gaborgsomogyi commented on a change in pull request #19096: [SPARK-21869][SS] A
cached Kafka producer should not be closed if any task is using it - adds inuse
tracking.
URL: https://github.com/apache/spark/pull/19096#discussion_r264184414
##########
File path:
external/kafka-0-10-sql/src/test/scala/org/apache/spark/sql/kafka010/CachedKafkaProducerSuite.scala
##########
@@ -35,43 +36,72 @@ class CachedKafkaProducerSuite extends SharedSQLContext
with PrivateMethodTester
CachedKafkaProducer.clear()
}
- test("Should return the cached instance on calling getOrCreate with same
params.") {
- val kafkaParams = new ju.HashMap[String, Object]()
- kafkaParams.put("acks", "0")
- // Here only host should be resolvable, it does not need a running
instance of kafka server.
- kafkaParams.put("bootstrap.servers", "127.0.0.1:9022")
- kafkaParams.put("key.serializer", classOf[ByteArraySerializer].getName)
- kafkaParams.put("value.serializer", classOf[ByteArraySerializer].getName)
- val producer = CachedKafkaProducer.getOrCreate(kafkaParams)
- val producer2 = CachedKafkaProducer.getOrCreate(kafkaParams)
- assert(producer == producer2)
-
- val cacheMap = PrivateMethod[ConcurrentMap[Seq[(String, Object)],
KP]]('getAsMap)
- val map = CachedKafkaProducer.invokePrivate(cacheMap())
+ test("Should return the cached instance on calling acquire with same
params.") {
+ val kafkaParams: ju.HashMap[String, Object] = generateKafkaParams
+ val producer = CachedKafkaProducer.acquire(kafkaParams)
+ val producer2 = CachedKafkaProducer.acquire(kafkaParams)
+ assert(producer.kafkaProducer == producer2.kafkaProducer)
+ assert(producer.getInUseCount == 2)
+ val map = CachedKafkaProducer.getAsMap
assert(map.size == 1)
}
- test("Should close the correct kafka producer for the given kafkaPrams.") {
+ test("Should return the new instance on calling acquire with different
params.") {
+ val kafkaParams: ju.HashMap[String, Object] = generateKafkaParams
+ val producer = CachedKafkaProducer.acquire(kafkaParams)
+ kafkaParams.remove("ack") // mutate the kafka params.
+ val producer2 = CachedKafkaProducer.acquire(kafkaParams)
+ assert(producer.kafkaProducer != producer2.kafkaProducer)
+ assert(producer.getInUseCount == 1)
+ assert(producer2.getInUseCount == 1)
+ val map = CachedKafkaProducer.getAsMap
+ assert(map.size == 2)
+ }
+
+ test("Should return the cached instance, even if auth tokens are set up.") {
+ // TODO.
+ // Question: What happens when a delegation token value is changed for a
given producer?
+ // we would need to recreate a kafka producer.
+ }
+
+ test("Remove an offending kafka producer from cache.") {
Review comment:
I see the intention here but this test can pass if a producer never ever
created.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
