Hello,
Stuck at this time. Any help will be appreciated.
I am able to create a queryable state and also query the state. Everything
works correctly.
KeyedStream<Tuple2<Key, VO>, Key> stream = sourceStream.keyBy(t2 -> t2.f0);
stream.asQueryableState("queryableVO");
I deploy this on a Kubernetes cluster with Flink standalone-job and
KubernetesHAFactory.
There are two states created. One is the operator and keyed state which is
stored in a RocksDB Backend in S3.
The other is the HA state maintained by Kubernetes in S3.
If anything changes in the job main class (like removing operators etc.), the
upgrade does not work seamlessly and I have to delete the HA state from S3.
If I delete the S3 state for HA, the queryable state becomes unusable i.e. I
cannot query from the state anymore. Interestingly, the other operator and
keyed states in RocksDB backend are still accessible! Just not the queryable
state.
When I check the UI, I see the checkpointed state for the queryable stream has
a data size of approx ~50-60KB. But I still cannot query it.
Thanks,
Sandeep
