a399948 | hequn8128 | 02 December 2019, 13:57:00 UTC | Commit for release 1.8.3 | 02 December 2019, 13:57:00 UTC |
c58a2f5 | Chesnay Schepler | 29 November 2019, 15:50:05 UTC | [FLINK-14995][kinesis][legal] Update NOTICE | 01 December 2019, 16:47:31 UTC |
8747e3d | Mads Chr. Olesen | 27 November 2019, 13:56:16 UTC | [FLINK-14976][cassandra] Release semaphore on all Throwable's in send() | 29 November 2019, 18:50:25 UTC |
aa92ec5 | Stefan Richter | 28 November 2019, 16:21:52 UTC | [FLINK-10377] Support checkpoint overtaking a savepoint in TwoPhaseCommitSink The precondition checkState(pendingTransactionIterator.hasNext(), "checkpoint completed, but no transaction pending"); in TwoPhaseCommitSinkFunction.notifyCheckpointComplete() seems too strict, because checkpoints can overtake checkpoints and will fail the precondition. In this case the commit was already performed by the first notification and subsumes the late checkpoint. I think the check can be removed. This can happen in the following scenario: # savepoint is triggered # checkpoint is triggered # checkpoint completes (but it doesn't subsume the savepoint, because checkpoints subsume only other checkpoints). # savepoint completes | 28 November 2019, 16:23:35 UTC |
a77d6fb | chunpinghe | 27 November 2019, 11:01:38 UTC | [FLINK-11835][tests] Wait until job was recovered before unblocking task | 27 November 2019, 11:04:04 UTC |
b7e532a | Chesnay Schepler | 14 November 2019, 13:51:19 UTC | [FLINK-13995][legal] Properly exclude netty license directory | 26 November 2019, 10:48:17 UTC |
b4c387b | Chesnay Schepler | 24 November 2019, 17:21:12 UTC | [FLINK-14104][build] Use flink-shaded-jackson 9.0 | 25 November 2019, 12:26:53 UTC |
f5200c2 | Chesnay Schepler | 20 November 2019, 10:40:27 UTC | bump shade plugin version in dependencies/quickstarts | 21 November 2019, 11:02:42 UTC |
fef8a97 | Chesnay Schepler | 20 November 2019, 10:39:55 UTC | Add building instructions | 21 November 2019, 11:02:42 UTC |
95aca66 | Chesnay Schepler | 13 November 2019, 10:40:04 UTC | [FLINK-14104][build] Add opt-in jackson 2.10.1 profile | 21 November 2019, 11:02:42 UTC |
6e099b5 | Chesnay Schepler | 14 November 2019, 11:06:39 UTC | [FLINK-14104][build] Bump shade-plugin version to 3.1.1 Earlier versions fail with jackson 2.10.1, likely due to java9+ features present in the jar. | 21 November 2019, 11:02:42 UTC |
15ca8f0 | danrtsey.wy | 10 November 2019, 07:30:45 UTC | [FLINK-13184][yarn] Use dynamic properties instead of uploading taskmanager-conf.yaml to hdfs. This will reduce the time cost of launching a TaskExecutor so that YarnResourceManager could start a large number of TaskExecutors timely. | 20 November 2019, 06:15:32 UTC |
46e22ce | danrtsey.wy | 10 November 2019, 07:19:32 UTC | [FLINK-13184][yarn] Use NMClientAsync instead of NMClient to avoid starting TaskExecutor blocking call. The start container requests will be executed in a thread pool of NMClientAsync. | 20 November 2019, 06:12:09 UTC |
29c6454 | Paul Lam | 18 November 2019, 09:31:15 UTC | [FLINK-13747][client] Add integration tests for classloading in program execution | 19 November 2019, 09:09:40 UTC |
540458e | Paul Lam | 18 November 2019, 09:30:27 UTC | [FLINK-13747] Make client respect classloading policy | 19 November 2019, 09:09:40 UTC |
