Revert the decision for the sendOffsetsToTransaction(groupMetadata) API to fail with old version of brokers for the sake of making the application easier to adapt between versions. This PR silently downgrade the TxnOffsetCommit API when the build version is small than 3.
Reviewers: Matthias J. Sax <matthias@confluent.io>, Guozhang Wang <wangguoz@gmail.com>
Startup scripts for the early version of Kafka contain removed JVM options like `-XX:+PrintGCDateStamps` or `-XX:UseParNewGC`.
When system tests run on JVM that doesn't support these options we should set up
environment variables with correct options.
Reviewers: Guozhang Wang <guozhang@confluent.io>, Ron Dagostino <rdagostino@confluent.io>, Ismael Juma <ismael@juma.me.uk
* Replace Prev/Standby task lists with a representation of the current poasition
of all tasks, where each task is encoded as the sum of the positions of all the
changelogs in that task.
* Only the protocol change is implemented, not actual positions, and the
assignor is updated to translate the new protocol back to lists of Prev/Standby
tasks so that the current assignment protocol still functions without modification.
Implements: KIP-441
Reviewers: John Roesler <vvcephei@apache.org>, Bruno Cadonna <bruno@confluent.io>
As described in KIP-568.
Waiting on acceptance of the KIP to write the tests, on the off chance something changes. But rest assured unit tests are coming ⚡️
Will also kick off existing Streams system tests which leverage this new API (eg version probing, sometimes broker bounce)
Reviewers: Boyang Chen <boyang@confluent.io>, Guozhang Wang <wangguoz@gmail.com>
what/why
the throttling_test was broken by this PR (#7785) since it depends on the consumer having partitions-assigned before starting the producer
this PR provides the ability to wait for partitions to be assigned in the console consumer before considering it started.
caveat
this does not support starting up the JmxTool inside the console-consumer for custom metrics while using this wait_until_partitions_assigned flag since the code assumes one JmxTool running per node.
I think a proper fix for this would be to make JmxTool its own standalone single-node service
alternatives
we could use the EndToEnd test suite which uses the verifiable producer/consumer under the hood but I found that there were more changes necessary to get this working unfortunately (specifically doesn't seem like this test suite plays nicely with the ProducerPerformanceService)
Reviewers: Mathew Wong <mwong@confluent.io>, Bill Bejeck <bbejeck.com>
The test_throttled_reassignment test fails because the consumer that is used to validate reassignment does not start on time to consume all messages. This does not seem like an issue with the throttling of the reassignment, since increasing the timeout allowed the test to pass multiple consecutive runs locally.
This test seemed to rely on the default JmxTool for the console consumer that was removed in this commit: 179d0d7
The console consumer would check to see if it had partitions assigned to it before beginning to consume. Although the test occasionally failed with the JmxTool, it began to fail much more after the removal.
Error messages of failures followed the below format with varying numbers of missed messages. They are the first messages by the producer.
535 acked message did not make it to the Consumer. They are: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19...plus 515 more. Total Acked: 192792, Total Consumed: 192259. We validated that the first 535 of these missing messages correctly made it into Kafka's data files. This suggests they were lost on their way to the consumer.
In the scope of the test, this error suggests that the test is falling into the race condition described in produce_consume_validate.py, which has the timeout to prevent the consumer from missing initial messages.
This can serve as a temporary fix until the logic of consumer startup is addressed further.
Reviewers: Jason Gustafson <jason@confluent.io>, Bill Bejeck <bbejeck@gmail.com>
Newer versions of Java have added checks to ensure that trust anchors are CA certificates and contain proper extensions. This PR adds Basic Constraints extension with the CA field set to true for system tests.
Reviewers: ajini Sivaram <rajinisivaram@googlemail.com>
This change mainly have 2 components:
1. extend the existing transactions_test.py to also try out new sendTxnOffsets(groupMetadata) API to make sure we are not introducing any regression or compatibility issue
a. We shrink the time window to 10 seconds for the txn timeout scheduler on broker so that we could trigger expiration earlier than later
2. create a completely new system test class called group_mode_transactions_test which is more complicated than the existing system test, as we are taking rebalance into consideration and using multiple partitions instead of one. For further breakdown:
a. The message count was done on partition level, instead of global as we need to visualize
the per partition order throughout the test. For this sake, we extend ConsoleConsumer to print out the data partition as well to help message copier interpret the per partition data.
b. The progress count includes the time for completing the pending txn offset expiration
c. More visibility and feature improvements on TransactionMessageCopier to better work under either standalone or group mode.
