CODEWITHSUNDEEP
apache-flinkstateful
PengKun Liu
PengKun Liu

Reputation: 41

Why stateful source in flink require synchronized lock

I am reading the flink examples here https://ci.apache.org/projects/flink/flink-docs-release-1.8/dev/stream/state/state.html, and can not figure out the point of the existence of the synchronized block.

I have googled a lot and can not find useful answers, from flink's documentation, for each parallelism there will be a state instance, and also the run method in one instance of CounterSource will not be called from different threads, so what is the point for the below code 

val lock = ctx.getCheckpointLock

while (isRunning) {
  // output and state update are atomic
  lock.synchronized({
    ctx.collect(offset)

    offset += 1
  })
}

Upvotes: 4

Views: 999

Answers (1)

David Anderson
David Anderson

Reputation: 43499

Checkpointing occurs in another thread. The purpose of this lock is to prevent the source from modifying its state while a checkpoint is taking a snapshot of the state.

Upvotes: 3

Related Questions