0b43d8d | Nico Kruber | 19 August 2019, 21:48:57 UTC | [FLINK-13791][docs] Speed up sidenav by using group_by _includes/sidenav.html parses through pages_by_language over and over again trying to find children when building the (recursive) side navigation. By doing this once with a group_by, we can gain considerable savings in building the docs via `./build_docs.sh` without any change to the generated HTML pages: This closes #9487 | 14 November 2019, 13:52:08 UTC |
adbf065 | Nico Kruber | 14 August 2019, 21:05:00 UTC | [FLINK-13726][docs] Build docs with jekyll 4.0.0.pre.beta1 This significantly reduces the build times, on my machine from 140s to 47s! This closes #9444 | 14 November 2019, 13:50:21 UTC |
369ac37 | Nico Kruber | 14 August 2019, 21:00:09 UTC | [hotfix][docs] Temporarily disable liveserve ./build_docs.sh -i previously did not only enable incremental documentation building while serving the docs, it also enabled a 'liveserve' mode that automatically reloaded pages in the browser when they changed. This is based on the 'hawkins' module which is not (yet) compatible with jekyll 4.0 which we need to (significantly) improve build times. This disables the liveserve mode and remove the hawkins module until a new version is available. | 14 November 2019, 13:49:36 UTC |
ff954b5 | Nico Kruber | 14 August 2019, 15:29:43 UTC | [FLINK-13725][docs] use sassc for faster doc generation Jekyll requires sass but can optionally also use a C-based implementation provided by sassc. Although we do not use sass directly, there may be some indirect use inside jekyll. It doesn't seem to hurt to upgrade here. This closes #9443 | 14 November 2019, 13:49:36 UTC |
70640a8 | Nico Kruber | 14 August 2019, 14:57:14 UTC | [FLINK-13729][docs] Update website generation dependencies This seems to come with a much nicer code highlighting. This closes #9442 | 14 November 2019, 13:49:27 UTC |
d9b0c4b | Nico Kruber | 14 August 2019, 13:20:28 UTC | [FLINK-13723][docs] Use liquid-c for faster doc generation Jekyll requires liquid and only optionally uses liquid-c if available. The latter uses natively-compiled code and reduces generation time by ~5% for me. This closes #9441 | 14 November 2019, 13:47:19 UTC |
6d55aba | Nico Kruber | 14 August 2019, 14:18:06 UTC | [FLINK-13724][docs] Remove unnecessary whitespace from the generated pages Starting command tags with "{%-" will drop all whitespace to the left and ending with "-%}" will drop all whitespace to the right (including newlines!). Code like the following would otherwise create quite some unnecessary whitespace: {% if parent_id %} {% assign parent_id = current[0].nav-parent_id %} {% else %} {% break %} {% endif %} This closes #9440 | 14 November 2019, 13:45:42 UTC |
bd6b2e2 | Nico Kruber | 14 August 2019, 13:59:50 UTC | [FLINK-13728][docs] Fix wrong closing tag order in sidenav This closes #9439 [ci skip] | 14 November 2019, 13:33:03 UTC |
48f2d63 | Nico Kruber | 12 November 2019, 14:09:52 UTC | [hotfix][docs] clarify that a RocksDB dependency in pom.xml may not be needed | 13 November 2019, 14:18:06 UTC |
db552a1 | Nico Kruber | 12 November 2019, 14:07:20 UTC | [hotfix][docs] clarify RocksDB thread options applicability per operator/TM | 13 November 2019, 14:18:06 UTC |
ba94f84 | Till Rohrmann | 08 November 2019, 09:49:13 UTC | [hotfix] Fix checkstyle violations in KeyGroupRangeAssignment | 08 November 2019, 09:56:47 UTC |