Reviewers: Matthias J. Sax <matthias@confluent.io>, Guozhang Wang <wangguoz@gmail.com>
This PR is collaborated by Guozhang Wang and John Roesler. It is a significant tech debt cleanup on task management and state management, and is broken down by several sub-tasks listed below:
Extract embedded clients (producer and consumer) into RecordCollector from StreamTask.
guozhangwang#2
guozhangwang#5
Consolidate the standby updating and active restoring logic into ChangelogReader and extract out of StreamThread.
guozhangwang#3
guozhangwang#4
Introduce Task state life cycle (created, restoring, running, suspended, closing), and refactor the task operations based on the current state.
guozhangwang#6
guozhangwang#7
Consolidate AssignedTasks into TaskManager and simplify the logic of changelog management and task management (since they are already moved in step 2) and 3)).
guozhangwang#8
guozhangwang#9
Also simplified the StreamThread logic a bit as the embedded clients / changelog restoration logic has been moved into step 1) and 2).
guozhangwang#10
Reviewers: A. Sophie Blee-Goldman <sophie@confluent.io>, Bruno Cadonna <bruno@confluent.io>, Boyang Chen <boyang@confluent.io>
Deprecate existing metadata query APIs in favor of new
ones that include standby hosts as well as partition
information.
Closes: #7960
Implements: KIP-535
Co-authored-by: Navinder Pal Singh Brar <navinder_brar@yahoo.com>
Reviewed-by: John Roesler <vvcephei@apache.org>
Do not initialize `JmxTool` by default when running console consumer. In order to support this, we remove `has_partitions_assigned` and its only usage in an assertion inside `ProduceConsumeValidateTest`, which did not seem to contribute much to the validation.
Reviewers: David Arthur <mumrah@gmail.com>, Jason Gustafson <jason@confluent.io>
In some system tests a Streams app is started and then prints a message to stdout, which the system test waits for to confirm the node has successfully been brought up. It then greps for certain log messages in a retriable loop.
But waiting on the Streams app to start/print to stdout does not mean the log file has been created yet, so the grep may return an error. Although this occurs in a retriable loop it is assumed that grep will not fail, and the result is piped to wc and then blindly converted to an int in the python function, which fails since the error message is a string (throws ValueError)
We should catch the ValueError and return a 0 so it can try again rather than immediately crash
Reviewers: Bill Bejeck <bbejeck@gmail.com>, John Roesler <vvcephei@users.noreply.github.com>, Guozhang Wang <wangguoz@gmail.com>
The upgrade system test correctly rolls by upgrading the broker and
leaving the IBP, and then rolling again with the latest IBP version.
Unfortunately, this is not sufficient to pick up many problems in our IBP
gating as we charge through the rolls and after the second roll all of
the brokers will rejoin the ISR and the test will be treated as a
success.
This test adds two new checks:
1. We wait for the ISR to stabilize for all partitions. This is best
practice during rolls, and is enough to tell us if a broker hasn't
rejoined after each roll.
2. We check the broker logs for some common protocol errors. This is a
fail safe as it's possible for the test to be successful even if some
protocols are incompatible and the ISR is rejoined.
Reviewers: Nikhil Bhatia <nikhil@confluent.io>, Jason Gustafson <jason@confluent.io>
The --enable-autocommit argument is a flag. It does not take a parameter. This was broken in #7724.
Reviewers: Ismael Juma <ismael@juma.me.uk>, Manikumar Reddy <manikumar.reddy@gmail.com>
Two tests using 50k replicas on 8 brokers:
* Do a rolling restart with clean shutdown, delete topics
* Run produce bench and consumer bench on a subset of topics
Reviewed-By: David Jacot <djacot@confluent.io>, Vikas Singh <vikas@confluent.io>, Jason Gustafson <jason@confluent.io>
This patch adds a basic downgrade system test. It verifies that producing and consuming continues to work before and after the downgrade.
Reviewers: Ismael Juma <ismael@juma.me.uk>, David Arthur <mumrah@gmail.com>
Recently, system tests test_rebalance_[simple|complex] failed
repeatedly with a verfication error. The cause was most probably
the missing clean-up of a state directory of one of the processors.
A node is cleaned up when a service on that node is started and when
a test is torn down.
If the clean-up flag clean_node_enabled of a EOS Streams service is
unset, the clean-up of the node is skipped.
The clean-up flag of processor1 in the EOS tests should stay set before
its first start, so that the node is cleaned before the service is started.
Afterwards for the multiple restarts of processor1 the cleans-up flag should
be unset to re-use the local state.
After the multiple restarts are done, the clean-up flag of processor1 should
again be set to trigger node clean-up during the test teardown.
A dirty node can lead to test failures when tests from Streams EOS tests are
scheduled on the same node, because the state store would not start empty
since it reads the local state that was not cleaned up.
Reviewers: Matthias J. Sax <mjsax@apache.org>, Andrew Choi <andchoi@linkedin.com>, Bill Bejeck <bbejeck@gmail.com>
* Add rate limiting to tc
* Feedback from PR
* Add a sanity test for tc
* Add iperf to vagrant scripts
* Dynamically determine the network interface
* Add some temp code for testing on AWS
* Temp: use hostname instead of external IP
* Temp: more AWS debugging
* More AWS WIP
* More AWS temp
* Lower latency some
* AWS wip
* Trying this again now that ping should work
* Add cluster decorator to tests
* Fix broken import
* Fix device name
* Fix decorator arg
* Remove errant import
* Increase timeouts
* Fix tbf command, relax assertion on latency test
* Fix log line
* Final bit of cleanup
* Newline
* Revert Trogdor retry count
* PR feedback
* More PR feedback
* Feedback from PR
* Remove unused argument
When Trogdor wants to clear all the faults injected to Kibosh, it sends the empty JSON object {}. However, Kibosh expects {"faults":[]} instead. Kibosh should handle the empty JSON object, since that's consistent with how Trogdor handles empty JSON fields in general (if they're empty, they can be omitted). We should also have a test for this.
Reviewers: David Arthur <mumrah@gmail.com>, Stanislav Kozlovski <stanislav_kozlovski@outlook.com>
Instead of caching the checkpoint map during StandbyTask
initialization, use the latest checkpoints (which would have
been updated during suspend).
Reviewers: Bill Bejeck <bill@confluent.io>
Author: Colin P. Mccabe <cmccabe@confluent.io>
Reviewers: Vikas Singh <vikas@confluent.io>, Jason Gustafson <jason@confluent.io>
Closes#7477 from cmccabe/KAFKA-8984
The consumer's `committed` API does not return an entry in the response map for a requested partition if there is no committed offset. The transactional message copier, which is used in the transaction system test, did not account for this. If the first transaction attempted by the copier was randomly aborted, then we would not seek to the beginning as expected, which means we would fail to copy some of the records.
This patch fixes the problem by iterating over the assignment rather than the result of `committed` when resetting offsets. It also adds enables additional logging in the transaction message copier service to make finding problems easier in the future.
Author: Jason Gustafson <jason@confluent.io>
Reviewers: Manikumar Reddy <manikumar.reddy@gmail.com>
Closes#7653 from hachikuji/fix-transaction-system-test
With KIP-444 the metrics definitions are refactored. Thus, Streams' SimpleBenchmark needs to be updated to correctly access the refactored metrics.
Reviewers: Guozhang Wang <wangguoz@gmail.com>, Matthias J. Sax <mjsax@apache.org>, Bill Bejeck <bbejeck@gmail.com>
Inside onLeavePrepare we would look into the assignment and try to revoke the owned tasks and notify users via RebalanceListener#onPartitionsRevoked, and then clear the assignment.
However, the subscription's assignment is already cleared in this.subscriptions.unsubscribe(); which means user's rebalance listener would never be triggered. In other words, from consumer client's pov nothing is owned after unsubscribe, but from the user caller's pov the partitions are not revoked yet. For callers like Kafka Streams which rely on the rebalance listener to maintain their internal state, this leads to inconsistent state management and failure cases.
Before KIP-429 this issue is hidden away since every time the consumer re-joins the group later, it would still revoke everything anyways regardless of the passed-in parameters of the rebalance listener; with KIP-429 this is easier to reproduce now.
Our fixes are following:
• Inside unsubscribe, first do onLeavePrepare / maybeLeaveGroup and then subscription.unsubscribe. This we we are guaranteed that the streams' tasks are all closed as revoked by then.
• [Optimization] If the generation is reset due to fatal error from join / hb response etc, then we know that all partitions are lost, and we should not trigger onPartitionRevoked, but instead just onPartitionsLost inside onLeavePrepare. This is because we don't want to commit for lost tracks during rebalance which is doomed to fail as we don't have any generation info.
Reviewers: Matthias J. Sax <matthias@confluent.io>, A. Sophie Blee-Goldman <sophie@confluent.io>, Bill Bejeck <bill@confluent.io>, Guozhang Wang <guozhang@confluent.io>
MM2 added a few connector classes in Connect's classpath and given that the assertion in the Connect REST system tests need to be adjusted to account for these additions.
This fix makes sure that the loaded Connect plugins are a superset of the expected by the test connectors.
Testing: The change is straightforward. The fix was tested with local system test runs.
Author: Konstantine Karantasis <konstantine@confluent.io>
Reviewers: Manikumar Reddy <manikumar.reddy@gmail.com>
Closes#7578 from kkonstantine/minor-fix-connect-test-after-mm2-classes
This system test was marked @Ignore around a year and a half ago pending the version probing work, but never turned on again.
These days, it is made redundant by the suite of system tests in streams_upgrade_test, which cover rolling upgrades (including version probing and metadata change).
Reviewers: Boyang Chen <boyang@confluent.io>, Bill Bejeck <bbejeck@gmail.com>
All four flavors of the repartition/optimization tests have been reported as flaky and failed in one place or another:
* RepartitionOptimizingIntegrationTest.shouldSendCorrectRecords_OPTIMIZED
* RepartitionOptimizingIntegrationTest.shouldSendCorrectRecords_NO_OPTIMIZATION
* RepartitionWithMergeOptimizingIntegrationTest.shouldSendCorrectRecords_OPTIMIZED
* RepartitionWithMergeOptimizingIntegrationTest.shouldSendCorrectRecords_NO_OPTIMIZATION
They're pretty similar so it makes sense to knock them all out at once. This PR does three things:
* Switch to in-memory stores wherever possible
* Name all operators and update the Topology accordingly (not really a flaky test fix, but had to update the topology names anyway because of the IM stores so figured might as well)
* Port to TopologyTestDriver -- this is the "real" fix, should make a big difference as these repartition tests required multiple roundtrips with the Kafka cluster (while using only the default timeout)
Reviewers: Bill Bejeck <bill@confluent.io>, Guozhang Wang <wangguoz@gmail.com>
Key improvements with this PR:
* tasks will remain available for IQ during a rebalance (but not during restore)
* continue restoring and processing standby tasks during a rebalance
* continue processing active tasks during rebalance until the RecordQueue is empty*
* only revoked tasks must suspended/closed
* StreamsPartitionAssignor tries to return tasks to their previous consumers within a client
* but do not try to commit, for now (pending KAFKA-7312)
Reviewers: John Roesler <john@confluent.io>, Boyang Chen <boyang@confluent.io>, Guozhang Wang <wangguoz@gmail.com>
Boyang Chen
Ron Dagostino
Michal T
Matthias J. Sax
Bill Bejeck
Nikolay
A. Sophie Blee-Goldman
Rajini Sivaram
Brian Bushree
Matthew Wong
Guozhang Wang
David Arthur
vinoth chandar
Lucas Bradstreet
Bruno Cadonna
Manikumar Reddy
John Roesler
Randall Hauch
Jason Gustafson
Colin Patrick McCabe
Colin P. Mccabe
Konstantine Karantasis
Bill Bejeck