[jira] [Commented] (FLINK-964) Integrate profiling code with web interface

["Daniel Warneke (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-964?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14053896#comment-14053896 ] 

Daniel Warneke commented on FLINK-964:
--------------------------------------

Unfortunately, I don’t live in Berlin, so a face-to-face meeting won’t be easily possible. However, I’m sure we can figure it out online, too. Currently, the profiling code continuously generates events containing the following information:

- CPU utilization (USR/SYS/…) per server (ranging from 0-100%)
- Memory utilization per server (in MB)
- Network utilization per server (send/received MBit/s)
- The above three metrics aggregated for the whole cluster
- CPU utilization per thread (ranging from 0-100%), derived from how much time a JVM thread spends in its possible states RUNNING, WAITING, BLOCKED, …

Every event comes with a timestamp, so it’s relatively easy to plot them in a chart.
If you tell me where you need these events to arrive, I can make sure they get there. Maybe you could take the web chart drawing from there? I’m not really an expert in this area…


> Integrate profiling code with web interface
> -------------------------------------------
>
>                 Key: FLINK-964
>                 URL: https://issues.apache.org/jira/browse/FLINK-964
>             Project: Flink
>          Issue Type: Bug
>    Affects Versions: 0.6-incubating
>            Reporter: Stephan Ewen
>
> This issue is subject to discussion.
> The profiling code currently needs to be kept in sync with the job graph code, execution graph code, and runtime code.
> Since that part of the code is undergoing quite some changes and the profiling code is not used right now, I suggest to remove it, or move it to an artifact repository.



--
This message was sent by Atlassian JIRA
(v6.2#6252)