I found this defect while inspecting [KAFKA-7293: Merge followed by groupByKey/join might violate co-partitioning](https://issues.apache.org/jira/browse/KAFKA-7293); This flag is never used now. Instead, `KStreamImpl#repartitionRequired` is now covering its functionality.
Reviewers: Matthias J. Sax <mjsax@apache.org>, Bill Bejeck <bbejeck@gmail.com>
First PR in series to inline the generic parameters of the following bytes stores:
[x] InMemoryKeyValueStore
[ ] RocksDBWindowStore
[ ] RocksDBSessionStore
[ ] MemoryLRUCache
[ ] MemoryNavigableLRUCache
[ ] (awaiting merge) InMemoryWindowStore
A number of tests took advantage of the generic InMemoryKeyValueStore and had to be reworked somewhat -- this PR covers everything related to the in-memory key-value store.
Reviewers: Guozhang Wang <wangguoz@gmail.com>, Matthias J. Sax <mjsax@apache.org>, John Roesler <john@confluent.io>, Bill Bejeck <bbejeck@gmail.com>
This fix is aiming for #2 issue pointed out within https://issues.apache.org/jira/browse/KAFKA-7672
In the current setup, we do offset checkpoint file write when EOS is turned on during #suspend, which introduces the potential race condition during StateManager #closeSuspend call. To mitigate the problem, we attempt to always write checkpoint file in #suspend call.
Reviewers: Guozhang Wang <wangguoz@gmail.com>, Matthias J. Sax <mjsax@apache.org>, John Roesler <john@confluent.io>, Bill Bejeck <bbejeck@gmail.com>
* In activeTasks.suspend, we should also close all restoring tasks as well. Closing restoring tasks would not require `task.close` as in `closeNonRunningTasks `, since the topology is not initialized yet, instead only state stores are initialized. So we only need to call `task.closeStateManager`.
* Also add @linyli001 's fix.
* Unit tests updated accordingly.
Reviewers: Matthias J. Sax <mjsax@apache.org>, John Roesler <john@confluent.io>
This is an update to the existing javadocs for KGroupedStream class.
Guozhang Wang <wangguoz@gmail.com>, Matthias J. Sax <mjsax@apache.org>, John Roesler <john@confluent.io>, Bill Bejeck <bbejeck@gmail.com>
Implemented an in-memory window store allowing for range queries. A finite retention period defines how long records will be kept, ie the window of time for fetching, and the grace period defines the window within which late-arriving data may still be written to the store.
Unit tests were written to test the functionality of the window store, including its insert/update/delete and fetch operations. Single-record, all records, and range fetch were tested, for both time ranges and key ranges. The logging and metrics for late-arriving (dropped)records were tested as well as the ability to restore from a changelog.
Reviewers: John Roesler <john@confluent.io>, Matthias J. Sax <matthias@confluent.io>, Bill Bejeck <bill@confluent.io>, Guozhang Wang <wangguoz@gmail.com>
This PR fixes the issue found in the soak testing cluster regarding using RocksDBTimestampedStore when a regular RocksDB store should have been used.
Reviewers: Guozhang Wang <wangguoz@gmail.com>, Matthias J. Sax <mjsax@apache.org>, John Roesler <john@confluent.io>
* MINOR: add test for StreamsSmokeTestDriver
Hi @bbejeck@mjsax@ableegoldman@guozhangwang ,
Please take a look at this when you get the chance.
The primary concern is adding the test. It will help us verify changes to the smoke test (such as adding suppression).
I've also added some extra output to the smoke test stdout, which will hopefully aid us in diagnosing the flaky tests.
Finally, I bundled in some cleanup. It was my intention to do that in a separate PR, but it wound up getting smashed together during refactoring.
Please let me know if you'd prefer for me to pull any of these out into a separate request.
Thanks,
-John
Also, add more output for debuggability
* cleanup
* cleanup
* refactor
* refactor
* remove redundant printlns
* Update EmbeddedKafkaCluster.java
* move to integration package
* replace early-exit on pass
* use classrule for embedded kafka
* pull in smoke test improvements from side branch
* try-with-resources
* format events instead of printing long lines
* minor formatting fix
Reviewers: Matthias J. Sax <mjsax@apache.org>, Bill Bejeck <bbejeck@gmail.com>
The change-logging stores should not bypass methods in underlying stores.
