1d76285-964f-48b1-b87d-2e2fafa96877  us-east-1a
UN  10.105.171.34  201.48 KB  256          ?       c2204fec-b713-4fd5-ba47-d8b72505ac12  us-east-1a

Note: Non-system keyspaces don't have the same replication settings, effective ownership information is meaningless

Scala代码连接尝试 #

使用Phantom库编写的连接代码：

schema.scala

abstract class MyDataSchema extends Table[MyDataSchema, MyData] {

  override def tableName: String = "mydata"

  object myColumn extends OptionalStringColumn with PartitionKey {
    override def name: String = "myColumn"
  }
}

connection.scala

class ScyllaDBConnection extends ScyllaDBConnectionTrait {

  override def connectionProvider: CassandraConnection = {
    val applicationConf: Config = ConfigFactory.load("application.conf")
    val keyspace: String = applicationConf.getString("keyspace.keyspaceName")
    val connection: CassandraConnection =
      ContactPoints(List("10.98.139.252")) //10.98.139.252 is IP of Service
        .keySpace(keyspace)
    connection
  }

}

database.scala

class MyDB(override val connector: CassandraConnection)
  extends Database[MyDB](connector) {
  object MyData extends MyDataSchema with Connector
}

运行后出现错误：

All host(s) tried for query failed (tried: /10.98.139.252 (com.datastax.driver.core.exceptions.TransportException: [/10.98.139.252] Cannot connect))

解决方法 #

1. 本地运行Scala代码的网络配置 #

如果Scala代码在minikube集群外部（本地机器）运行，ClusterIP类型的Service无法直接访问，需要先做端口转发：

kubectl port-forward -n scylla svc/simple-cluster-client 9042:9042

然后修改Scala代码中的连接地址为localhost：

ContactPoints(List("localhost"))
  .keySpace(keyspace)

2. 同集群内部署Scala服务的网络配置 #

如果Scala服务部署在同一个K8s集群中，直接使用Service的DNS名称即可，无需IP：

ContactPoints(List("simple-cluster-client.scylla.svc.cluster.local"))
  .keySpace(keyspace)

DNS格式为[service-name].[namespace].svc.cluster.local，若Scala服务与ScyllaDB不在同一命名空间，必须使用完整DNS地址。

3. 检查ScyllaDB端口状态 #

确认ScyllaDB的CQL端口（9042）是否正常监听，可进入Pod验证：

kubectl exec -n scylla simple-cluster-us-east-1-us-east-1a-0 -- nc -zv localhost 9042

另外，第三个Pod处于Pending状态，可能是minikube资源不足导致，可调整minikube资源配置或删除该Pending Pod，避免影响集群状态。

4. 优化Phantom库连接配置 #

Phantom基于Datastax Java Driver，可添加超时配置提升连接稳定性：

val connection: CassandraConnection =
  ContactPoints(List("localhost"))
    .withClusterBuilder(_.withPort(9042)
      .withSocketOptions(SocketOptions.builder().setConnectTimeoutMillis(10000).build()))
    .keySpace(keyspace)

内容的提问来源于stack exchange，提问作者Zaryab Ali