You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@toree.apache.org by "Chip Senkbeil (JIRA)" <ji...@apache.org> on 2016/02/19 01:10:18 UTC

[jira] [Comment Edited] (TOREE-264) Support multiple sessions in the kernel each with their own isolated interpreters

    [ https://issues.apache.org/jira/browse/TOREE-264?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15153411#comment-15153411 ] 

Chip Senkbeil edited comment on TOREE-264 at 2/19/16 12:09 AM:
---------------------------------------------------------------

One issue with multiple interpreters is that you can have only one REPL class server per Spark Context. We'd need to alter the package names of classes generated by the REPL to not collide with other REPLs.


was (Author: senkwich):
One issue with multiple interpreters is that you can have only one REPL class server for all of your Spark cluster. We'd need to alter the package names of classes generated by the REPL to not collide with other REPLs.

> Support multiple sessions in the kernel each with their own isolated interpreters
> ---------------------------------------------------------------------------------
>
>                 Key: TOREE-264
>                 URL: https://issues.apache.org/jira/browse/TOREE-264
>             Project: TOREE
>          Issue Type: Improvement
>            Reporter: Gino Bustelo
>
> Many folks are very interested in having multiple users connecting to one Kernel/Toree process and sharing the same SparkContext. 
> I've always considered us already supporting thins use-case, but with the caveat that all users also share the same interpreter and can lead to lots of race-conditions.
> I'm thinking that we could more sanely support multiple users by routing them to isolated interpreters that share the SparkContext but have independent variable namespace.
> Thoughs?



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