A broken can have more than one instance of ZooKeeperClient. For example, SimpleAclAuthorizer creates a separate ZooKeeperClient instance when configured.
This commit makes it possible to optionally specify the name for the ZooKeeperClient instance. The name is specified only for a broker's ZooKeeperClient instances, but not for commands' and tests'.
Reviewers: Jun Rao <junrao@gmail.com>
`map` was being used to convert `Iterable[Integer]` to `Iterable[Int`]. That
operation represented 11% of total CPU time measured under load for us.
We also expect a positive impact on GC.
Reviewers: Joel Koshy <jjkoshy@gmail.com>, Ismael Juma <ismael@juma.me.uk>
- The flaky failure is caused by the fact that the main thread sometimes issues DescribeConsumerGroup request before the consumer assignment takes effect. Added a latch to make sure such situation is not going to happen.
Author: huxihx <huxi_2b@hotmail.com>
Author: huxi <huxi_2b@hotmail.com>
Author: Manikumar Reddy <manikumar.reddy@gmail.com>
Reviewers: Manikumar Reddy <manikumar.reddy@gmail.com>
Closes#6441 from huxihx/KAFKA-8098
This patch adds additional DEBUG statements in AbstractIndex.scala, OffsetIndex.scala, and TimeIndex.scala. It also changes the logging on append from DEBUG to TRACE to make DEBUG logging less disruptive, and it ensures that exceptions raised from index classes include file/offset information.
Reviewers: Jason Gustafson <jason@confluent.io>
Shutdown session expiry thread prior to closing ZooKeeper client to ensure that new clients are not created by the expiry thread and left active when returning from ZooKeeperClient.close().
Reviewers: Ismael Juma <ismael@juma.me.uk>
We verify that ZK clients are closed in tests since these can affect subsequent tests and that makes it hard to debug test failures. But because of changes to ZooKeeper client, we were checking the wrong thread name. The thread name used now is <creatorThreadName>-EventThread where creatorThreadName varies depending on the test. Fixed ZooKeeperTestHarness to check this format and fixed tests which were leaving ZK clients behind. Also added a test to make sure we can detect changes to the thread name when we update ZK clients in future.
Reviewers: Ismael Juma <ismael@juma.me.uk>, Manikumar Reddy <manikumar.reddy@gmail.com>
https://issues.apache.org/jira/browse/KAFKA-7813
Running the JMX tool without --object-name parameter, results in a NullPointerException.
*More detailed description of your change,
if necessary. The PR title and PR message become
the squashed commit message, so use a separate
comment to ping reviewers.*
*Summary of testing strategy (including rationale)
for the feature or bug fix. Unit and/or integration
tests are expected for any behaviour change and
system tests should be considered for larger changes.*
Author: huxihx <huxi_2b@hotmail.com>
Reviewers: Ewen Cheslack-Postava <ewen@confluent.io>
Closes#6139 from huxihx/KAFKA-7813
The use of consumer.poll() made the test flaky since in some cases, it doesn't wait for coordinator connection.
Reviewers: Manikumar Reddy <manikumar.reddy@gmail.com>
Adds a new listener config `max.connections` to limit the number of active connections on each listener. The config may be prefixed with listener prefix. This limit may be dynamically reconfigured without restarting the broker.
This is one of the PRs for KIP-402 (https://cwiki.apache.org/confluence/display/KAFKA/KIP-402%3A+Improve+fairness+in+SocketServer+processors). Note that this is currently built on top of PR #6022
Author: Rajini Sivaram <rajinisivaram@googlemail.com>
Reviewers: Gwen Shapira <cshapi@gmail.com>
Closes#6034 from rajinisivaram/KAFKA-7730-max-connections
Just a minor cleanup to use Java 8 lambdas vs anonymous classes in this test.
I ran all tests in the streams test suite
Reviewers: Matthias J. Sax <mjsax@apache.org>, Guozhang Wang <wangguoz@gmail.com>
When unclean leader election is enabled dynamically on brokers, we notify controller of the update before updating KafkaConfig. When processing this event, controller's decision to elect unclean leaders is based on the current KafkaConfig, so there is a small timing window when the controller may not elect unclean leader because KafkaConfig of the server was not yet updated. The commit fixes this timing window by using the existing BrokerReconfigurable interface used by other classes which rely on the current value of KafkaConfig.
Reviewers: Manikumar Reddy <manikumar.reddy@gmail.com>
To keep align with the way it handles the offset topic, TopicCommand should not be able to alter transaction topic partition count.
Author: huxihx <huxi_2b@hotmail.com>
Reviewers: Viktor Somogyi <viktorsomogyi@gmail.com>, Ismael Juma <ismael@juma.me.uk>, Manikumar Reddy <manikumar.reddy@gmail.com>
Closes#6109 from huxihx/KAFKA-7801
DynamicBrokerReconfigurationTest.testAddRemoveSaslListeners removes a listener, waits for the config to be propagated to all brokers and then validates that connections to the removed listener fail. But there is a small timing window between config update and Processor shutdown. Before validating that connections to a removed listener fail, this commit waits for all metrics of the removed listener to be deleted, ensuring that the Processors of the listener have been shutdown.
