Re: State Access Beyond RichCoFlatMapFunction

Posted by Sandeep khanzode on
URL: http://deprecated-apache-flink-user-mailing-list-archive.369.s1.nabble.com/State-Access-Beyond-RichCoFlatMapFunction-tp41373p41579.html

Hello,

Is there an example setup of Queryable State for a Local Embedded Environment?

I am trying to execute Flink programs from within IntelliJ. Any help would be appreciated!

Even if not, if there are other examples where QueryableState can be executed in a standalone cluster, that would also be good help. Thanks.

On 10-Feb-2021, at 9:05 AM, Kezhu Wang <[hidden email]> wrote:

(a) It is by design. For keyed state, you can only access state for that key, not others. If you want one value per key, ValueState fits more appropriate that MapState.

(b) state-processor-api aims to access/create/modify/upgrade offline savepoint but not running state. Queryable state may meet your requirement, but it is not actively developed for a while according to my observation and still beta. 

Queryable state: https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/stream/state/queryable_state.html

On February 9, 2021 at 22:09:29, Sandeep khanzode ([hidden email]) wrote:

Hello,

I am creating a class that extends RichCoFlatMapFunction. I need to connect() two streams to basically share the state of one stream in another. 

This is what I do:

private transient MapState<KeyClass, ValueClass> state;

@Override

public void open(Configuration parameters) throws Exception {
    MapStateDescriptor<KeyClass, ValueClass> stateDescriptor =
            new MapStateDescriptor<>(“abc-saved-state",
                    Types.POJO(KeyClass.class), Types.POJO(ValueClass.class));
    state = getRuntimeContext().getMapState(stateDescriptor);

This works correctly.

I have two questions:

(a) Whenever I debug, I can only see the current key in the MapState, not all the possible keys that were created before and saved. Next time, I get a hit for another key, I will only see the other key and not the rest of previous keys. Is it by design or am I missing something?

(b) Can I somehow access this state beyond the class that holds the state? I.e. can I access the state in some other class? If not, can I use the https://ci.apache.org/projects/flink/flink-docs-release-1.9/dev/libs/state_processor_api.html to do this? Is that the correct way to access the running state of one stream elsewhere in the program without corrupting it?

Your response will be greatly appreciated. I will be happy to add more details if required.

Thanks,

Sandeep Ramesh Khanzode