Mailing-List: contact dev-help@hive.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hive.apache.org
Received-SPF: pass (nike.apache.org: domain of thejas.nair@gmail.com
 designates 209.85.220.173 as permitted sender)
MIME-Version: 1.0
In-Reply-To: <8B5F7C57-D42D-4A20-BF4E-0B8A5A838D9D@premierinc.com>
References: <8B5F7C57-D42D-4A20-BF4E-0B8A5A838D9D@premierinc.com>
Date: Fri, 5 Jun 2015 09:14:46 -0700
Message-ID: 
 <CA+mdTd27c0eZ_nMd_dpsP_wZK+9iOGN56OYnu9_JYoDABv6-Rw@mail.gmail.com>
Subject: Re: connection pooling for hive JDBC client
From: Thejas Nair <thejas.nair@gmail.com>
To: dev <dev@hive.apache.org>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

It is better to use re-use the hive jdbc connection if possible, as
there are costs associated with creating new connection. However,
there are known issues in running multiple queries simultaneously
using a single connection. For now, it woudl be safer to run queries
sequentially within a connection (ie, don't share it across threads
and run concurrent queries). (You can ofcourse run queries in parallel
using different connections).

Regarding connection pooling libraries, I am not sure if that has been
tested. There might be issues like that unsupported api call you
mentioned, that need to be fixed.


On Fri, Jun 5, 2015 at 6:50 AM, McWhorter, David
<David_McWhorter@premierinc.com> wrote:
> Hello, I sent this email to the users list a few days ago but no one ther=
e seems to be able to help, so I am trying the dev list.  I am working on a=
n application that queries and interacts with hive using the JDBC API.  In =
many other cases, using a JDBC connection pool such as commons-dbcp or Bone=
CP or HikariCP is a recommended practice and results in much better perform=
ance.  All of the examples I=E2=80=99ve found of using accessing Hive throu=
gh the JDBC API (such as https://cwiki.apache.org/confluence/display/Hive/H=
iveServer2+Clients#HiveServer2Clients-JDBCClientSampleCode) use a raw JDBC =
connection to Hive directly.  A few questions:
>
>   1.  Should hive work with connection pools such as commons-dbcp, BoneCP=
, or HikariCP?
>      *   Note: I have tried all BoneCP and HikariCP, and cannot seem to g=
et either to work because HiveConnection.setReadOnly throws a java.sql.SQLE=
xception saying =E2=80=9CMethod not supported=E2=80=9D.  I am using hive 0.=
14.0.
>   2.  Do you recommended using a JDBC connection pool for interacting wit=
h hive from an application that will execute many repeated and concurrent q=
ueries/statements?
>
> Also, just to be clear, I am not asking about configuring the Hive metast=
ore to use a connection pool to connect to its underlying database (MySQL, =
PostgreSQL, etc), but about using a connection pool to interact with and qu=
ery Hive through the JDBC api from an application.
>
> Thank you,
> David McWhorter
>
> =E2=80=94
>
> David McWhorter
> Senior Developer, Foundations
> Informatics and Technology Services
> Office: 434.260.5232 | Mobile: 434.227.2551
> david_mcwhorter@premierinc.com<mailto:david_mcwhorter@premierinc.com>  | =
 Premier, Inc. (NASDAQ: PINC)