[jira] [Commented] (FLINK-20125) Test memory configuration display in the web ui

["Xintong Song (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-20125?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17236037#comment-17236037 ] 

Xintong Song commented on FLINK-20125:
--------------------------------------

I have tested with the following deployment modes.
 * Native Kubernetes (session & application)
 * Yarn (session, single job & application)
 * Standalone
 * Local (where JM/TM runs in the client's process)

I have verified the following behaviors.
 * JM metrics page looks as proposed in FLIP-104
 * TM metrics page looks as proposed in FLIP-102
 * Changing memory configurations, it shows correct configured values for both JM & TM
 * Running a job, the usage metrics change as expected.
 * The cluster configuration page shows memory related configurations correctly.
 * All 8 newly introduced metrics for a TM can be fetched via rest API.

There are a few issues that I would like bring attention to. Not all of them are necessarily problems.
 # Configured memory sizes for JM are displayed in bytes. It would be better to use proper units here, same as what we do for TM. (FLINK-20258)
 # The column title `Configured Values` might be a bit confusing. Not all of the values are explicitly configured by users. There could be values that are automatically derived from users' configuration. I would suggest to add a bit explanation (e.g., a (i) with some hidden texts) for both JM and TM. (FLINK-20259)
 # The cluster configurations page shows memory sizes related to JM configurations in bytes, even if users have explicitly configured them with different units. This is probably due to the client/scripts derives the accurate bytes from the configuration, and overwrite it to the JM start up arguments. Not sure if this necessarily a problem that needs to be fixed.
 # For local execution, the displayed memory sizes and metrics are not correct (see screenshot). 
 ## Client, JM & TM shares the same process. The heap/metaspace usages displayed on the JM/TM metrics page are not only used by JM/TM. 
 ## Configured values are for Heap/Offheap/Metaspace/Overhead are not taking effect, because the JVM process is started before creating JM/TM.

For (4), ideally I would suggest not to display the memory model, configured values and metrics for local execution if we cannot guarantee the values are correct and meaningful. If this cannot be achieved easily, I'm also ok with the current way. We did mentioned in the docs that some of the memory configurations will be ignored for local execution.
https://ci.apache.org/projects/flink/flink-docs-master/ops/memory/mem_setup_jobmanager.html#local-execution
https://ci.apache.org/projects/flink/flink-docs-master/ops/memory/mem_setup_tm.html#local-execution

!local execution screenshot.png!

> Test memory configuration display in the web ui
> -----------------------------------------------
>
>                 Key: FLINK-20125
>                 URL: https://issues.apache.org/jira/browse/FLINK-20125
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Runtime / Coordination
>    Affects Versions: 1.12.0
>            Reporter: Robert Metzger
>            Assignee: Xintong Song
>            Priority: Critical
>             Fix For: 1.12.0
>
>         Attachments: local execution screenshot.png
>
>
> The following questions should be answered for the different deployment modes (application cluster, per-Job cluster, session cluster) and deployment backends (Standalone/Local, Kubernetes, Yarn, Mesos):
> - Does the TaskManager metrics page (path: /#/task-manager/<taskmanager-id>/metrics) looks as proposed in FLIP-102 (minor improvements where introduced in )? It helps to have a job running to see whether the metrics in the right column are actually dynamically changing.
> - Check whether all 8 newly introduced metrics are exposed through the TaskManager's metric endpoint:
> - Does the JobManager metrics page (path: /#/job-manager/metrics) as proposed in FLIP-104?
> - Does the Cluster configuration endpoint include /#/job-manager/config include memory-related parameters like jobmanager.memory.heap.size, jobmanager.memory.jvm-metaspace.size, or jobmanager.memory.off-heap.size? Do they match configuration parameters possibly set in the config file?
> Improved unaligned checkpointing times under backpressure
> (See wiki for more details!)
> ----
> [General Information about the Flink 1.12 release testing|https://cwiki.apache.org/confluence/display/FLINK/1.12+Release+-+Community+Testing]
> When testing a feature, consider the following aspects:
> - Is the documentation easy to understand
> - Are the error messages, log messages, APIs etc. easy to understand
> - Is the feature working as expected under normal conditions
> - Is the feature working / failing as expected with invalid input, induced errors etc.
> If you find a problem during testing, please file a ticket (Priority=Critical; Fix Version = 1.12.0), and link it in this testing ticket.
> During the testing keep us updated on tests conducted, or please write a short summary of all things you have tested in the end.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)