[
https://issues.apache.org/jira/browse/FLINK-39494?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18079604#comment-18079604
]
Yuepeng Pan edited comment on FLINK-39494 at 5/9/26 2:06 AM:
-------------------------------------------------------------
Thanks [~chshining] so much for the reporting.
I have confirmed the issue you mentioned; it is an intermittent and unintended
behavior.
I believe we should fix it.
Of course, this requires at least a two-fold assessment: *
{*}Short-term:{*} Can we identify the root cause within a short window, say
one or two business days?
* {*}If not:{*} If it cannot be fixed quickly and isn't urgent, we should at
least document this known issue in the feature description to keep users
informed.
Regardless of whether we decide to fix it immediately, I think it’s necessary
to create a new Jira ticket to track this. Would you mind helping to create one?
Please evaluate based on your availability and let me know your thoughts.
BTW, the unstable order issue doesn't affect the correctness for showing
rescales related data.
was (Author: JIRAUSER301104):
Thanks [~chshining] so much for the reporting.
I have confirmed the issue you mentioned; it is an intermittent and unintended
behavior.
I believe we should fix it.
Of course, this requires at least a two-fold assessment: *
{*}Short-term:{*} Can we identify the root cause within a short window, say
one or two business days?
*
{*}If not:{*} If it cannot be fixed quickly and isn't urgent, we should at
least document this known issue in the feature description to keep users
informed.
Regardless of whether we decide to fix it immediately, I think it’s necessary
to create a new Jira ticket to track this. Would you mind helping to create one?
Please evaluate based on your availability and let me know your thoughts.
> Release Testing: Verify FLIP-487: Show history of rescales in Web UI for
> AdaptiveScheduler
> ------------------------------------------------------------------------------------------
>
> Key: FLINK-39494
> URL: https://issues.apache.org/jira/browse/FLINK-39494
> Project: Flink
> Issue Type: Sub-task
> Components: Tests
> Affects Versions: 2.3.0
> Reporter: Yuepeng Pan
> Assignee: Chanhae Oh
> Priority: Blocker
> Fix For: 2.3.0
>
> Attachments: image-2026-04-29-22-45-23-546.png,
> image-2026-04-29-22-49-25-250.png, image-2026-04-29-22-51-54-495.png,
> image-2026-04-29-22-52-08-177.png, image-2026-04-29-22-52-20-391.png,
> image-2026-04-29-22-52-36-259.png, image-2026-04-29-22-59-13-102.png,
> image-2026-04-29-22-59-27-876.png, image-2026-04-29-22-59-36-834.png,
> image-2026-04-29-23-00-12-453.png, image-2026-04-29-23-00-23-748.png,
> image-2026-04-29-23-01-35-076.png, image-2026-04-29-23-04-24-498.png,
> image-2026-04-29-23-04-33-591.png, image-2026-04-29-23-06-40-867.png,
> image-2026-04-29-23-06-52-568.png, image-2026-04-29-23-07-21-359.png,
> image-2026-04-29-23-07-44-964.png, image-2026-04-29-23-08-00-925.png,
> image-2026-04-29-23-09-05-876.png, image-2026-04-29-23-09-21-305.png,
> image-2026-04-29-23-09-31-426.png, image-2026-04-29-23-09-41-317.png,
> image-2026-04-29-23-09-53-011.png, image-2026-04-29-23-10-02-825.png,
> image-2026-05-09-02-00-49-444.png
>
>
> h1. 1 Prepare to build Flink binaries and test jobs
>
> 1.1. Environment requirements:
> Standard JDK and Maven setups for building Flink from source will suffice.
> h3.
> 1.2. Build flink binary file
> {code:java}
> # cd into you target directory
> git clone https://github.com/apache/flink # $repo_path, the placeholder name
> will be used later.
> cd flink
> git checkout ${target_tag or target_branch}
> mvn clean package -DskipTests -T 4C {code}
> h3.
> 1.3. Build an example job
> {code:java}
> mvn archetype:generate \
> -DarchetypeGroupId=org.apache.flink \
> -DarchetypeArtifactId=flink-walkthrough-datastream-java \
> -DarchetypeVersion=2.2.0 \
> -DgroupId=frauddetection \
> -DartifactId=frauddetection \
> -Dversion=0.1 \
> -Dpackage=frauddetection \
> -DinteractiveMode=false
> cd frauddetection
> mvn clean package -DskipTests {code}
> h1. 2 Testing cases for streaming jobs
> h2. 2.1 For disabled adaptive scheduler and disabled max rescale history
> 2.1.1. run shell command:
>
> {code:java}
> cd $repo_path/build-target
> ./bin/start-cluster.sh {code}
> 2.1.2. Visit [http://localhost:8081/]
> 2.1.3. Use flink WEB UI to submit the frauddetection jar as a job.
> 2.1.4. Check if the UI shown logic is correct
> The testing case will be expected if the there isn't Rescale tab page in UI
> like follows,
> If not , this test would be tagged failed.
>
> !image-2026-04-29-22-45-23-546.png!