15031fc | Jiayi | 07 November 2019, 16:54:56 UTC | [FLINK-14646] Add non-null checks to KeyGroupRangeAssignment This closes #10120. | 08 November 2019, 09:55:54 UTC |
80b54bf | Chesnay Schepler | 07 November 2019, 09:48:16 UTC | [FLINK-14639][metrics][docs] Fix methods signature | 07 November 2019, 09:50:14 UTC |
d252469 | Hwanju Kim | 05 November 2019, 07:13:03 UTC | [FLINK-14589] Redundant slot requests with the same AllocationID leads to inconsistent slot table When a slot request is redundantly made with the same AllocationID to a slot index other than the already allocated one, slot table becomes inconsistent having two slot indices allocated but one AllocationID assigned to only the latest slot index. This can lead to slot leakage. This patch prevents such redundent slot request from rendering inconsistent slot allocation state by rejecting the request. This closes #10099. | 06 November 2019, 22:55:49 UTC |
cb31954 | Till Rohrmann | 05 November 2019, 14:30:13 UTC | [FLINK-12342][yarn] Remove container requests in order to reduce excess containers This commit changes the order in which the container requests are removed when onContainersAllocated is being called. The idea is to remove the container requests as fast as possible in order to avoid allocating excess containers as described in YARN-1902. This closes #10109. | 06 November 2019, 22:52:48 UTC |
9296ba5 | Nicolas Deslandes | 29 October 2019, 14:41:59 UTC | [FLINK-14562] Let RabbitMQ source close consumer and channel on close Closing method of RabbitMQ source must close consumer and channel in order to prevent leaving idle consumer This closes #10036. | 03 November 2019, 20:12:58 UTC |
e1e7226 | Jark Wu | 30 October 2019, 06:52:49 UTC | [hotfix][FLINK-14398][table-planner] Remove parentheses for Types.INT in SqlITCase | 30 October 2019, 06:53:38 UTC |
f6b855d | Hao Dang | 30 October 2019, 03:14:24 UTC | [FLINK-14398][table-planner] Further split input unboxing code into separate methods (#10000) | 30 October 2019, 06:04:26 UTC |
443ffae | Jiangjie (Becket) Qin | 26 October 2019, 10:55:44 UTC | [hotfix][kafka][test-stability] Accelerate the KafkaProducerTest by reducing the timeout values. | 29 October 2019, 02:24:44 UTC |
cf7509b | Jiangjie (Becket) Qin | 26 October 2019, 10:56:59 UTC | [FLINK-14370][kafka][test-stability] Fix the cascading test failure in KafkaProducerTestBase. | 29 October 2019, 02:24:35 UTC |
e0387a8 | Piotr Nowojski | 18 October 2019, 14:38:26 UTC | [FLINK-14235][kafka,tests] Change source in at-least-once test from finite to infinite Previously it was possible that the source would end before a first chcekpoint could complete. If that was the case, any exceptions thrown during checkpointing are swallowed, which could explain the apparent data loss from FLINK-14235. | 28 October 2019, 09:04:34 UTC |
148dedd | Abhilasha Seth | 09 October 2019, 23:22:10 UTC | Upgrade KPL version to 0.13.1 This commit mitigates the issue - https://github.com/awslabs/amazon-kinesis-producer/issues/224 This closes #14175 | 24 October 2019, 18:15:45 UTC |
9add2b1 | tison | 18 October 2019, 18:03:02 UTC | [FLINK-14434][coordination] Dispatcher#createJobManagerRunner returns on creation succeed This closes #9940. | 21 October 2019, 14:06:47 UTC |