Reviewers: Manikumar Reddy <manikumar.reddy@gmail.com>
*Handle OptionException while parsing options when using console producer and print usage before die.*
Author: Suman BN <sumannewton@gmail.com>
Reviewers: Manikumar Reddy <manikumar.reddy@gmail.com>
Closes#6386 from sumannewton/console-producer-parse-printusage
After the 2.1 release, if the broker hasn't been upgrade to the latest inter-broker protocol version, the committed offsets stored in the __consumer_offset topic will get cleaned up way earlier than it should be when the offsets are loaded back from the __consumer_offset topic in GroupCoordinator, which will happen during leadership transition or after broker bounce. This patch fixes the bug by setting expireTimestamp to None if it is the default value after loading v1 offset records from __consumer_offsets.
Reviewers: Ismael Juma <ismael@juma.me.uk>, Jason Gustafson <jason@confluent.io>
Test currently checks that there were at least 5 polls when 5 connections are established with connectionQueueSize=1. But we could be doing the check just after the 5th connection before the 5th poll, so updated the check to verify that there were at least 4 polls.
Reviewers: Ismael Juma <ismael@juma.me.uk>
Metadata may be updated from the background thread, so we need to protect access to SubscriptionState. This patch restructures the metadata handling so that we only check pattern subscriptions in the foreground. Additionally, it improves the following:
1. SubscriptionState is now the source of truth for the topics that will be fetched. We had a lot of messy logic previously to try and keep the the topic set in Metadata consistent with the subscription, so this simplifies the logic.
2. The metadata needs for the producer and consumer are quite different, so it made sense to separate the custom logic into separate extensions of Metadata. For example, only the producer requires topic expiration.
3. We've always had an edge case in which a metadata change with an inflight request may cause us to effectively miss an expected update. This patch implements a separate version inside Metadata which is bumped when the needed topics changes.
4. This patch removes the MetadataListener, which was the cause of https://issues.apache.org/jira/browse/KAFKA-7764.
Reviewers: David Arthur <mumrah@gmail.com>, Rajini Sivaram <rajinisivaram@googlemail.com>
Added additional description for the "time" parameter for GetOffsetShell which adds " No offset is returned if timestamp provided is greater than recently committed record timestamp." in the description.
Author: KartikVK <karthikkalaghatgi123@gmail.com>
Reviewers: huxi <huxi_2b@hotmail.com>, Manikumar Reddy <manikumar.reddy@gmail.com>
Closes#6357 from Kartikvk1996/kartik-branch
Disable forceSync in EmbeddedZookeeper.
Increase ZK tick to allow longer maxSessionTimeout in tests.
Increase ZK client session timeout in tests.
Handle transient ZK session expiration exception in test utils for createTopic.
Author: Jun Rao <junrao@gmail.com>
Reviewers: Guozhang Wang <wangguoz@gmail.com>, Manikumar Reddy <manikumar.reddy@gmail.com>, Ismael Juma <ismael@juma.me.uk>
Closes#6354 from junrao/KAFKA-8018
Stack trace generated from the test failure shows that test failed even though threads were runnable and making progress, indicating that the timeout may be too small when test machine is slow. Increasing timeout from 10 to 15 seconds, consistent with the default wait in other tests. Thread dump also showed a lot of left over threads from other tests, so added clean up of those as well.
Reviewers: Ismael Juma <ismael@juma.me.uk>
Port 22 is used by ssh, which causes the AdminClient to throw an OOM:
> java.lang.OutOfMemoryError: Java heap space
> at java.nio.HeapByteBuffer.<init>(HeapByteBuffer.java:57)
> at java.nio.ByteBuffer.allocate(ByteBuffer.java:335)
> at org.apache.kafka.common.memory.MemoryPool$1.tryAllocate(MemoryPool.java:30)
> at org.apache.kafka.common.network.NetworkReceive.readFrom(NetworkReceive.java:112)
> at org.apache.kafka.common.network.KafkaChannel.receive(KafkaChannel.java:424)
> at org.apache.kafka.common.network.KafkaChannel.read(KafkaChannel.java:385)
> at org.apache.kafka.common.network.Selector.attemptRead(Selector.java:640)
> at org.apache.kafka.common.network.Selector.pollSelectionKeys(Selector.java:561)
> at org.apache.kafka.common.network.Selector.poll(Selector.java:472)
> at org.apache.kafka.clients.NetworkClient.poll(NetworkClient.java:535)
> at org.apache.kafka.clients.admin.KafkaAdminClient$AdminClientRunnable.run(KafkaAdminClient.java:1140)
> at java.lang.Thread.run(Thread.java:748)
>
>
Author: Manikumar Reddy <manikumar.reddy@gmail.com>
Author: Ismael Juma <ismael@juma.me.uk>
Reviewers: Ismael Juma <ismael@juma.me.uk>
Closes#6360 from omkreddy/KAFKA-7312
This patch fixes a bug in log dir reassignment where Partition.maybeReplaceCurrentWithFutureReplica would compare the entire LogEndOffsetMetadata of each replica to determine whether the reassignment has completed. If the active segments of both replicas have different base segments (for example, if the current replica had previously been cleaned and the future replica rolled segments at different points), the reassignment will never complete. The fix is to compare only the LogEndOffsetMetadata.messageOffset for each replica. Tested with a unit test that simulates the compacted current replica case.
