Mailing-List: contact dev-help@hive.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hive.apache.org
Content-Type: multipart/alternative;
 boundary="===============8188639404099720634=="
MIME-Version: 1.0
Subject: Re: Review Request 59325: Cleanup of structures required when LLAP
 access from external clients completes
From: Siddharth Seth <sseth@apache.org>
To: Sergey Shelukhin <sergey@hortonworks.com>,
 Jason Dere <jdere@hortonworks.com>
Cc: hive <dev@hive.apache.org>, Siddharth Seth <sseth@apache.org>
Date: Wed, 17 May 2017 21:58:44 -0000
Message-ID: <20170517215844.55001.46502@reviews-vm2.apache.org>
Auto-Submitted: auto-generated
Sender: Siddharth Seth <noreply@reviews.apache.org>
References: <20170517020251.61684.14018@reviews-vm2.apache.org>
In-Reply-To: <20170517020251.61684.14018@reviews-vm2.apache.org>
Reply-To: Siddharth Seth <sseth@apache.org>
archived-at: Wed, 17 May 2017 21:58:48 -0000

--===============8188639404099720634==
MIME-Version: 1.0
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit


> On May 17, 2017, 2:02 a.m., Sergey Shelukhin wrote:
> > llap-server/src/java/org/apache/hadoop/hive/llap/daemon/impl/QueryTracker.java
> > Lines 464 (patched)
> > <https://reviews.apache.org/r/59325/diff/1/?file=1722071#file1722071line483>
> >
> >     will there be a callable created after every single fragment that finishes for the query with no other fragments running? Perhaps one callable should wait for the entire query and the fragments coming it should just push back the timestamp at which the query would time out and be cleaned?
> >     
> >     also what if readLock blocks all cleanup attempts?
> 
> Siddharth Seth wrote:
>     In terms of the callable - was trying to keep it simple. Can try making use of a single callable.
>     
>     Why would a readLock block all cleanup attempts? If I'm not mistaken, a readLock means something else is running - which would cause another cleanup to be scheduled.

Haven't made the Callable change here. Made small improvements to when cleanup runs (it will not create a new lock).
Will file a follow up for the single Callable. Think it adds a bit of code since we can't have a thread wait, so it'll really be a one callable optionally schedules the next callable etc.


- Siddharth


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/59325/#review175184
-----------------------------------------------------------


On May 17, 2017, 9:56 p.m., Siddharth Seth wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/59325/
> -----------------------------------------------------------
> 
> (Updated May 17, 2017, 9:56 p.m.)
> 
> 
> Review request for hive, Jason Dere and Sergey Shelukhin.
> 
> 
> Bugs: HIVE-14052
>     https://issues.apache.org/jira/browse/HIVE-14052
> 
> 
> Repository: hive-git
> 
> 
> Description
> -------
> 
> Cleanup of structures required when LLAP access from external clients completes
> 
> 
> Diffs
> -----
> 
>   llap-common/src/gen/protobuf/gen-java/org/apache/hadoop/hive/llap/daemon/rpc/LlapDaemonProtocolProtos.java ece31ed 
>   llap-common/src/protobuf/LlapDaemonProtocol.proto 3a3a2b8 
>   llap-server/src/java/org/apache/hadoop/hive/llap/daemon/impl/QueryInfo.java ce2f457 
>   llap-server/src/java/org/apache/hadoop/hive/llap/daemon/impl/QueryTracker.java daeb555 
>   llap-server/src/test/org/apache/hadoop/hive/llap/daemon/impl/TaskExecutorTestHelpers.java 27c426c 
>   ql/src/java/org/apache/hadoop/hive/ql/udf/generic/GenericUDTFGetSplits.java 868eec7 
> 
> 
> Diff: https://reviews.apache.org/r/59325/diff/2/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Siddharth Seth
> 
>


--===============8188639404099720634==--