tillrohrmann commented on a change in pull request #9438: [WIP] [FLINK-13712] [docs] Add release notes for Flink 1.9.0 URL: https://github.com/apache/flink/pull/9438#discussion_r314289274
########## File path: docs/release-notes/flink-1.9.md ########## @@ -0,0 +1,175 @@ +--- +title: "Release Notes - Flink 1.9" +--- +<!-- +Licensed to the Apache Software Foundation (ASF) under one +or more contributor license agreements. See the NOTICE file +distributed with this work for additional information +regarding copyright ownership. The ASF licenses this file +to you under the Apache License, Version 2.0 (the +"License"); you may not use this file except in compliance +with the License. You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + +Unless required by applicable law or agreed to in writing, +software distributed under the License is distributed on an +"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +KIND, either express or implied. See the License for the +specific language governing permissions and limitations +under the License. +--> + + +These release notes discuss important aspects, such as configuration, behavior, +or dependencies, that changed between Flink 1.8 and Flink 1.9. It also provides an overview +on known shortcoming or limitations with new experimental features introduced in 1.9. + +Please read these notes carefully if you are planning to upgrade your Flink version to 1.9. + +* This will be replaced by the TOC +{:toc} + +## Known shortcomings or limitations for new features + +### New Table / SQL Blink runner + +TBD + +### Java 9 support + +Since Flink 1.9.0, Flink can now be compiled and run on Java 9. Note that certain components interacting +with eternal systems (connectors, filesystems, metric reporters, etc.) may not work since the respective projects may +have skipped Java 9 support. + +Related issues: +- [FLINK-8033: JDK 9 support](https://issues.apache.org/jira/browse/FLINK-8033) + +## Deprecations and breaking changes + +### Failover strategies + +As a result of completion of fine-grained recovery ([FLIP-1](https://cwiki.apache.org/confluence/display/FLINK/FLIP-1+%3A+Fine+Grained+Recovery+from+Task+Failures)), +Flink will now attempt to only restart tasks that are +connected to failed tasks through a pipelined connection. By default, the `region` failover strategy is used. + +Users who were not using a restart strategy or have already configured a failover strategy should not be affected. +Moreover, users who already enabled the `region` failover strategy, along with a restart strategy that enforces a +certain number of restarts or introduces a restart delay, will see changes in behavior. +The `region` failover strategy now correctly respects constraints that are defined by the restart strategy. + +Streaming users who were not using a failover strategy may be affected if their jobs are embarrassingly parallel or +contain multiple independent jobs. In this case, only the failed parallel pipeline or affected jobs will be restarted. + +Batch users may be affected if their job contains blocking exchanges (usually happens for shuffles) or the +`ExecutionMode` was set to `BATCH` or `BATCH_FORCED` via the `ExecutionConfig`. + +Overall, users should see an improvement in performance. + +Related issues: +- [FLINK-13223: Set jobmanager.execution.failover-strategy to region in default flink-conf.yaml](https://issues.apache.org/jira/browse/FLINK-13223) +- [FLINK-13060: FailoverStrategies should respect restart constraints](https://issues.apache.org/jira/browse/FLINK-13060) + +### Job termination via CLI + +With the support of graceful job termination with savepoints for semantic correctness +([FLIP-34](https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=103090212)), a few changes +related to job termination has been made to the CLI. + +From now on, the `stop` command with no further arguments stops the job with a savepoint targeted at the +default savepoint location (as configured via the `state.savepoints.dir` property in the job configuration), +or a location explicitly specified using the `-p <savepoint-path>` option. +Since job terminations are now always accompanied with a savepoint, stopping jobs is expected to take longer now. + +Related issues: +- [FLINK-13123: Align Stop/Cancel Commands in CLI and REST Interface and Improve Documentation](https://issues.apache.org/jira/browse/FLINK-13123) +- [FLINK-11458: Add TERMINATE/SUSPEND Job with Savepoint](https://issues.apache.org/jira/browse/FLINK-11458) + +### Network stack + +A few changes in the network stack related to changes in the threading model of `StreamTask` to a mailbox-based approach +requires close attention to some related configuration: + +- Due to changes in the lifecycle management of result partitions, partition requests as well as re-triggers will now +happen sooner. Therefore, it is possible that some jobs with long deployment times and large state might start failing +more frequently with `PartitionNotFound` exceptions compared to previous versions. If that's the case, users should +increase the value of `taskmanager.network.request-backoff.max` in order to have the same effective partition +request timeout as it was prior to 1.9.0. + +- To avoid a potential deadlock, a timeout has been added for how long a task will wait for assignment of exclusive +memory segments. The default timeout is 30 seconds, and is configurable via `taskmanager.network.memory.exclusive-bubffers-request-timeout-ms`. Review comment: typo in `bubffers` -> `buffers` ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
