Github user rmetzger commented on a diff in the pull request:
https://github.com/apache/flink/pull/2131#discussion_r68404148
--- Diff:
flink-streaming-connectors/flink-connector-kinesis/src/main/java/org/apache/flink/streaming/connectors/kinesis/proxy/KinesisProxy.java
---
@@ -159,42 +217,65 @@ public String getShardIterator(KinesisStreamShard
shard, String shardIteratorTyp
return kinesisClient.getShardIterator(shard.getStreamName(),
shard.getShardId(), shardIteratorType, startingSeqNum).getShardIterator();
}
+ private List<KinesisStreamShard> getShardsOfStream(String streamName,
String lastSeenShardId) throws InterruptedException {
+ List<KinesisStreamShard> shardsOfStream = new ArrayList<>();
+
+ DescribeStreamResult describeStreamResult;
+ do {
+ describeStreamResult = describeStream(streamName,
lastSeenShardId);
+
+ List<Shard> shards =
describeStreamResult.getStreamDescription().getShards();
+ for (Shard shard : shards) {
+ shardsOfStream.add(new
KinesisStreamShard(streamName, shard));
+ }
+
+ if (shards.size() != 0) {
+ lastSeenShardId = shards.get(shards.size() -
1).getShardId();
+ }
+ } while
(describeStreamResult.getStreamDescription().isHasMoreShards());
+
+ return shardsOfStream;
+ }
+
/**
* Get metainfo for a Kinesis stream, which contains information about
which shards this Kinesis stream possess.
*
+ * This method is using a "full jitter" approach described in
+ * <a
href="http://google.com";>https://www.awsarchitectureblog.com/2015/03/backoff.html</a>.
This is necessary
+ * because concurrent calls will be made by all parallel subtask's
{@link ShardDiscoverer}s. This jitter backoff
+ * approach will help distribute calls across the discoverers over time.
+ *
* @param streamName the stream to describe
* @param startShardId which shard to start with for this describe
operation (earlier shard's infos will not appear in result)
* @return the result of the describe stream operation
*/
- private DescribeStreamResult describeStream(String streamName, String
startShardId) {
+ private DescribeStreamResult describeStream(String streamName, String
startShardId) throws InterruptedException {
final DescribeStreamRequest describeStreamRequest = new
DescribeStreamRequest();
describeStreamRequest.setStreamName(streamName);
describeStreamRequest.setExclusiveStartShardId(startShardId);
DescribeStreamResult describeStreamResult = null;
String streamStatus = null;
- int remainingRetryTimes = Integer.valueOf(
-
configProps.getProperty(KinesisConfigConstants.CONFIG_STREAM_DESCRIBE_RETRIES,
Integer.toString(KinesisConfigConstants.DEFAULT_STREAM_DESCRIBE_RETRY_TIMES)));
- long describeStreamBackoffTimeInMillis = Long.valueOf(
-
configProps.getProperty(KinesisConfigConstants.CONFIG_STREAM_DESCRIBE_BACKOFF,
Long.toString(KinesisConfigConstants.DEFAULT_STREAM_DESCRIBE_BACKOFF)));
- // Call DescribeStream, with backoff and retries (if we get
LimitExceededException).
- while ((remainingRetryTimes >= 0) && (describeStreamResult ==
null)) {
+ // Call DescribeStream, with full-jitter backoff (if we get
LimitExceededException).
+ Random seed = null;
+ int attemptCount = 0;
+ while (describeStreamResult == null) { // retry until we get a
result
try {
describeStreamResult =
kinesisClient.describeStream(describeStreamRequest);
streamStatus =
describeStreamResult.getStreamDescription().getStreamStatus();
} catch (LimitExceededException le) {
- LOG.warn("Got LimitExceededException when
describing stream " + streamName + ". Backing off for "
- + describeStreamBackoffTimeInMillis + "
millis.");
- try {
-
Thread.sleep(describeStreamBackoffTimeInMillis);
- } catch (InterruptedException ie) {
- LOG.debug("Stream " + streamName + " :
Sleep was interrupted ", ie);
+ if (seed == null) {
+ seed = new Random();
}
+ long backoffMillis = fullJitterBackoff(
+ describeStreamBaseBackoffMillis,
describeStreamMaxBackoffMillis, describeStreamExpConstant, attemptCount++,
seed);
+ LOG.warn("Got LimitExceededException when
describing stream " + streamName + ". Backing off for "
+ + backoffMillis + " millis.");
+ Thread.sleep(backoffMillis);
} catch (ResourceNotFoundException re) {
throw new RuntimeException("Error while getting
stream details", re);
}
- remainingRetryTimes--;
}
if (streamStatus == null) {
--- End diff --
The `RuntimeException` below has a typo.
`Can't get stream info from ____ after`.
Also, I wonder where the number 3 is coming from.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
