flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (FLINK-5941) Let handlers take part in job archiving
Date Thu, 02 Mar 2017 12:50:45 GMT

    [ https://issues.apache.org/jira/browse/FLINK-5941?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15892186#comment-15892186
] 

ASF GitHub Bot commented on FLINK-5941:
---------------------------------------

Github user zentol commented on a diff in the pull request:

    https://github.com/apache/flink/pull/3444#discussion_r103916234
  
    --- Diff: flink-runtime-web/src/main/java/org/apache/flink/runtime/webmonitor/WebRuntimeMonitor.java
---
    @@ -424,6 +425,33 @@ protected void initChannel(SocketChannel ch) {
     		LOG.info("Web frontend listening at " + address + ':' + port);
     	}
     
    +	public static Archiver[] getArchivers() {
    --- End diff --
    
    i wanted to do something like this originally. However, since the archiving is done in
the ``┬┤MemoryArchivist``` i would have to pass the ```WebRuntimeMonitor``` reference to it.
This is problematic since you now add an order in which they have to created. This specifically
caused issues with the Yarn/Mesos runner that create the JobManager before the WebRuntimeMonitor.
    
    It's a stupid problem and i agree that the static method approach is rather brittle, but
i couldn't find a good solution for the above issue.


> Let handlers take part in job archiving
> ---------------------------------------
>
>                 Key: FLINK-5941
>                 URL: https://issues.apache.org/jira/browse/FLINK-5941
>             Project: Flink
>          Issue Type: New Feature
>          Components: Webfrontend
>            Reporter: Chesnay Schepler
>            Assignee: Chesnay Schepler
>             Fix For: 1.3.0
>
>
> The key idea behind the HistoryServer is to pre-compute all JSON responses which the
WebFrontend could request and store them as files in a directory structure resembling the
REST-API.
> For this require a mechanism to generate the responses and their corresponding REST URL.
> FLINK-5852 made it easier to re-use the JSON generation code, while FLINK-5870 made handlers
aware of the REST URLs that they are registered one.
> The aim of this JIRA is to extend job-related handlers, building on the above JIRAs,
enabling them to generate a number of (Path, Json) pairs for a given ExecutionGraph, containing
all responses that they could generate for the given graph and their respective REST URL..



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Mime
View raw message