[
https://issues.apache.org/jira/browse/FLINK-8971?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16458462#comment-16458462
]
ASF GitHub Bot commented on FLINK-8971:
---------------------------------------
GitHub user tzulitai opened a pull request:
https://github.com/apache/flink/pull/5941
[FLINK-8971] [e2e] Include broadcast / union state in general purpose
DataStream job
## What is the purpose of the change
This PR extends the general purpose DataStream job to contain an operator
that uses broadcast and union state.
The operator is self-verifiable such that on restore of the job, if
restored broadcast state / union state is incorrect, an exception will be
thrown to fail the job.
The fact that the `test_resume_savepoint` job uses the general purpose
DataStream job ensures that we have an e2e test that covers broadcast / union
state savepointing and resuming.
## Brief change log
- 8315969 a preliminary fix that allows the sequence generator to have only
one key
- eb8df23 Introduces an operator that uses broadcast and union state
## Verifying this change
This is an extension to existing tests.
## Does this pull request potentially affect one of the following parts:
- Dependencies (does it add or upgrade a dependency): (yes / **no**)
- The public API, i.e., is any changed class annotated with
`@Public(Evolving)`: (yes / **no**)
- The serializers: (yes / **no** / don't know)
- The runtime per-record code paths (performance sensitive): (yes /
**no** / don't know)
- Anything that affects deployment or recovery: JobManager (and its
components), Checkpointing, Yarn/Mesos, ZooKeeper: (yes / **no** / don't know)
- The S3 file system connector: (yes / **no** / don't know)
## Documentation
- Does this pull request introduce a new feature? (yes / **no**)
- If yes, how is the feature documented? (**not applicable** / docs /
JavaDocs / not documented)
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/tzulitai/flink FLINK-8971
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/5941.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #5941
----
commit 83159697b4d856b109863baee9d880bd2f1b4c86
Author: Tzu-Li (Gordon) Tai <tzulitai@...>
Date: 2018-04-30T10:04:43Z
[hotfix] [e2e-tests] Make SequenceGeneratorSource usable for 0-size key
ranges
commit eb8df230d1963ac7ca0e819aa5eb203d8a8682bf
Author: Tzu-Li (Gordon) Tai <tzulitai@...>
Date: 2018-04-30T10:05:46Z
[FLINK-8971] [e2e-tests] Include broadcast / union state in general purpose
DataStream job
----
> Create general purpose testing job
> ----------------------------------
>
> Key: FLINK-8971
> URL: https://issues.apache.org/jira/browse/FLINK-8971
> Project: Flink
> Issue Type: Task
> Components: Tests
> Affects Versions: 1.5.0
> Reporter: Till Rohrmann
> Priority: Critical
> Fix For: 1.5.0
>
>
> In order to write better end-to-end tests we need a general purpose testing
> job which comprises as many Flink aspects as possible. These include
> different types for records and state, user defined components, state types
> and operators.
> The job should allow to activate a certain misbehavior, such as slowing
> certain paths down or throwing exceptions to simulate failures.
> The job should come with a data generator which generates input data such
> that the job can verify it's own behavior. This includes the state as well as
> the input/output records.
> We already have the [heavily misbehaved
> job|https://github.com/rmetzger/flink-testing-jobs/blob/flink-1.4/basic-jobs/src/main/java/com/dataartisans/heavymisbehaved/HeavyMisbehavedJob.java]
> which simulates some misbehavior. There is also the [state machine
> job|https://github.com/dataArtisans/flink-testing-jobs/tree/master/streaming-state-machine]
> which can verify itself for invalid state changes which indicate data loss.
> We should incorporate their characteristics into the new general purpose job.
> Additionally, the general purpose job should contain the following aspects:
> * Job containing a sliding window aggregation
> * At least one generic Kryo type
> * At least one generic Avro type
> * At least one Avro specific record type
> * At least one input type for which we register a Kryo serializer
> * At least one input type for which we provide a user defined serializer
> * At least one state type for which we provide a user defined serializer
> * At least one state type which uses the AvroSerializer
> * Include an operator with ValueState
> * Value state changes should be verified (e.g. predictable series of values)
> * Include an operator with operator state
> * Include an operator with broadcast state
> * Broadcast state changes should be verified (e.g. predictable series of
> values)
> * Include union state
> * User defined watermark assigner
> The job should be made available in the {{flink-end-to-end-tests}} module.
> This issue is intended to serve as an umbrella issue for developing and
> extending this job.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