If some of you have a minute, can you take a quick look at this? I happened to notice during some other refactoring that the change-logging store layer sometimes bypasses the underlying store and instead calls across to a different layer.
It seems unexpected that it should do so, and it might actually cause problems. There was one spot where it's impossible to avoid it (in the windowed store), but I added a note justifying why we bypass the underlying store.
Thanks,
-John
* MINOR: fix bypasses in ChangeLogging stores
* fix test
Reviewers: Guozhang Wang <wangguoz@gmail.com>, Matthias J. Sax <mjsax@apache.org>, Bill Bejeck <bbejeck@gmail.com>
Refactor internal store wrapping for improved maintainability.
Reviewers: Matthias J. Sax <matthias@confluent.io>, Guozhang Wang <guozhang@confluent.io>
This condition is a fatal error, so error level is warranted, to provide more context on why Streams shuts down.
Reviewers: Matthias J. Sax <matthias@confluent.io>, Guozhang Wang <guozhang@confluent.io>
1. In the caching layer's flush listener call, we should always write to the underlying store, before flushing (see #4331 's point 4) for detailed explanation). When fixing 4331, it only touches on KV stores, but it turns out that we should fix for window and session store as well.
2. Also apply the optimization that was in session-store already: when the new value bytes and old value bytes are all null (this is possible e.g. if there is a put(K, V) followed by a remove(K) or put(K, null) and these two operations only hit the cache), upon flushing this mean the underlying store does not have this value at all and also no intermediate value has been sent to downstream as well. We can skip both putting a null to the underlying store as well as calling the flush listener sending `null -> null` in this case.
Modifies corresponding unit tests.
Reviewers: John Roesler <john@confluent.io>, Matthias J. Sax <matthias@confluent.io>, Bill Bejeck <bill@confluent.io>
JUnit 4.13 fixes the issue where `Category` and `Parameterized` annotations
could not be used together. It also deprecates `ExpectedException` and
`assertThat`. Given this, we:
- Replace `ExpectedException` with the newly introduced `assertThrows`.
- Replace `Assert.assertThat` with `MatcherAssert.assertThat`.
- Annotate `AbstractLogCleanerIntegrationTest` with `IntegrationTest` category.
Reviewers: Ewen Cheslack-Postava <ewen@confluent.io>, David Arthur <mumrah@gmail.com>
This PR addressed the following test failure:
```
java.lang.AssertionError:
Expected: a string starting with "process-state-manager-test Failed to write offset checkpoint file to ["
but: was "[AdminClient clientId=adminclient-874] Connection to node -1 (localhost/127.0.0.1:8080) could not be established. Broker may not be available."
```
Reviewers: Bill Bejeck <bill@confluent.io>, John Roesler <john@confluent.io>, Guozhang Wang <guozhang@confluent.io>
This PR adds a upgrade notes and changes examples to use the bootstrap-server.
Author: Viktor Somogyi-Vass <viktorsomogyi@gmail.com>
Reviewers: Srinivas <srinivas96alluri@gmail.com>, Manikumar Reddy <manikumar.reddy@gmail.com>
Closes#6118 from viktorsomogyi/topiccommand-adminclient-doc
Reviewers: Damian Guy <damian@confluent.io>, John Roesler <john@confluent.io>, Boyang Chen <bchen11@outlook.com>, Matthias J. Sax <matthias@confluent.io>, Bill Bejeck <bill@confluent.io>
#2972 tried to fix a bug about flushing operation, but it was not complete, since findSessions(key, earliestEnd, latestStart) does not guarantee to only return a single entry since its semantics are to return any sessions whose end > earliestEnd and whose start < latestStart.
I've tried various ways to fix it completely and I ended up having to add a single-point query to the public ReadOnlySessionStore API for the exact needed semantics. It is used for flushing to read the old values (otherwise the wrong old values will be sent downstreams, hence it is a correctness issue) and also for getting the value for value-getters (it is for perf only).
This PR enables BloomFilters for RocksDB to speed up point lookups.
