[
https://issues.apache.org/jira/browse/FLINK-40162?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18097035#comment-18097035
]
Milind L commented on FLINK-40162:
----------------------------------
I've taken a look at the code, and if this seems valid, I would like to try
fixing this.
> Fast-running FlinkSessionJob does not reflect actual state of the job
> ---------------------------------------------------------------------
>
> Key: FLINK-40162
> URL: https://issues.apache.org/jira/browse/FLINK-40162
> Project: Flink
> Issue Type: Bug
> Components: Kubernetes Operator
> Affects Versions: kubernetes-operator-1.15.0
> Reporter: Milind L
> Priority: Major
>
> On creating a FlinkSessionJob which runs quickly (1-5s for me, I suppose the
> exact time depends on the cluster), the `status.jobStatus.state` is stuck at
> `RECONCILING` and `status.state` is stuck at `DEPLOYED`, even though the
> Flink Web UI shows the job state as "FINISHED"
> Exact reproduction:
> The FlinkDeployment:
> {code:yaml}
> apiVersion: flink.apache.org/v1beta1
> kind: FlinkDeployment
> metadata:
> name: sessioncluster
> namespace: default
> spec:
> image: sleep-job:1.0 # this image has my job jar on the classpath
> imagePullPolicy: Never
> flinkVersion: v2_1
> serviceAccount: flink
> flinkConfiguration:
> taskmanager.numberOfTaskSlots: "2"
> execution.checkpointing.interval: "2000"
> jobManager:
> resource:
> cpu: 0.5
> memory: 1024m
> taskManager:
> resource:
> cpu: 0.5
> memory: 1024m
> {code}
> The FlinkSessionJob and the code for the job:
> {code:yaml}
> apiVersion: flink.apache.org/v1beta1
> kind: FlinkSessionJob
> metadata:
> name: sleep-job
> spec:
> deploymentName: sessioncluster
> job:
> entryClass: whichever.sleepjob.SleepJob
> args:
> - "1"
> parallelism: 1
> {code}
>
> {code:java}
> package whichever.sleepjob;
> import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
> public class SleepJob {
> public static void main(String[] args) throws Exception {
> if (args.length < 1) {
> System.err.println("Usage: SleepJob <seconds>");
> System.exit(1);
> }
> long sleepMs = Long.parseLong(args[0]) * 1000L;
> StreamExecutionEnvironment env =
> StreamExecutionEnvironment.getExecutionEnvironment();
> env.fromElements("sleep")
> .map(value -> {
> Thread.sleep(sleepMs);
> return "done";
> })
> .print();
> env.execute("SleepJob");
> }
> }
> {code}
>
> To reproduce this, the execution.checkpointing.interval must be set to any
> value and state.checkpoints.dir should be unset. This seems to be a valid
> configuration, because there are no issues with running the job itself.
>
> The `status.error` field gives more information:
> {code}
> {\"type\":\"org.apache.flink.kubernetes.operator.exception.UpgradeFailureException\",\"message\":\"Latest
> checkpoint not externally addressable, Manual restore
> required.\",\"additionalMetadata\":{},\"throwableList\":[]}",
> {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