Reviewers: Anna Povzner <anna@confluent.io>, Jason Gustafson <jason@confluent.io>
This patch fixes a regression in the replica fetcher which occurs when the replica fetcher manager simultaneously calls `removeFetcherForPartitions`, removing the corresponding partitionStates, while a replica fetcher thread attempts to truncate the same partition(s) in `truncateToHighWatermark`. This causes an NPE which causes the fetcher to crash.
This change simply checks that the `partitionState` is not null first. Note that a similar guard exists in `truncateToEpochEndOffsets`.
Reviewers: Stanislav Kozlovski <stanislav_kozlovski@outlook.com>, Jason Gustafson <jason@confluent.io>
This patch adds several new log messages to provide more information about errors during log dir movement and to make it clear when each partition movement is finished.
Reviewers: Jason Gustafson <jason@confluent.io>
In some test cases it's desirable to instantiate a subclass of `ShutdownableThread` without starting it. Since most subclasses of `ShutdownableThread` put cleanup logic in `ShutdownableThread.shutdown()`, being able to call `shutdown()` on the non-running thread would be useful.
This change allows us to avoid blocking in `ShutdownableThread.shutdown()` if the thread's `run()` method has not been called. We also add a check that `initiateShutdown()` was called before `awaitShutdown()`, to protect against the case where a user calls `awaitShutdown()` before the thread has been started, and unexpectedly is not blocked on the thread shutting down.
Reviewers : Dhruvil Shah <dhruvil@confluent.io>, Jun Rao <junrao@gmail.com>
In order to debug problems with log directory reassignments, it is helpful to know when the fetcher thread begins moving a particular partition. This patch refactors the fetch logic so that we stick to a selected partition as long as it is available and log a message when a different partition is selected.
Reviewers: Viktor Somogyi-Vass <viktorsomogyi@gmail.com>, Dong Lin <lindong28@gmail.com>, Jun Rao <junrao@gmail.com>
Currently, commitTransaction and abortTransaction wait indefinitely for the respective operation to be completed. This patch uses the producer's max block time to limit the time that we will wait. If the timeout elapses, we raise a TimeoutException, which allows the user to either close the producer or retry the operation.
Reviewers: Guozhang Wang <wangguoz@gmail.com>, Jason Gustafson <jason@confluent.io>
Use of `MetadataRequest.isAllTopics` is not consistently defined for all versions of the api. For v0, it evaluates to false. This patch makes the behavior consistent for all versions.
Reviewers: Rajini Sivaram <rajinisivaram@googlemail.com>
Per the KIP-263 discussion, we think we can improve broker restart time by avoiding performing costly disk operations when sanity checking index files for segments below recovery point on broker startup.
This PR includes the following changes:
1. Mmap the index file and populate fields of the index file on-demand rather than performing costly disk operations when creating the index object on broker startup.
2. Skip sanity checks on the time index and offset index of segments.
1. For segment with offset below the flushed point (recovery point), these segments are safely flushed so we don't need to sanity check the index files. if there are indeed data corruption on disk, given that we don't sanity check the segment file, sanity checking only the indexes adds little benefit.
2. For segment with offset above the flushed point (recovery point), we will recover these segments in `recoveryLog()` (Log.scala) in any case so sanity checking the index files for these segments is redundant.
We did experiments on a cluster with 15 brokers, each of which has ~3k segments (and there are 31.8k partitions with RF=3 which are evenly distributed across brokers; total bytes-in-rate is around 400 MBps). The results show that rolling bounce time reduces from 135 minutes to 55 minutes.
Reviewers: Ismael Juma <ismael@juma.me.uk>, Jun Rao <junrao@gmail.com>
Fail produce requests using zstd until the inter.broker.protocol.version is large enough that replicas are ensured to support it. Otherwise, followers receive the `UNSUPPORTED_COMPRESSION_TYPE` when fetching zstd data and ISRs shrink.
Reviewers: Jason Gustafson <jason@confluent.io>
Ivan Yurchenko
Radai Rosenblatt
Anna Povzner
huxihx
Bob Barrett
Boyang Chen
Rajini Sivaram
Manikumar Reddy
Matthias J. Sax
Bill Bejeck
Suman BN
Zhanxiang (Patrick) Huang
Jason Gustafson
KartikVK
Jun Rao
Mickael Maison
Colin Hicks
lambdaliu
Stanislav Kozlovski
Gardner Vickers
Gwen (Chen) Shapira
Ryan Chen
Chia-Ping Tsai
Murad
Lee Dongjin