The thread no longer dies. When encountering an unexpected error, it marks the partition as "uncleanable" which means it will not try to clean its logs in subsequent runs.
Reviewers: Dhruvil Shah <dhruvil@confluent.io>, Jun Rao <junrao@gmail.com>
OAuthBearerLoginModule is used both on the server-side and client-side (similar to login modules for other mechanisms). OAUTHBEARER tokens are client credentials used only on the client-side to authenticate with servers, but the current implementation requires tokens to be provided on the server-side even if OAUTHBEARER is not used for inter-broker communication. This commit makes tokens optional for server-side login context to allow brokers to be configured without a token when OAUTHBEARER is not used for inter-broker communication.
Reviewers: Ron Dagostino <rndgstn@gmail.com>, Jun Rao <junrao@gmail.com>
This patch contains the broker-side support for the fencing improvements from KIP-320. This includes the leader epoch validation in the ListOffsets, OffsetsForLeaderEpoch, and Fetch APIs as well as the changes needed in the fetcher threads to maintain and use the current leader epoch. The client changes from KIP-320 will be left for a follow-up.
One notable change worth mentioning is that we now require the read lock in `Partition` in order to read from the log or to query offsets. This is necessary to ensure the safety of the leader epoch validation. Additionally, we forward all leader epoch changes to the replica fetcher thread and go through the truncation phase. This is needed to ensure the fetcher always has the latest epoch and to guarantee that we cannot miss needed truncation if we missed an epoch change.
Reviewers: Jun Rao <junrao@gmail.com>
The parent classloader of the DelegatingClassLoader and therefore the classloading scheme used by Connect does not have to be fixed to the System classloader.
Setting it the same as the one that was used to load the DelegatingClassLoader class itself is more flexible and, while in most cases will result in the System classloader to be used, it will also work in othr managed environments that control classloading differently (OSGi, and others).
The fix is minimal and the mainstream use is tested via system tests.
Reviewers: Randall Hauch <rhauch@gmail.com>, Jason Gustafson <jason@confluent.io>
This patch adds checks before reading the first record of a control batch. If the batch is empty, it is treated as having already been cleaned. In the case of LogCleaner this means it is safe to discard. In the case of ProducerStateManager it means it shouldn't cause state to be stored because the relevant transaction has already been cleaned. In the case of Fetcher, it just preempts the check for an abort. In the case of GroupMetadataManager, it doesn't process the offset as a commit. The patch also adds isControl to the output of DumpLogSegments. Changes were tested with unit tests, except the DumpLogSegments change which was tested manually.
In `ConsumerBuilder.build`, if `awaitInitialPositions` raises an exception, the consumer will not be closed properly. We should add the consumer instance to the `consumers` collection immediately after construction.
Reviewers: Guozhang Wang <wangguoz@gmail.com>
Pricing for m3.xlarge: On-Demand is at $0.266. Reserved is at about $0.16 (40% discount). And Spot is at $0.0627 (76% discount relative to On-Demand, or 60% discount relative to Reserved). Insignificant fluctuation in the past 3 months.
Ran on branch builder and works as expected -- each worker is created using spot instances (https://jenkins.confluent.io/job/system-test-kafka-branch-builder/1982/console)
This can be safely backported to 0.10.2 (tested using https://jenkins.confluent.io/job/system-test-kafka-branch-builder/1983/)
Author: Max Zheng <maxzheng.os@gmail.com>
Reviewers: Ewen Cheslack-Postava <ewen@confluent.io>
Closes#5707 from maxzheng/minor-switch@trunk
In recent PRs, we have been confused about the proper usage of
StatefulProcessorNode (#5731 , #5737 )
This change disambiguates it.
Reviewers: Matthias J. Sax <matthias@confluent.io>, Bill Bejeck <bill@confluent.io>, Guozhang Wang <guozhang@confluent.io>
Various converters (AvroConverter and JsonConverter) produce a
SchemaAndValue consisting of a logical schema type and a java.util.Date.
This is a fix for SchemaProjector to properly handle the Date.
Author: Robert Yokota <rayokota@gmail.com>
Reviewers: Konstantine Karantasis <konstantine@confluent.io>, Ewen Cheslack-Postava <ewen@confluent.io>
Closes#5736 from rayokota/KAFKA-7476
In unrelated recent work, I noticed some warnings about the missing type parameters on ProcessorParameters.
While investigating it, it seems like there was a bug in the creation of repartition topics.
Reviewers: Bill Bejeck <bill@confluent.io>, Guozhang Wang <guozhang@confluent.io>
This patch ensures that the leader epoch cache is updated when a broker becomes leader with the latest epoch and the log end offset as its starting offset. This guarantees that the leader will be able to provide the right truncation point even if the follower has data from leader epochs which the leader itself does not have. This situation can occur when there are back to back leader elections.
Additionally, we have made the following changes:
1. The leader epoch cache enforces monotonically increase epochs and starting offsets among its entry. Whenever a new entry is appended which violates requirement, we remove the conflicting entries from the cache.
2. Previously we returned an unknown epoch and offset if an epoch is queried which comes before the first entry in the cache. Now we return the smallest . For example, if the earliest entry in the cache is (epoch=5, startOffset=10), then a query for epoch 4 will return (epoch=4, endOffset=10). This ensures that followers (and consumers in KIP-320) can always determine where the correct starting point is for the active log range on the leader.
