Any reason not to call KafkaConsumer.poll() with Long.MAX_VALUE? - apache-kafka

Question

For my app that consume Kafka messages, is there any reason not to call KafkaConsumer.poll(millisecs) with Long.MAX_VALUE for millsecs?
I see many code pass in something like 1000milli, but wouldn't Long.MAX_VALUE be ideal, in that, just wait there and be trigged when there's a message appears in the topic.
What's the point of having poll() be called with 1sec since if you can pass in Long.MAX_VALUE?
Just to clearly illustrate the code:
public class MyConsumer implements Runnable {
#Override
public void run() {
while (true) {
ConsumerRecords<String, LogLine> records = consumer.poll(Long.MAX_VALUE);
// does something
}
}
}

Answer 1

Basically you can call poll() with Long.MAX_VALUE in every queue, unless you want do something special if queue does not have data in appropriate time.
For example suppose you have two topics A and B in Kafka and topic A has higher priority. In your code you want poll on A, and if A has not record, poll and (possibly) retrieve data from B. In this case you must pass a reasonable timeout to poll()
while (true) {
ConsumerRecords<String, LogLine> recordsA = consumerA.poll(1000);
if (!records.isEmpty()) {
// does something with recordsA
continue;
}
else {
ConsumerRecords<String, LogLine> recordsB = consumerB.poll(1000);
// does something with recordsB
}
}