> 2.1.5. run shell command to reset the env
> {code:java}
> cd $repo_path/build-target
> ./bin/stop-cluster.sh {code}
>
>
> h2. 2.2 For disabled adaptive scheduler and enabled rescale history
>
> 2.2.1. Set the rescale history option in
> `$repo_path/build-target/conf/config.yaml` as follows:
>
> {code:java}
> web:
> adaptive-scheduler:
> rescale-history:
> size: 2 {code}
> 2.2.2. run shell command:
>
> {code:java}
> cd $repo_path/build-target
> ./bin/start-cluster.sh {code}
> 2.2.3. then visit [http://localhost:8081/]
> 2.2.4. Use flink WEB UI to submit the frauddetection jar as a job.
> 2.2.5. Check that the testing case will be expected if the there isn't
> Rescales tab page in UI like follows:
> after that,
> The testing case will be expected if the there isn't Rescale tab page in UI
> like follows, If not , this test would be tagged failed.
> !image-2026-04-29-22-49-25-250.png!
> 2.2.6. Delete the following changed config content in
> `$repo_path/build-target/conf/config.yaml`
> {code:java}
> web:
> adaptive-scheduler:
> rescale-history:
> size: 2 {code}
> 2.2.7 run shell command to stop flink cluster:
> {code:java}
> cd $repo_path/build-target
> ./bin/stop-cluster.sh {code}
>
>
> h2. 2.3 For enabled adaptive scheduler and disabled rescale history
>
> 2.3.1. Add the rescale history option into
> `$repo_path/build-target/conf/config.yaml` as follows:
>
> {code:java}
> jobmanager:
> scheduler: adaptive {code}
> 2.3.2. run shell command:
>
> {code:java}
> cd $repo_path/build-target
> ./bin/start-cluster.sh {code}
> 2.3.3. Visit [http://localhost:8081/]
> 2.3.4. Use flink WEB UI to submit the frauddetection jar as a job.
> 2.3.5. Check UI
> The testing case will be expected if the there isn't Rescale tab page in UI
> like follows, if not , this test would be tagged failed：
> - For Overview page
>
> !image-2026-04-29-22-51-54-495.png!
> - For History page
>
> !image-2026-04-29-22-52-08-177.png!
> - For Summary page
>
> !image-2026-04-29-22-52-20-391.png!
> - For Configuration page
> Only the Configuration sub-page is not empty.
> !image-2026-04-29-22-52-36-259.png!
> 2.3.6. Delete the config changed content in
> `$repo_path/build-target/conf/config.yaml`
>
> {code:java}
> jobmanager:
> scheduler: adaptive {code}
> 2.3.7. run shell command to stop cluster
> {code:java}
> cd $repo_path/build-target
> ./bin/stop-cluster.sh {code}
> h2. 2.4 For enabled adaptive scheduler and enabled rescale history
>
> 2.4.1. Add the rescale history option in
> `$repo_path/build-target/conf/config.yaml` as follows:
>
> {code:java}
> jobmanager:
> scheduler: adaptive
> web:
> adaptive-scheduler:
> rescale-history:
> size: 3
> taskmanager:
> numberOfTaskSlots: 8 {code}
> 2.4.2. run shell command:
>
> {code:java}
> cd $repo_path/build-target
> ./bin/start-cluster.sh {code}
> 2.4.3. Visit [http://localhost:8081/]
> 2.4.4. Use flink WEB UI to submit the frauddetection jar as a job.
> 2.4.5. Manually adjust the job parallelism to trigger a rescale.
> After the job is successfully scheduled for the first time, as shown in the
> figure below, manually adjust the job parallelism twice via the subsequent
> prompts to actively generate two new rescales.
>
> !image-2026-04-29-22-59-13-102.png!
> Please note that the interval between the two manual rescale triggers must be
> very short. As shown in the figure below, immediately after sending the first
> request to increase parallelism (a notification will appear in the
> upper-right corner), initiate the second parallelism increase without delay.
> Then, verify whether an 'ignored' type rescale is generated. Since the
> minimum interval between two manual rescales can be difficult to gauge, it is
> recommended to repeat the attempt several times until an 'ignored' rescale
> record appears. This is primarily to test whether the generation of 'Ignored'
> type Rescales is functioning correctly
>
> !image-2026-04-29-22-59-27-876.png!
> !image-2026-04-29-22-59-36-834.png!
> 2.4.6. Check the related UI pages
> - Overview
> !image-2026-04-29-23-04-24-498.png!
> - Overview-View details and table header hints.
> !image-2026-04-29-23-06-40-867.png!
> - History
> -- for brief rescale view
> !image-2026-04-29-23-06-52-568.png!
> – View details and table header hints.
>
> !image-2026-04-29-23-07-21-359.png!
> – for detailed rescale
>
> !image-2026-04-29-23-07-44-964.png!
> !image-2026-04-29-23-08-00-925.png!
> – View details and table header hints.
>
> !image-2026-04-29-23-09-05-876.png!
> - Summary
> !image-2026-04-29-23-09-21-305.png!
> !image-2026-04-29-23-09-31-426.png!
> - Configuration
>
> !image-2026-04-29-23-09-41-317.png!
> 2.4.7. Manually trigger another rescale and verify whether the History page
> retains the most recent rescale records as specified by the
> `web.adaptive-scheduler.rescale-history.size` configuration value.
> !image-2026-04-29-23-09-53-011.png!
> !image-2026-04-29-23-10-02-825.png!
> 2.4.8. Run shell command to stop cluster:
> ```shell
> cd $repo_path/build-target
> ./bin/stop-cluster.sh
> ```
>
>
> h1. 3 Please record anything unexpected by comments in the current ticket.
> Thank you.
>
>
>
>
>
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
