Re: How to setup HA properly with Kubernetes Standalone Application Cluster

[陳昌倬]

On Fri, May 14, 2021 at 02:00:41PM +0200, Fabian Paul wrote:
> Hi Chen,
> 
> Can you tell us a bit more about the job you are using? 
> The intended behaviour you are seeking can only be achieved 
> If the Kubernetes HA Services are enabled [1][2].
> Otherwise the job cannot recall past checkpoints.

Hi Fabian,

Thanks for the response, the following is our cluster setup.

The HA setting we used is the same one as we used to used in session
cluster, so HA shall work. The following is setting we used for HA:

    high-availability: 
org.apache.flink.kubernetes.highavailability.KubernetesHaServicesFactory
    high-availability.storageDir: gs://a-bucket/recovery
    kubernetes.cluster-id: a-cluster-id
    kubernetes.context: a-context
    kubernetes.namespace: a-namespace

The following is Job we used for standalone application cluster.
Basically the content is almost the same as we used to used in session
cluster, except that it is Job, not Deployment, and the args are
different:

    apiVersion: batch/v1
    kind: Job
    spec:
      template:
        spec:
          restartPolicy: OnFailure
          containers:
            - name: jobmanager
              env:
                - name: FLINK_CONF_FILE
                  value: /opt/flink/conf/flink-conf.yaml
                - name: HADOOP_CLASSPATH
                  value: 
/opt/flink/conf:/opt/hadoop-3.1.1/share/hadoop/common/lib/*:/opt/hadoop-3.1.1/share/hadoop/common/*:/opt/hadoop-3.1.1/share/hadoop/hdfs:/opt/hadoop-3.1.1/share/hadoop/hdfs/lib/*:/opt/hadoop-3.1.1/share/hadoop/hdfs/*:/opt/hadoop-3.1.1/share/hadoop/mapreduce/lib/*:/opt/hadoop-3.1.1/share/hadoop/mapreduce/*:/opt/hadoop-3.1.1/share/hadoop/yarn:/opt/hadoop-3.1.1/share/hadoop/yarn/lib/*:/opt/hadoop-3.1.1/share/hadoop/yarn/*:/contrib/capacity-scheduler/*.jar
                - name: HADOOP_CONF_DIR
                  value: /opt/flink/conf
                - name: GOOGLE_APPLICATION_CREDENTIALS
                  value: /opt/flink/conf/gcp-service-account.json
              args:
                - standalone-job
                - --fromSavepoint
                - gs://a-savepoint
                - --job-classname
                - com.example.my.application

-- 
ChangZhuo Chen (陳昌倬) czchen@{czchen,debian}.org
http://czchen.info/
Key fingerprint = BA04 346D C2E1 FE63 C790  8793 CC65 B0CD EC27 5D5B

signature.asc
Description: PGP signature