[ZEPPELIN-619] Shared Resource pool across interpreter processes#655
[ZEPPELIN-619] Shared Resource pool across interpreter processes#655Leemoonsoo wants to merge 17 commits into
Conversation
b146820 to
9f1105d
Compare
9f1105d to
9d288fe
Compare
|
A question @Leemoonsoo, it resource sharing only available through the ZeppelinContext object of the Spark interpreter or is it also available to other interpreters ? |
|
@doanduyhai |
|
Great news! We have finally a way to share variables between interpreters! |
There was a problem hiding this comment.
Consider using ConcurrentHashMap to remove the synchronize block.
Same remark for other synchronize blocks within this class
|
Another question, now that we have a resource pull to share data between interpreters, should we deprecate AngularObjectRegistry and use resource pool instead ? I have the feeling that resource pool is more general purpose. Furthermore with AngularObject we have the notion of scope (paragraph or note) which we don't have with resource pool. It helps segregate data properly Anyway it is just a question, it's not blocking for this PR |
|
@doanduyhai Thanks for review and i think i addressed all your comment. Please take a look. Regarding ResourcePool and AngularObjectRegistry, here's differences
I agree, if we improve ResourcePool more and replace AngularObjectRegistry, that would be simpler and better. |
|
@Leemoonsoo Thanks for the detailed explanation of differences between AngularObjectRegistry and ResourcePool. I guess that the AngularObjectRegistry is still very relevant since it has a scope system and is saved with notebooks in json. Otherwise, the code looks good to me so +1 for merge |
|
Thanks @doanduyhai for the review. |
### What is this PR for? Master branch build failure after merging #655 #591 ### What type of PR is it? Hot Fix ### Questions: * Does the licenses files need update? no * Is there breaking changes for older versions? no * Does this needs documentation? no Author: Lee moon soo <moon@apache.org> Closes #680 from Leemoonsoo/ZEPPELIN-619_followup and squashes the following commits: beac930 [Lee moon soo] Fix test
What is this PR for?
This is sub task of https://issues.apache.org/jira/browse/ZEPPELIN-533.
It provides shared resource pool to exchange data across interpreter processes.
What type of PR is it?
Feature
Is there a relevant Jira issue?
https://issues.apache.org/jira/browse/ZEPPELIN-619
How should this be tested?
create two different spark interpreter settings.
create two different notebooks each bind different spark interpreter setting.
put an object from one notebook.
read the object from the other notebook. (from the other interpreter process)
See screenshot
Screenshots (if appropriate)
Questions: