KAFKA-13823 Feature flag changes from KIP-778 by mumrah · Pull Request #12036 · apache/kafka

mumrah · 2022-04-12T19:58:32Z

This PR includes the changes to feature flags that were outlined in KIP-778.

Single finalized version level

The UpdateFeatures RPC and the FeatureLevelRecord were updated to remove the max version level. Other changes to the RPC from KIP-778 include:

Adding dry-run to the RPC so the controller can actually attempt the upgrade (rather than the client)
Introduce a upgrade type enum instead of "allowDowngrade" boolean

kafka-features.sh changes

This tool was overhauled in KIP-778 and now includes sub-commands.

$ ./bin/kafka-features.sh --help
usage: kafka-features [-h] --bootstrap-server BOOTSTRAP_SERVER [--command-config COMMAND_CONFIG] {describe,upgrade,downgrade,disable} ...

This tool manages feature flags in Kafka.

positional arguments:
  {describe,upgrade,downgrade,disable}
    describe             Describe one or more feature flags.
    upgrade              Upgrade one or more feature flags.
    downgrade            Upgrade one or more feature flags.
    disable              Disable one or more feature flags. This is the same as downgrading the version to zero.

optional arguments:
  -h, --help             show this help message and exit
  --bootstrap-server BOOTSTRAP_SERVER
                         A comma-separated list of host:port pairs to use for establishing the connection to the Kafka cluster.
  --command-config COMMAND_CONFIG
                         Property file containing configs to be passed to Admin Client.

Help output from the "upgrades" sub-command:

$ ./bin/kafka-features.sh upgrade --help
usage: kafka-features --bootstrap-server BOOTSTRAP_SERVER upgrade [-h] [--feature FEATURE] [--version VERSION] [--release RELEASE] [--dry-run]

optional arguments:
  -h, --help             show this help message and exit
  --dry-run              Perform a dry-run of this upgrade operation.

Upgrade specific features:
  --feature FEATURE      A feature flag to upgrade. This option may be repeated for upgrading multiple feature flags.
  --version VERSION      The version to upgrade to.

Upgrade to feature level defined for a given release:
  --release RELEASE

Refer to KIP-778 for more details on the new command structure.

* Remove finalized max version * Add support for UpdateFeatures on KRaft controller * Changes to UpdateFeatures RPC * Changes to FeatureLevelRecord

cmccabe · 2022-04-12T21:48:58Z

+        return validateOnly;
+    }
+
+    public UpdateFeaturesOptions dryRun(boolean dryRun) {


looks like this is a lingering reference to dryRun. it should be validateOnly like the others...

I wasn't sure if we needed to leave it as-is for compatibility reasons.

Changing the name of the parameter doesn't change the binary compatibility, so it should be fine. One advantage of Java not having named parameters, I guess.

Looking at this again, I can see that I forgot that it isn't just the parameter name, but also the method name.

I guess then let's just deprecate this one and add the validateOnly one

cmccabe · 2022-04-12T21:53:29Z


    private final ClusterControlManager clusterControlManager;
    private final TimelineLong nextProducerId; // Initializes to 0
+    private final TimelineInteger brokerId;


Since these are all changed at the same time, it would be better to have a TimelineObject<LatestProducerIdBlock>. Then you could just have a a long, int, long or whatever.

cmccabe · 2022-04-12T22:05:03Z

looks like there is a build error

dengziming

Thanks for this PR, left a small suggestion.

dengziming · 2022-04-13T10:25:35Z

      lifecycleManager.start(() => metadataListener.highestMetadataOffset,
        BrokerToControllerChannelManager(controllerNodeProvider, time, metrics, config,
          "heartbeat", threadNamePrefix, config.brokerSessionTimeoutMs.toLong),
-        metaProps.clusterId, networkListeners, supportedFeatures)


supportedFeatures is no longer used so we can remove it from BrokerServer

dengziming · 2022-04-13T10:27:14Z

 import org.apache.kafka.common.{ClusterResource, Endpoint}
-import org.apache.kafka.controller.{Controller, QuorumController, QuorumControllerMetrics}
-import org.apache.kafka.metadata.{KafkaConfigSchema, VersionRange}
+import org.apache.kafka.controller.{Controller, QuorumController, QuorumControllerMetrics, QuorumFeatures}


It seems QuorumFeatures and FinalizedControllerFeatures is missing?

Could you clarify what you mean by "missing"?

FinalizedControllerFeatures is the return type we use in Controller for the finalizedFeatures() call.

QuorumFeatures is used in this class down on line 163

val quorumFeatures = QuorumFeatures.create(config.nodeId, QuorumFeatures.defaultFeatureMap())

Sorry, I realized that I was missing these entire files. I've pushed them now

cmccabe · 2022-04-13T19:28:49Z

-      {"name": "AllowDowngrade", "type": "bool", "versions": "0+",
-        "about": "When set to true, the finalized feature version level is allowed to be downgraded/deleted. The downgrade request will fail if the new maximum version level is a value that's not lower than the existing maximum finalized version level."}
-    ]}
+      {"name": "AllowDowngrade", "type": "bool", "versions": "0",


So if this is set to true and UpgradeType == 1 then we'll change UpgradeType to 2, right?

I don't think these can be set together. AllowDowngrade is just for version 0, and UpgradeType is version 1+

If you're talking about UpdateFeatureRequest, we just ignore UpgradeType when the message version is zero.

cmccabe · 2022-04-13T19:33:44Z

I think you forgot to include the QuorumFeatures file.

Do we have a test for sending a v0 UpdateFeaturesRequest, with and without AllowDowngrade set to true?

Doesn't have to be an integration test (unit test is fine) but I think we should have this...

cmccabe

LGTM once you address the comment about consolidating the nextProducerId, brokerId, brokerEpoch and adding a test for the old RPC

mumrah · 2022-04-14T01:32:38Z

@cmccabe Fixed a few integration tests and producer ID block thing in latest commit.

cmccabe · 2022-04-14T17:04:46Z

Thanks, @mumrah. Committed.

revert test commit test commit revert test commit MINOR: Supplement the description of `Valid Values` in the documentation of `compression.type` (apache#11985) Because a validator is added to ProducerConfig.COMPRESSION_TYPE_CONFIG and KafkaConfig.CompressionTypeProp, the corresponding testCase is improved to verify whether the wrong value of compression.type will throw a ConfigException. Reviewers: Mickael Maison <mickael.maison@gmail.com>, Guozhang Wang <wangguoz@gmail.com> MINOR: Correct Connect docs on connector/task states (apache#11914) The `DESTROYED` state is represented internally as a tombstone record when running in distributed mode and by the removal of the connector/task from the in-memory status map when running in standalone mode. As a result, it will never appear to users of the REST API, and we should remove mention of it from our docs so that developers creating tooling against the REST API don't write unnecessary logic to account for that state. Reviewers: Mickael Maison <mickael.maison@gmail.com> KAFKA-13828; Ensure reasons sent by the consumer are small (apache#12043) This PR reworks the reasons used in the ConsumerCoordinator to ensure that they remain reasonably short. Reviewers: Bruno Cadonna <bruno@confluent.io> KAFKA-13542: Add rebalance reason in Kafka Streams (apache#12018) Reviewers: Bruno Cadonna <bruno@confluent.io>, David Jacot <djacot@confluent.io> MINOR: Verify stopReplica if broker epoch not stale (apache#12040) Verify that ReplicaManager.stopReplica is called if the stop replica request doesn't result in a stale broker epoch error. Reviewers: Mickael Maison <mimaison@users.noreply.github.com> KAFKA-13651; Add audit logging to `StandardAuthorizer` (apache#12031) This patch adds audit support through the kafka.authorizer.logger logger to StandardAuthorizer. It follows the same conventions as AclAuthorizer with a similarly formatted log message. When logIfAllowed is set in the Action, then the log message is at DEBUG level; otherwise, we log at trace. When logIfDenied is set, then the log message is at INFO level; otherwise, we again log at TRACE. Reviewers: Colin P. McCabe <cmccabe@apache.org> KAFKA-13743: Prevent topics with conflicting metrics names from being created in KRaft mode apache#11910 In ZK mode, the topic "foo_bar" will conflict with "foo.bar" because of limitations in metric names. We should implement this in KRaft mode. This PR also changes TopicCommandIntegrationTest to support KRaft mode. Reviewers: Colin P. McCabe <cmccabe@apache.org> KAFKA-12613: Fix inconsistent validation logic between KafkaConfig and LogConfig (apache#10472) Reviewers: Mickael Maison <mickael.maison@gmail.com> KAFKA-13823 Feature flag changes from KIP-778 (apache#12036) This PR includes the changes to feature flags that were outlined in KIP-778. Specifically, it changes UpdateFeatures and FeatureLevelRecord to remove the maximum version level. It also adds dry-run to the RPC so the controller can actually attempt the upgrade (rather than the client). It introduces an upgrade type enum, which supersedes the allowDowngrade boolean. Because FeatureLevelRecord was unused previously, we do not need to introduce a new version. The kafka-features.sh tool was overhauled in KIP-778 and now includes the describe, upgrade, downgrade, and disable sub-commands. Refer to [KIP-778](https://cwiki.apache.org/confluence/display/KAFKA/KIP-778%3A+KRaft+Upgrades) for more details on the new command structure. Reviewers: Colin P. McCabe <cmccabe@apache.org>, dengziming <dengziming1993@gmail.com> MINOR: Update LICENSE-binary (apache#12051) Updates the license file. Reviewer: Bill Bejeck <bbejeck@apache.org> MINOR: Remove redundant conditional judgments in Selector.clear() (apache#12048) Condition 'sendFailed' is always 'false' when reached. Reviewers: Guozhang Wang <wangguoz@gmail.com> MINOR: ignore unused configuration when ConsumerCoordinator is not constructed (apache#12041) Following PR apache#11940, ignore unused config when ConsumerCoordinator is not constructed. Reviewers: Guozhang Wang <wangguoz@gmail.com> [MINOR] Update upgrade documentation for 3.2 (apache#12055) Reviewer: Bruno Cadonna <cadonna@apache.org> MINOR: Move some TopicCommand and ConfigCommand integration tests to unit tests (apache#12024) Move some TopicCommand and ConfigCommand integration tests to unit tests to speed up the tests Reviewers: Luke Chen <showuon@gmail.com> MINOR: Improve the description of principal under different mechanisms of sasl (apache#11947) Reviewers: Mickael Maison <mickael.maison@gmail.com> MINOR: update comment in LocalLog.replaceSegments() (apache#12054) Reviewers: Luke Chen <showuon@gmail.com> MINOR: Make link in quickstart dynamic (apache#12057) Reviewer: Matthias J. Sax <mjsax@apache.org> KAFKA-13769: Explicitly route FK join results to correct partitions (apache#11945) Prior to this commit FK response sink routed FK results to SubscriptionResolverJoinProcessorSupplier using the primary key. There are cases, where this behavior is incorrect. For example, if KTable key serde differs from the data source serde which might happen without a key changing operation. Instead of determining the resolver partition by serializing the PK this patch includes target partition in SubscriptionWrapper payloads. Default FK response-sink partitioner extracts the correct partition from the value and routes the message accordingly. Reviewers: Matthias J. Sax <matthias@confluent.io> KAFKA-13807: Fix incrementalAlterConfig and refactor some things (apache#12033) Ensure that we can set log.flush.interval.ms at the broker or cluster level via IncrementalAlterConfigs. This was broken by KAFKA-13749, which added log.flush.interval.ms as the second synonym rather than the first. Add a regression test to DynamicConfigChangeTest. Create ControllerRequestContext and pass it to every controller API. This gives us a uniform way to pass through information like the deadline (if there is one) and the Kafka principal which is making the request (in the future we will want to log this information). In ControllerApis, enforce a timeout for broker heartbeat requests which is equal to the heartbeat request interval, to avoid heartbeats piling up on the controller queue. This should have been done previously, but we overlooked it. Add a builder for ClusterControlManager and ReplicationControlManager to avoid the need to deal with a lot of churn (especially in test code) whenever a new constructor parameter gets added for one of these. In ControllerConfigurationValidator, create a separate function for when we just want to validate that a ConfigResource is a valid target for DescribeConfigs. Previously we had been re-using the validation code for IncrementalAlterConfigs, but this was messy. Split out the replica placement code into