Issue in applying network policy as random ports in Jobmanager and Taskmanager are visible

[Taranpreet Kaur via dev <de...@flink.apache.org>]

Dear Team,

I have raised a jira [FLINK-40516] Needed to support network policy in Jobmanager and Task Manager - ASF Jira<https://issues.apache.org/jira/browse/FLINK-40516>. I am using flink inside a Kubernetes pod and need to apply newtwork policy but random ports are getting opened although I had provided some configuration in conf.yaml as well. The example is shown below:

bash-4.4$ ss -nap | grep LISTEN
tcp   LISTEN    0      50                           :9249                               *:           users[cid:image001.png@01DD3F7A.8627ACD0]("java",pid=224,fd=271))
tcp   LISTEN    0      4096                         :8081                               *:           users[cid:image001.png@01DD3F7A.8627ACD0]("java",pid=224,fd=293))
tcp   LISTEN    0      4096                         :6123                               *:           users[cid:image001.png@01DD3F7A.8627ACD0]("java",pid=224,fd=265))
tcp   LISTEN    0      1000                         :6124                               *:           users[cid:image001.png@01DD3F7A.8627ACD0]("java",pid=224,fd=268))
tcp   LISTEN    0      4096                         :35669                              *:           users[cid:image001.png@01DD3F7A.8627ACD0]("java",pid=224,fd=286)) - random port
in taskmanager:
bash-4.4$ ss -nap |grep LISTEN
tcp   LISTEN    0      50                         :9249                            *:        users[cid:image001.png@01DD3F7A.8627ACD0]("java",pid=358,fd=277))
tcp   LISTEN    0      4096                       :41355                           *:        users[cid:image001.png@01DD3F7A.8627ACD0]("java",pid=358,fd=300)) - random port
tcp   LISTEN    0      4096                       :6121                            *:        users[cid:image001.png@01DD3F7A.8627ACD0]("java",pid=358,fd=313))
tcp   LISTEN    0      4096                       :6122                            *:        users[cid:image001.png@01DD3F7A.8627ACD0]("java",pid=358,fd=275))

Earlier two random ports were getting opened but after following configurations, it reduces to one:
jobManagerConfig:
    pekko.ask.timeout: "120s"
    client.timeout: "120s"
    state.checkpoints.num-retained: 1
    metrics.internal.query-service.port: "6126"    # ← PIN the metric query service
 taskManagerConfig:
   taskmanager.data.port: "6121"                  # ← PIN the data shuffle port
   metrics.internal.query-service.port: "6126"


Thanks & Regards
Taranpreet Kaur