[
https://issues.apache.org/jira/browse/FLINK-7607?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16598545#comment-16598545
]
Mike Pedersen commented on FLINK-7607:
--------------------------------------
Yeah, I'm having the same problem. We are building a machine learning pipeline
with a series of transforms for each of many features. This results in a graph
of approx. 1300 tasks, which Flink itself handles just fine, but totally hangs
the UI.
> Web Frontend Hangs with Large Numbers of Tasks
> ----------------------------------------------
>
> Key: FLINK-7607
> URL: https://issues.apache.org/jira/browse/FLINK-7607
> Project: Flink
> Issue Type: Bug
> Components: Webfrontend
> Affects Versions: 1.3.2
> Environment: Attempted to load the web frontend on a MacBook Pro 15"
> (late 2016) with 16 GB of memory using both Chrome 60.0 and Safari 10.1.2.
> Reporter: Joshua Griffith
> Assignee: Steven Langbroek
> Priority: Major
> Labels: performance
>
> Viewing a job with a high number of tasks in the web front-end causes the
> page to hang, consuming 100% CPU on a core. At 200 tasks the page slows
> noticeably and scrolling results in long, non-responsive pauses. At 400 tasks
> the page only updates once per minute and is almost entirely non-responsive.
> Initially, I thought this was caused by rendering a complex job graph but
> opening the inspector and deleting the canvas did not improve page
> performance. Further inspection indicated that the page was redrawing every
> DOM element in the task list on every update.
> A possible solution is to use an approach similar to
> [react-list|https://github.com/orgsync/react-list] and only request
> data/render list items that are in view and only update DOM nodes that have
> changed.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
