[jira] [Commented] (HIVE-24316) Upgrade ORC from 1.5.6 to 1.5.8 in branch-3.1

2021-08-24 Thread Sungwoo (Jira)


[ 
https://issues.apache.org/jira/browse/HIVE-24316?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17403893#comment-17403893
 ] 

Sungwoo commented on HIVE-24316:


Hello,

It seems that with ORC-361, the use of MemoryManagerImpl in 
LlapAwareMemoryManager is inconsistent. 

Before merging ORC-361, LlapAwareMemoryManager sets its own totalMemoryPool and 
MemoryManagerImpl accesses totalMemoryPool via getTotalMemoryPool(), so 
everything is fine.

With ORC-361 merged, we have the following:

1. LlapAwareMemoryManager sets its own totalMemoryPool as a private field.
 2. MemoryManagerImpl sets its own totalMemoryPool as a private field.
 3. LlapAwareMemoryManager overrides getTotalMemoryPool() using its own 
totalMemoryPool.

Now it is unclear whether or not getTotalMemoryPool() should be overridden.

Here are my thoughts on ORC-361:

1. Is MemoryManagerImpl intended to coordinate all threads writing to ORC files 
inside a process (like LLAP Daemon)? Then is it necessary to create 
LlapAwareMemoryManager as a ThreadLocal object? Why not just call 
OrcFile.getStaticMemoryManager() to obtain the shared MemoryManagerImpl?

3. LlapAwareMemoryManager sets its own totalMemoryPool:
{code:java}
  long memPerExecutor = LlapDaemonInfo.INSTANCE.getMemoryPerExecutor();
  totalMemoryPool = (long) (memPerExecutor * maxLoad);
{code}
>From my understanding, this has no effect because MemoryManagerImpl sets its 
>own totalMemoryPool.

Any comment would be appreciated.

> Upgrade ORC from 1.5.6 to 1.5.8 in branch-3.1
> -
>
> Key: HIVE-24316
> URL: https://issues.apache.org/jira/browse/HIVE-24316
> Project: Hive
>  Issue Type: Bug
>  Components: ORC
>Affects Versions: 3.1.3
>Reporter: Dongjoon Hyun
>Assignee: Dongjoon Hyun
>Priority: Major
>  Labels: pull-request-available
> Fix For: 3.1.3
>
>  Time Spent: 4h 50m
>  Remaining Estimate: 0h
>
> This will bring eleven bug fixes.
>  * ORC 1.5.7: [https://issues.apache.org/jira/projects/ORC/versions/12345702]
>  * ORC 1.5.8: [https://issues.apache.org/jira/projects/ORC/versions/12346462]



--
This message was sent by Atlassian Jira
(v8.3.4#803005)


[jira] [Commented] (HIVE-24316) Upgrade ORC from 1.5.6 to 1.5.8 in branch-3.1

2021-08-24 Thread Dongjoon Hyun (Jira)


[ 
https://issues.apache.org/jira/browse/HIVE-24316?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17403902#comment-17403902
 ] 

Dongjoon Hyun commented on HIVE-24316:
--

cc [~omalley]

> Upgrade ORC from 1.5.6 to 1.5.8 in branch-3.1
> -
>
> Key: HIVE-24316
> URL: https://issues.apache.org/jira/browse/HIVE-24316
> Project: Hive
>  Issue Type: Bug
>  Components: ORC
>Affects Versions: 3.1.3
>Reporter: Dongjoon Hyun
>Assignee: Dongjoon Hyun
>Priority: Major
>  Labels: pull-request-available
> Fix For: 3.1.3
>
>  Time Spent: 4h 50m
>  Remaining Estimate: 0h
>
> This will bring eleven bug fixes.
>  * ORC 1.5.7: [https://issues.apache.org/jira/projects/ORC/versions/12345702]
>  * ORC 1.5.8: [https://issues.apache.org/jira/projects/ORC/versions/12346462]



--
This message was sent by Atlassian Jira
(v8.3.4#803005)


[jira] [Commented] (HIVE-24316) Upgrade ORC from 1.5.6 to 1.5.8 in branch-3.1

2021-08-24 Thread Panagiotis Garefalakis (Jira)


[ 
https://issues.apache.org/jira/browse/HIVE-24316?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17403937#comment-17403937
 ] 

Panagiotis Garefalakis commented on HIVE-24316:
---

Hey [~glapark] thanks for bringing this up -- taking a look at 
MemoryManagerImpl looks like checkMemory() is the new method that determines if 
the scale has changed and since ORC-361 removed getTotalMemoryPool() calls from 
multiple places we are loosing the effect of controlling the memory pool.

The intention behind  LlapAwareMemoryManager  was to have memory per executor 
instead of the entire heap since multiple writers are involved. An idea could 
be to restore getTotalMemoryPool calls where needed .

> Upgrade ORC from 1.5.6 to 1.5.8 in branch-3.1
> -
>
> Key: HIVE-24316
> URL: https://issues.apache.org/jira/browse/HIVE-24316
> Project: Hive
>  Issue Type: Bug
>  Components: ORC
>Affects Versions: 3.1.3
>Reporter: Dongjoon Hyun
>Assignee: Dongjoon Hyun
>Priority: Major
>  Labels: pull-request-available
> Fix For: 3.1.3
>
>  Time Spent: 4h 50m
>  Remaining Estimate: 0h
>
> This will bring eleven bug fixes.
>  * ORC 1.5.7: [https://issues.apache.org/jira/projects/ORC/versions/12345702]
>  * ORC 1.5.8: [https://issues.apache.org/jira/projects/ORC/versions/12346462]



--
This message was sent by Atlassian Jira
(v8.3.4#803005)