The request for this has been around for some time - https://issues.apache.org/jira/browse/KAFKA-4850
For testing, I've done the following
Ran the standard streams suite of unit and integration tests
Kicked off the simple benchmark test with bloom filters enabled
Kicked off the simple benchmark test with bloom filters not enabled
Kicked off streams system tests
Matthias J. Sax <matthias@confluent.io>, Guozhang Wang <wangguoz@gmail.com>, John Roesler <john@confluent.io>
Let findSessions(final K key) to call on underlying bytes store directly, using the more restricted range.
Fix the conservative upper range for multi-key range in session schema.
Minor: removed unnecessary private WrappedSessionStoreBytesIterator class as it is only used in unit test.
Minor: removed unnecessary schema#init function by using the direct bytes-to-binary function.
Please read the original PR for more detailed explanation of the root cause of the bug.
Reviewers: Bill Bejeck <bill@confluent.io>, Damian Guy <damian@confluent.io>, John Roesler <john@confluent.io>
Note we can only remove this handling in 2.2 but not in 2.1 since #6124 is only in 2.2.
Reviewers: Bill Bejeck <bill@confluent.io>, Damian Guy <damian@confluent.io>, Matthias J. Sax <matthias@confluent.io>
In RocksDBStore.openDB we call
Files.createDirectories(dir.getParentFile().toPath());
return RocksDB.open(options, dir.getAbsolutePath());
We would also add the absolute file path as well to avoid the extra logging.
Reviewers: Matthias J. Sax <matthias@confluent.io>, Guozhang Wang <wangguoz@gmail.com>
I've observed several reports of sudden unexpected streamthread shutdown with the log entry like:
State transition from PENDING_SHUTDOWN to DEAD
but there is no related error logs before this line at all. I suspect this is because we intentionally do not log for KafkaException and there's some edge cases where we miss internally and hence caused this. I'm adding the ERROR level log entry here in order to reveal more information in case I saw this again in the future.
Reviewers: Matthias J. Sax <matthias@confluent.io>
Reviewers: Bill Bejeck <bill@confluent.io>, John Roesler <john@confluent.io>, Guozhang Wang <guozhang@confluent.io>, Ryanne Dolan <ryannedolan@gmail.com>, Ismael Juma <ismael@confuent.io>
The existing javadoc for PartitionGroup is a little confusing.
It's relatively important for these concepts to be clear, since
they form the basis for stream-time in Kafka Streams.
Reviewers: Matthias J. Sax <matthias@confluent.io>, Guozhang Wang <wangguoz@gmail.com>
KIP-291 Implementation : Added code to separate controller connections and requests from the data plane.
Tested with local deployment that the controller request are handled by the control plane and other requests are handled by the data plane.
Also added unit tests in order to test the functionality.
Author: Lucas Wang <luwang@linkedin.com>,
Author: Mayuresh Gharat <gharatmayuresh15@gmail.com>
Reviewers: Joel Koshy <jjkoshy@gmail.com>, Jun Rao <junrao@gmail.com>
Using AdminClient#alterConfigs, topic `retention.ms` property can be assigned to a value lesser than -1. This leads to inconsistency while describing the topic configuration. We should not allow values lesser than -1.
Author: Kamal Chandraprakash <kamal.chandraprakash@gmail.com>
Reviewers: Manikumar Reddy <manikumar.reddy@gmail.com>,Matthias J. Sax <matthias@confluent.io>
Closes#6082 from kamalcph/KAFKA-7781
We recently improved the handling of the InternalTopicManager retries with #6085. The AdminClient will throw an InvalidTopicException if the topic is not found. We need to ignore that exception as when calling AdminClient#describe we may not have had a chance to create the topic yet, especially with the case of internal topics
I've created a new test asserting that when an InvalidTopicException is thrown when the topic is not found we continue on.
Reviewers: John Roesler <john@confluent.io>, Guozhang Wang <guozhang@confluent.io>
Lee Dongjin
A. Sophie Blee-Goldman
Bill Bejeck
Boyang Chen
Guozhang Wang
asutosh936
Chia-Ping Tsai
Matthias J. Sax
John Roesler
Filipe Agapito
Ismael Juma
Viktor Somogyi-Vass
cadonna
Richard Yu
Jingguo Yao
Mayuresh Gharat
Kamal Chandraprakash