[jira] [Commented] (STORM-2305) STORM-2279 calculates task index different from grouper code

["Jungtaek Lim (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/STORM-2305?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15831119#comment-15831119 ] 

Jungtaek Lim commented on STORM-2305:
-------------------------------------

Turns out 1.x and 1.0.x branch were correct. STORM-2279 should be reverted for 1.x and 1.0.x branches, and this issue and STORM-2279 should be only applied to master.

> STORM-2279 calculates task index different from grouper code
> ------------------------------------------------------------
>
>                 Key: STORM-2305
>                 URL: https://issues.apache.org/jira/browse/STORM-2305
>             Project: Apache Storm
>          Issue Type: Bug
>          Components: storm-core
>    Affects Versions: 2.0.0
>            Reporter: Jungtaek Lim
>            Priority: Blocker
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> Arun reported this from https://github.com/apache/storm/pull/1866#discussion_r95326528
> Quoting his comment:
> {quote}
> This should match the task selected by fields grouping https://github.com/apache/storm/blob/master/storm-core/src/jvm/org/apache/storm/daemon/GrouperFactory.java#L159.
> The modulo technique will return a different value than Math.abs. Probably we should move it to some common utility function and use it in both places.
> In 1.x branch the fields grouper doesn't seem to account for negative hashCode, so not sure why didn't it come up before. https://github.com/apache/storm/blob/1.x-branch/storm-core/src/clj/org/apache/storm/daemon/executor.clj#L52
> {quote}
> So we should fix Nimbus code to use Math.abs on master, 1.x, 1.0.x branches, and also fix executor.clj to use Math.abs on 1.x, 1.0.x branches.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)