[
https://issues.apache.org/jira/browse/DRILL-4246?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15097550#comment-15097550
]
ASF GitHub Bot commented on DRILL-4246:
---------------------------------------
GitHub user jacques-n opened a pull request:
https://github.com/apache/drill/pull/323
DRILL-4246: Fix Allocator concurrency bug and improve error detection
- Rename the internal DrillBuf field to udle to better express its purpose.
- Rename AllocatorManager to AllocationManager to better express its
purpose.
- Address situation where dangling ledger could be transferred into while
it was being released released by protecting association and release inside the
AllocationManager.
- Add allocator assertions to ensure allocator operations are done while
the allocator is open.
- Simplify AllocationManager locking model.
- Exclude HDFS reference to netty-all
- Improve debugging messages for allocators (and fix debug message bugs)
Tests pending but evaluation shows previously frequently failing flatten
tests to consistently pass.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/jacques-n/drill DRILL-4246
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/drill/pull/323.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #323
----
commit 4b193d71a9f3e06aadf59c66832bcbcb4b060c92
Author: Jacques Nadeau <[email protected]>
Date: 2016-01-11T04:08:03Z
DRILL-4246: Fix Allocator concurrency bug and improve error detection
- Rename the internal DrillBuf field to udle to better express its purpose.
- Rename AllocatorManager to AllocationManager to better express its
purpose.
- Address situation where dangling ledger could be transferred into while
it was being released released by protecting association and release inside the
AllocationManager.
- Add allocator assertions to ensure allocator operations are done while
the allocator is open.
- Simplify AllocationManager locking model.
- Exclude HDFS reference to netty-all
- Improve debugging messages for allocators (and fix debug message bugs)
----
> New allocator causing a flatten regression test to fail with
> IllegalStateException
> ----------------------------------------------------------------------------------
>
> Key: DRILL-4246
> URL: https://issues.apache.org/jira/browse/DRILL-4246
> Project: Apache Drill
> Issue Type: Bug
> Components: Execution - Relational Operators
> Affects Versions: 1.5.0
> Reporter: Deneche A. Hakim
> Assignee: Jacques Nadeau
> Priority: Blocker
>
> We are seeing the following error in the test cluster:
> {noformat}
> /framework/resources/Functional/flatten_operators/100000rows/filter3.q
> Query:
> select uid, flatten(events) from `data.json` where uid > 1
> Failed with exception
> java.sql.SQLException: SYSTEM ERROR: IllegalStateException: Unaccounted for
> outstanding allocation (851968)
> Allocator(op:0:0:0:Screen) 1000000/851968/1941504/10000000000
> (res/actual/peak/limit)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