Reviewers: Jun Rao <junrao@gmail.com>
Reviewers: Johne Roesler <john@confluent.io>, Matthias J. Sax <matthias@confluent.io>, Bill Bejeck <bill@confluent.io>, Guozhang Wang <guozhang@confluent.io>
During the consumer group rebalance, when the joining group phase finishes, the heartbeat delayed operation of the consumer that fails to rejoin the group should be removed from the purgatory. Otherwise, even though the member ID of the consumer has been removed from the group, its heartbeat delayed operation is still registered in the purgatory and the heartbeat delayed operation is going to timeout and then another unnecessary rebalance is triggered because of it.
Author: Lincong Li <lcli@linkedin.com>
Reviewers: Dong Lin <lindong28@gmail.com>
Closes#5556 from Lincong/remove_heartbeat_delayedOperation
KIP-372 (allow naming all internal topics) was designed and developed concurrently with suppression.
Since suppression introduces a new internal topic, it also needs to be nameable.
Reviewers: Guozhang Wang <guozhang@confluent.io>, Matthias J. Sax <matthias@confluent.io>
This is Part 4 of suppression (durability)
Part 1 was #5567 (the API)
Part 2 was #5687 (the tests)
Part 3 was #5693 (in-memory buffering)
Implement a changelog for the suppression buffer so that the buffer state may be recovered on restart or recovery.
As of this PR, suppression is suitable for general usage.
Reviewers: Bill Bejeck <bill@confluent.io>, Guozhang Wang <guozhang@confluent.io>, Matthias J. Sax <matthias@confluent.io>
In some tests, the check monitoring the JMX tool log output doesn’t quite wait long enough before failing. Increasing the timeout from 10 to 20 seconds.
`ConsumerCoordinator.onJoinPrepare()` currently makes multiple copies of the set of assigned partitions. We can let `subscriptions.assignedPartitions()` return a view of the underlying partition set, copy it only once and re-use the copied value.
Author: radai-rosenblatt <radai.rosenblatt@gmail.com>
Reviewers: Ismael Juma <ismael@juma.me.uk>, Dong Lin <lindong28@gmail.com>
Closes#5124 from radai-rosenblatt/copy-all-the-things
Allow to cast LogicalType to string by calling the serialized (Java) object's toString().
Added tests for `BigDecimal` and `Date` as whole record and as fields.
Author: Amit Sela <amitsela33@gmail.com>
Reviewers: Randall Hauch <rhauch@gmail.com>, Robert Yokota <rayokota@gmail.com>, Ewen Cheslack-Postava <ewen@confluent.io>
Closes#4820 from amitsela/cast-transform-bytes
Switches to normal year format instead of week date years and day of month instead of day of year.
This is directly from #4820, but separated into a different JIRA/PR to keep the fixes independent. Original authorship should be maintained in the commit.
Author: Amit Sela <amitsela33@gmail.com>
Reviewers: Ewen Cheslack-Postava <ewen@confluent.io>
Closes#5718 from ewencp/fix-header-converter-date-format
As part of KAFKA-6514, the `apiVersion` tag was added to the `RequestsPerSec`
metric. A thread unsafe `HashMap` was used in the implementation even though
it can be accessed by multiple threads. Fix it by replacing it with the thread-safe
`Pool`.
Reviewers: Ismael Juma <ismael@juma.me.uk>
The Suppression buffer stores the full record context, not just the key and value,
so its changelog/restore loop will also need to preserve this information.
This change is a precondition to that, creating an option to register a
state restore callback to receive the full consumer record.
Reviewers: Bill Bejeck <bill@confluent.io>, Matthias J. Sax <matthias@confluent.io>, Guozhang Wang <wangguoz@gmail.com>
*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: Michał Borowiecki <mbor81@gmail.com>
Reviewers: Arjun Satish <arjun@confluent.io>, Ewen Cheslack-Postava <ewen@confluent.io>
Closes#5700 from mihbor/KAFKA-7434
Lazy allocation helps when there are a large number of connections
that have been accepted, but where no data has been received from
the clients. Each buffer is often around 16k (max TLS record size).
Nulling the buffers should not make a difference in the current
implementation since we release the reference to the channel
and transport layer after we close them, but it's a good practice
to release medium/large buffers after `close` is called.
Reviewers: Jun Rao <junrao@gmail.com>, Ismael Juma <ismael@juma.me.uk>
Values for `message.format.version` and `log.message.format.version` should be verified before topic creation or config change.
Reviewers: Manikumar Reddy <manikumar.reddy@gmail.com>, Ismael Juma <ismael@juma.me.uk>, Jason Gustafson <jason@confluent.io>
This commit implements the changes described in KIP-320 for the persistence of leader epoch information in the offset commit protocol.
Reviewers: Rajini Sivaram <rajinisivaram@googlemail.com>
This is Part 2 of suppression.
Part 1 was #5567
In an effort to control the scope of the review, this PR is just the tests for buffered suppression.
Reviewers: Bill Bejeck <bill@confluent.io>, Guozhang Wang <wangguoz@gmail.com>
This fixes a regression caused by KAFKA-4682 (KIP-211) which caused offset commit failures after upgrading from an older version which used the v1 inter-broker format.
Stanislav Kozlovski
Rajini Sivaram
matus-cuper
John Roesler
Jason Gustafson
Konstantine Karantasis
Bob Barrett
Colin Patrick McCabe
Max Zheng
Robert Yokota
Dong Lin
Xiongqi Wesley Wu
Nikolay
Lincong Li
Bill Bejeck
Guozhang Wang
Kevin Lu
Randall Hauch
radai-rosenblatt
Amit Sela
Zhanxiang (Patrick) Huang
Ismael Juma
Michał Borowiecki
Lee Dongjin
huxi
Attila Sasvari
Vahid Hashemian
Manikumar Reddy