c9002d2 | Liupengcheng | 21 October 2019, 05:47:48 UTC | [FLINK-14123][docs] Update release notes to contain warning about OOM This closes #9949. | 21 October 2019, 09:38:19 UTC |
31622bb | Timo Walther | 17 October 2019, 15:16:57 UTC | [FLINK-12848][table] Fix invalid row type caching This closes #9930. | 18 October 2019, 16:05:30 UTC |
be477b2 | Chesnay Schepler | 09 October 2019, 08:21:28 UTC | [FLINK-14008][release] Remove redundant NOTICE-binary infrastructure | 17 October 2019, 13:08:40 UTC |
8997f85 | Chesnay Schepler | 09 October 2019, 08:20:57 UTC | [FLINK-14008][release] Generate binary licensing during release | 17 October 2019, 13:00:13 UTC |
4080da6 | Chesnay Schepler | 09 October 2019, 08:18:48 UTC | [hotfix][release] Use release version in path | 17 October 2019, 13:00:09 UTC |
1f18ab7 | Chesnay Schepler | 17 October 2019, 11:04:29 UTC | [FLINK-14413][build] Specify encoding for ApacheNoticeResourceTransformer | 17 October 2019, 12:57:12 UTC |
315c1cd | Chesnay Schepler | 16 October 2019, 13:23:57 UTC | [FLINK-13991][git] Exclude files from 1.9+ features Adds artifacts from various 1.9+ features to .gitignore to ease switching branches. Does unfortunately not cover compiled files from the new WebUI since they reside in the same directory as the old one. | 16 October 2019, 13:23:57 UTC |
4c4dc3d | Chesnay Schepler | 08 October 2019, 09:27:34 UTC | [FLINK-14337][hs] Only mark archives as processed on success | 16 October 2019, 13:12:34 UTC |
f44f598 | Chesnay Schepler | 08 October 2019, 08:48:39 UTC | [FLINK-14337][hs] Prevent NPE on corrupt archives | 16 October 2019, 13:12:30 UTC |
5ec9627 | Chesnay Schepler | 08 October 2019, 09:32:59 UTC | [hotfix][hs] Clarify write access to webJobDir | 16 October 2019, 13:12:24 UTC |
5501858 | Chesnay Schepler | 08 October 2019, 09:26:55 UTC | [hotfix][hs] Deduplicate variables | 16 October 2019, 13:12:17 UTC |
d5285d5 | hpeter | 12 October 2019, 23:45:34 UTC | [FLINK-14215][docs] Add how to configure environment variables to documentation This closes #9887. | 15 October 2019, 08:47:28 UTC |
dc90f43 | Dian Fu | 14 October 2019, 07:54:48 UTC | [FLINK-13999][cep][docs] Correct the example in the section of Aggregations of MATCH_RECOGNIZE | 14 October 2019, 14:36:30 UTC |
903ac21 | tison | 11 October 2019, 02:10:43 UTC | [FLINK-14347][test] Filter out expected exception string in YARN tests This closes #9880. | 11 October 2019, 16:34:23 UTC |
0281b73 | Chesnay Schepler | 07 October 2019, 08:40:58 UTC | [FLINK-14334][es][docs] Use ExceptionUtils#findThrowable This closes #9849. | 10 October 2019, 16:34:35 UTC |
766e250 | Till Rohrmann | 04 October 2019, 11:56:44 UTC | [FLINK-14315] Make heartbeat manager fields non-nullable This commit introduces the NoOpHeartbeatManager which can be used to initialize an unset heartbeat manager field. This allows to make the heartbeat manager fields non-nullable which in turn avoid NPE. Moreover, this commit makes the heartbeat manager fields of the TaskExecutor final. This closes #9839. | 08 October 2019, 12:17:15 UTC |
42cb6dd | tison | 19 September 2019, 08:29:29 UTC | [FLINK-14010][coordination] YarnResourceManager#onShutdownRequest triggers fatal error This closes #9719. | 24 September 2019, 12:55:01 UTC |