a separate package and reorganize it a bit. Reviewers: David Arthur <mumrah@gmail.com MINOR: Correct spelling errors in KafkaRaftClient (apache#12061) Correct spelling errors in KafkaRaftClient Reviewers: José Armando García Sancio <jsancio@users.noreply.github.com> MINOR: Fix TestDowngrade.test_upgrade_and_downgrade (apache#12027) The second validation does not verify the second bounce because the verified producer and the verified consumer are stopped in `self.run_validation`. This means that the second `run_validation` just spit out the same information as the first one. Instead, we should just run the validation at the end. Reviewers: Jason Gustafson <jason@confluent.io> KAFKA-8785: fix request timeout by waiting for metadata cache up-to-date (apache#11681) The reason why this test is flaky is because we have race condition at the beginning of the test, when brokers are staring up, and the adminClient is requesting for brokers metadata. Once the adminClient only got partial metadata, the test will fail, because in these tests, brokers will be shutdown to test leader election. Fix this issue by explicitly waiting for metadata cache up-to-date in waitForReadyBrokers, and let admin client get created after waitForReadyBrokers. Reviewers: Jason Gustafson <jason@confluent.io>, David Jacot <djacot@confluent.io>, dengziming <dengziming1993@gmail.com> KAFKA-13832: Fix flaky testAlterAssignment (apache#12060) In KRaft mode the metadata is not propagate in time, so we should should wait for it before make assertions. Reviewers: Luke Chen <showuon@gmail.com> KAFKA-13654: Extend KStream process with new Processor API (apache#11993) Updates the KStream process API to cover the use cases of both process and transform, and deprecate the KStream transform API. Implements KIP-820 Reviewer: John Roesler <vvcephei@apache.org> KAFKA-13835: Fix two bugs related to dynamic broker configs in KRaft (apache#12063) Fix two bugs related to dynamic broker configs in KRaft. The first bug is that we are calling reloadUpdatedFilesWithoutConfigChange when a topic configuration is changed, but not when a broker configuration is changed. This is backwards. This function must be called only for broker configs, and never for topic configs or cluster configs. The second bug is that there were several configurations such as max.connections which are related to broker listeners, but which do not involve changing the registered listeners. We should support these configurations in KRaft. This PR fixes the configuration change validation to support this case. Reviewers: Jason Gustafson <jason@confluent.io>, Matthew de Detrich <mdedetrich@gmail.com> KAFKA-10095: Add stricter assertion in LogCleanerManagerTest (apache#12004) Reviewers: Mickael Maison <mickael.maison@gmail.com> MINOR: Scala cleanups in core (apache#12058) Reviewers: Luke Chen <showuon@gmail.com>, Divij Vaidya <diviv@amazon.com> KAFKA-13799: Improve documentation for Kafka zero-copy (apache#12052) Improve documentation for Kafka zero-copy. Kafka combines pagecache and zero-copy to greatly improve message consumption efficiency. But zero-copy only works in PlaintextTransportLayer. Reviewers: Divij Vaidya <divijvaidya13@gmail.com>, Guozhang Wang <wangguoz@gmail.com> KAFKA-13785: [5/N][emit final] cache for time ordered window store (apache#12030) A new cache for RocksDBTimeOrderedWindowStore. Need this because RocksDBTimeOrderedWindowStore's key ordering is different from CachingWindowStore which has issues for MergedSortedCacheWindowStoreIterator Reviewers: Matthias J. Sax <mjsax@apache.org>, Guozhang Wang <wangguoz@gmail.com> KAFKA-13588: consolidate `changelogFor` methods to simplify the generation of internal topic names (apache#11703) Reviewers: Guozhang Wang <wangguoz@gmail.com> HOTFIX: fix broken trunk due to conflicting and overlapping commits (apache#12074) Reviewers: Victoria Xia <victoria.xia@confluent.io>, David Arthur <mumrah@gmail.com> MINOR: cleanup for postProcessAndValidateIdempotenceConfigs method (apache#12069) Reviewers: Luke Chen <showuon@gmail.com> KAFKA-13834: fix drain batch starving issue (apache#12066) In drainBatchesForOneNode method, there's possibility causing some partitions in a node will never get picked. Fix this issue by maintaining a drainIndex for each node. Reviewers: Luke Chen <showuon@gmail.com>, RivenSun <91005273+RivenSun2@users.noreply.github.com> MINOR: Improve Gradle Caching and Fix Deprecations (apache#12003) * Fix UP-TO-DATE check in `create*VersionFile` tasks `create*VersionFile` tasks explicitly declared output UP-TO-DATE status as being false. This change properly sets the inputs to `create*VersionFile` tasks to the `commitId` and `version` values and sets `receiptFile` locally rather than in an extra property. * Enable output caching for `process*Messages` tasks `process*Messages` tasks did not have output caching enabled. This change enables that caching, as well as setting a property name and RELATIVE path sensitivity. * Fix existing Gradle deprecations Replaces `JavaExec#main` with `JavaExec#mainClass` Replaces `Report#destination` with `Report#outputLocation` Adds a `generator` configuration to projects that need to resolve the `generator` project (rather than referencing the runtimeClasspath of the `generator` project from other project contexts. Reviewers: Mickael Maison <mickael.maison@gmail.com>  MINOR; Fix partition change record noop check (apache#12073) When LeaderRecoveryState was added to the PartitionChangeRecord, the check for being a noop was not updated. This commit fixes that and improves the associated test to avoid this oversight in the future. Reviewers: Colin Patrick McCabe <cmccabe@apache.org> MINOR: revert back to 60s session timeout for static membership test (apache#11881) Reviewers: Guozhang Wang <wangguoz@gmail.com> KAFKA-13841: Fix a case where we were unable to place on fenced brokers in KRaft mode (apache#12075) This PR fixes a case where we were unable to place on fenced brokers In KRaft mode. Specifically, if we had a broker registration in the metadata log, but no associated heartbeat, previously the HeartbeatManager would not track the fenced broker. This PR fixes this by adding this logic to the metadata log replay path in ClusterControlManager. Reviewers: David Arthur <mumrah@gmail.com>, dengziming <dengziming1993@gmail.com>

This test was removed in #11667 since UpdateFeatures is not properly handled in KRaft mode, now we can bring it back since UpdateFeatures is properly handled after #12036. Reviewers: Luke Chen <showuon@gmail.com>

Feature flag changes from KIP-778

38e4c18

* Remove finalized max version * Add support for UpdateFeatures on KRaft controller * Changes to UpdateFeatures RPC * Changes to FeatureLevelRecord

mumrah requested a review from cmccabe April 12, 2022 20:01