[jira] [Created] (KAFKA-10271) Performance degradation while fetching a key from a single partition

2020-07-14 Thread Dima R (Jira)
Dima R created KAFKA-10271:
--

 Summary: Performance degradation while fetching a key from a 
single partition
 Key: KAFKA-10271
 URL: https://issues.apache.org/jira/browse/KAFKA-10271
 Project: Kafka
  Issue Type: Bug
  Components: streams
Affects Versions: 2.5.0, 2.6.0, 2.5.1
Reporter: Dima R
Assignee: Dima R
 Fix For: 2.6.0, 2.5.1


This is follow-up bug for KAFKA-10030 

WrappingStoreProvider does not take into account withPartition parameter and 
always return all existing stores, thus causing significant performance 
degradation to the caller, in case state store has many partitions.

 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)


[jira] [Created] (KAFKA-10030) Throw exception while fetching a key from a single partition

2020-05-21 Thread Dima R (Jira)
Dima R created KAFKA-10030:
--

 Summary: Throw exception while fetching a key from a single 
partition
 Key: KAFKA-10030
 URL: https://issues.apache.org/jira/browse/KAFKA-10030
 Project: Kafka
  Issue Type: Bug
  Components: streams
Affects Versions: 2.5.0
 Environment: StreamsConfig.NUM_STREAM_THREADS_CONFIG=2
Reporter: Dima R
 Fix For: 2.6.0, 2.5.1


StreamThreadStateStoreProvider#stores throws exception whenever taskId is not 
found, which is not correct behaviour in multi-threaded env where state store 
partitions are distributed among several StreamTasks. 
{code:java}
final Task task = tasks.get(keyTaskId);
if (task == null) {
 throw new InvalidStateStoreException(
 String.format("The specified partition %d for store %s does not exist.",
 storeQueryParams.partition(),
 storeName));
}{code}
Reproducible with KStream number of threads more then 1 

StoreQueryIntegrationTest#streamsConfiguration

config.put(StreamsConfig.NUM_STREAM_THREADS_CONFIG, 2);

 

Suggested solution is to not throw exception if at least one state store is 
found, which is always true when using StoreQueryParameters.withPartition



--
This message was sent by Atlassian Jira
(v8.3.4#803005)