bd677f2 | tszkitlo40 | 21 September 2019, 05:29:57 UTC | [FLINK-13746][e2e] Whitelist [Terror] to avoid end to end test failure in es 2.3.5 This closes #9738. | 24 September 2019, 12:44:19 UTC |
4a4dae0 | Thomas Weise | 16 September 2019, 20:33:42 UTC | [FLINK-14107][kinesis] Erroneous queue selection in record emitter may lead to deadlock | 18 September 2019, 23:56:39 UTC |
517bd29 | yanghua | 16 September 2019, 12:06:55 UTC | [FLINK-13965] Keep hasDeprecatedKeys and deprecatedKeys methods in ConfigOption and mark it with @Deprecated annotation This closes #9691. | 18 September 2019, 09:42:35 UTC |
b03c943 | Till Rohrmann | 10 September 2019, 13:30:57 UTC | [FLINK-14043] Speed up SavepointMigrationTestBase sub classes Since all SavepointMigrationTestBase sub classes rely on the MigrationTestUtils.AccumulatorCountingSink which uses user code accumulators in order to communicate with the test driver, we set the heartbeat interval to 300ms in order to speed the test execution up. The reason this works is that Flink transports user code accumulators from the TM to the JM via the heartbeats. Hence, the heartbeat interval represents the lower boundary for the test completion. This closes #9666. | 11 September 2019, 12:26:34 UTC |
1d32e39 | Till Rohrmann | 09 September 2019, 07:50:41 UTC | [FLINK-14009][build] Ignore license file check for Scala version different than 2.11 This closes #9651. | 09 September 2019, 16:10:58 UTC |
7de0568 | Chesnay Schepler | 06 September 2019, 12:56:03 UTC | [hotfix][travis] Add detailed instructions for update binary licensing | 06 September 2019, 12:56:32 UTC |
68885b8 | Chesnay Schepler | 04 September 2019, 13:37:20 UTC | [FLINK-13968][travis] Check correctness of binary licensing | 06 September 2019, 11:28:51 UTC |
750648d | Chesnay Schepler | 05 September 2019, 11:40:06 UTC | [FLINK-13892][hs] Harden HistoryServerTest | 06 September 2019, 10:31:04 UTC |
1ebc30c | Chesnay Schepler | 06 September 2019, 10:08:45 UTC | [FLINK-13936][licensing] Update NOTICE-binary | 06 September 2019, 10:12:16 UTC |
863b92b | Chesnay Schepler | 04 September 2019, 13:37:07 UTC | [FLINK-13967][licensing] Fully generate binary licensing | 06 September 2019, 09:24:06 UTC |
ba1cd43 | Chesnay Schepler | 05 September 2019, 07:18:37 UTC | [FLINK-13966][licensing] Pin locale for deterministic sort order | 05 September 2019, 14:49:47 UTC |
d540794 | Chesnay Schepler | 29 August 2019, 11:15:08 UTC | [FLINK-13892][hs] Harden HistoryServerTest | 05 September 2019, 07:42:43 UTC |
655c78f | Aljoscha Krettek | 02 September 2019, 08:39:31 UTC | [FLINK-13586] Make ClosureCleaner.clean() backwards compatible with 1.8.0 | 04 September 2019, 08:16:02 UTC |
94ca735 | Kostas Kloudas | 02 September 2019, 12:35:57 UTC | [FLINK-13941][fs-connector] Do not delete partial part files from S3 upon restore. | 03 September 2019, 13:14:24 UTC |
b7ce7b8 | Mads Chr. Olesen | 03 September 2019, 09:49:10 UTC | [FLINK-13059][cassandra] Release semaphore on exception in send() | 03 September 2019, 09:50:05 UTC |
4fcbefb | Chesnay Schepler | 29 August 2019, 12:39:37 UTC | [FLINK-13897][oss] Move NOTICE file into META-INF directory | 03 September 2019, 09:20:45 UTC |
b79c199 | Fabian Hueske | 03 September 2019, 08:11:37 UTC | [hotfix][docs] Minor fixes in operations playground. [ci skip] | 03 September 2019, 08:11:41 UTC |
