akka/akka

Hi everyone, I think I am running into an issue because I may not understand something about the correct use of concurrency facilities in Akka. The scenario is this: I have some actor which is receiving messages from Kafka via Alpakka Kafka. It does this throughout the day. Message throughput is quite high I would say, at around 20 to 30 k msgs/sec. Periodically (three times a day) the actor receives a message that prompts it to upload a snapshot of the collected data to an AWS S3 bucket via Alpakka S3. This is done in the below code:

case PerformSnapshot => Effect
      .none
      .thenRun { s =>
        implicit val sys = ctx.system
        implicit val ec = ctx.executionContext
        val timeFormatter = time.format.DateTimeFormatter.ofPattern("yyyyMMdd").withZone(timeZone)
        val key = s"top-secred-data-bucket/${timeFormatter.format(s.currentSnapshot.timestamp)}/${s.currentSnapshot.metaData.id}"
        val self = ctx.self
        Source
          .single(s.currentCollection.values.toSeq)
          .runWith(blobSink(key))
          .onComplete {
            case Success(_) => self ! UpdateSnapshot(Some(s.currentSnapshot))
            case Failure(e) => self ! Shutdown("Failed to upload to S3", Some(e))
          }
      }

The actor and application are running fine throughout the day. The load on the application is not near its provisioned resource limits. But when the snapshot is triggered this log signals something is off and triggers the app to shut down:

Kafka commit failed after=5 ms,
commitsInProgress=0,
exception=org.apache.kafka.clients.consumer.CommitFailedException: Commit cannot be completed since the group has already rebalanced and assigned the partitions to another member. This means that the time between subsequent calls to poll() was longer than the configured max.poll.interval.ms, which typically implies that the poll loop is spending too much time message processing.

My question: Does that signal that I am not as concurrent as I think I am with the above code and the S3 upload actually blocks the actor from processing further messages? Or is that 5 ms setting simply too low?