d22e15c | haodang | 28 August 2019, 11:47:02 UTC | [FLINK-13789][kafka] Simplify transactional ID generation Remove String.format usage to prevent unexpected behaviors if the configured prefix contains format specifiers. | 28 August 2019, 11:49:11 UTC |
70dfb7a | tison | 28 August 2019, 09:11:36 UTC | [FLINK-13807][tests] Use UTF-8 charset in TestBaseUtils.getResultReader | 28 August 2019, 09:13:02 UTC |
9868ae1 | Fabian Hueske | 19 August 2019, 16:32:08 UTC | [FLINK-12749][docs] Add Flink Operations Playground This closes #9491. [ci skip] | 27 August 2019, 11:11:10 UTC |
c6b07c7 | 张志豪 | 18 August 2019, 13:16:20 UTC | [FLINK-13761][scala] Deprecate Scala SplitStream Deprecate Scala SplitStream which has been superseded by side outputs. This closes #9474. | 22 August 2019, 12:45:56 UTC |
b837e1c | Chesnay Schepler | 22 August 2019, 07:42:28 UTC | [FLINK-13806][metrics] Log all errors on DEBUG | 22 August 2019, 09:06:17 UTC |
481332e | Gary Yao | 12 August 2019, 13:56:31 UTC | [FLINK-13488][tests] Harden ConnectedComponents E2E test By default the tests starts 25 TMs with a single slot each. This is not sustainable on Travis CI. This commit changes the test so that it only starts 2 TMs that each offer 13 slots by default. Run 'set -Eexuo pipefail' at the beginning of the test as recommended by the README.md. | 13 August 2019, 20:15:23 UTC |
0130b86 | Tzu-Li (Gordon) Tai | 07 August 2019, 09:43:02 UTC | [FLINK-13159] [tests] Add subclass serialization cases to PojoSerializerSnapshotMigrationTest This closes #9375. | 08 August 2019, 08:05:50 UTC |
27be11a | Tzu-Li (Gordon) Tai | 08 August 2019, 05:34:04 UTC | [FLINK-13159] Fix incorrect subclass serializer reconfiguration in PojoSerializer | 08 August 2019, 08:05:44 UTC |
c70c0b0 | Yun Tang | 06 August 2019, 18:07:33 UTC | [FLINK-13159] Fix the NPE when PojoSerializer restored | 08 August 2019, 08:05:36 UTC |
954f3c0 | Aljoscha Krettek | 05 August 2019, 08:15:34 UTC | [FLINK-10368] Increase slot request timeout to harden YARN/Kerberos test Before, the tests were sometimes failing with NoResourceAvailableException. In the logs it was visible that the requested TaskExecutors (TMs) were connecting after the exception was thrown. Increasing the timeout therefore fixes the instability. | 06 August 2019, 15:08:41 UTC |
9881c45 | Aljoscha Krettek | 02 August 2019, 12:48:24 UTC | [hotfix] Print Flink logs from YARN in test_yarn_kerberos_docker.sh | 06 August 2019, 15:07:08 UTC |
9441505 | Aljoscha Krettek | 01 August 2019, 11:04:24 UTC | [FLINK-10368] Harden Dockerized Kerberos tests by waiting for NM to be up Before, we didn't wait for Yarn NodeManagers to be up. This meant that sometimes the Flink Job would not have enough resources to run. | 06 August 2019, 15:04:44 UTC |
a76b9e9 | Chesnay Schepler | 24 July 2019, 07:55:08 UTC | [FLINK-13394][travis] Use fallback unsafe MapR repository | 06 August 2019, 11:23:09 UTC |
a0d236f | Gary Yao | 31 July 2019, 11:41:42 UTC | [FLINK-13508][tests] Prevent waitUntilCondition() from sleeping negative time This fixes that CommonTestUtils#waitUntilCondition() may invoke Thread.sleep() with a negative argument. | 04 August 2019, 18:28:18 UTC |
80452f1 | Stephan Ewen | 30 July 2019, 10:00:07 UTC | [FLINK-13499][maprfs] Handle MapR dependency purely through reflection This allows us to remove the MapR dependency from the module. The MapR maven dependency has frequently caused issues. | 02 August 2019, 09:30:33 UTC |
99d0064 | Stephan Ewen | 30 July 2019, 09:04:23 UTC | [hotfix][tests] Refactor MapR FS Tests | 02 August 2019, 09:30:33 UTC |
c4e9743 | ifndef-SleePy | 23 July 2019, 12:29:36 UTC | [FLINK-9526][e2e] Fix unstable BucketingSink end-to-end test | 29 July 2019, 13:51:19 UTC |
9797fec | David Moravek | 22 July 2019, 11:59:08 UTC | [FLINK-13367] Recognize writeReplace in ClosureCleaner This closes #9201 | 25 July 2019, 15:58:03 UTC |
480875f | Kostas Kloudas | 23 July 2019, 13:09:35 UTC | [hotfix] fix checkstyle | 23 July 2019, 13:21:51 UTC |
616d1b8 | David Moravek | 23 July 2019, 12:56:26 UTC | [FLINK-13369] Track references of already visited object in ClosureCleaner | 23 July 2019, 13:21:36 UTC |
91d036f | Paul Lam | 05 June 2019, 11:33:40 UTC | [FLINK-12741] [docs] Update Kafka producer fault tolerance guarantees | 23 July 2019, 09:58:46 UTC |
330516c | Chesnay Schepler | 19 July 2019, 08:43:48 UTC | [FLINK-12578][build] Use fallback unsafe MapR repository | 19 July 2019, 08:45:03 UTC |
148e0dd | Chesnay Schepler | 18 July 2019, 11:01:11 UTC | [FLINK-12578][build] Add fallback unsafe MapR repository | 18 July 2019, 11:09:55 UTC |
fe80cb5 | Chesnay Schepler | 18 July 2019, 11:00:50 UTC | [FLINK-12578][build] Use secure MapR repository by default | 18 July 2019, 11:05:09 UTC |
54c44eb | Andrey Zagrebin | 08 July 2019, 08:51:19 UTC | [hotfix][tests][coordination] Move idle task manager release tests into a separate suite | 10 July 2019, 20:51:05 UTC |
48f5c78 | Andrey Zagrebin | 04 July 2019, 14:12:32 UTC | [FLINK-12736][coordination] Release TaskExecutor in SlotManager only if there were no slot allocations after the partition check The ResourceManager looks out for TaskManagers that have not had any slots allocated on them for a while, as these could be released to safe resources. If such a TM is found, the RM checks via an RPC call whether the TM still holds any partitions. If no partition is held then the TM is released. However, in the RPC callback no check is made whether the TM is actually still idle. In the meantime a slot could have been allocated on the TM. Even if the slot has been freed, there can be newly allocated partitions not included in check result. To make sure there was no resource allocation in between, we can mark the taskManagerRegistration.getIdleSince() time before starting the async 'no partition' check. The TM can be released only if the idle time after the check matches the previously marked one. Otherwise we discard the release and start over after the next timeout. This closes #9041. | 10 July 2019, 20:50:53 UTC |
755ab6f | liyafan82 | 05 July 2019, 06:44:45 UTC | [FLINK-12319][Library/CEP]Change the logic of releasing node from recursive to non-recursive | 05 July 2019, 10:56:19 UTC |
fbfe7d4 | Till Rohrmann | 02 July 2019, 12:57:21 UTC | [FLINK-12889] Set FatalExitExceptionHandler for StreamTask#asyncOperationsThreadPool In order to avoid the swallowing of uncaught exceptions in asynchronous checkpoint operations, this commit sets the FatalExitExceptionHandler for the StreamTask#asyncOperationsThreadPool. For testing purposes the uncaught exception handler was made configurable in the StreamTask. | 02 July 2019, 13:20:55 UTC |