ibatis-user-java mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mario Ds Briggs <mario.bri...@in.ibm.com>
Subject Re: Idle Threads - Glassfish/DB2
Date Wed, 15 Apr 2009 11:07:25 GMT
Did you do a compare of the JCC log for the same application run with
  a - the glassfish connection pool
  b - the ibatis SIMPLE datasource

thanks
Mario



                                                                           
             "Jeff Hibbs"                                                  
             <jhibbs@bop.gov>                                              
                                                                        To 
             10/04/2009 22:20          <user-java@ibatis.apache.org>       
                                                                        cc 
                                                                           
             Please respond to                                     Subject 
             user-java@ibatis.         Re: Idle Threads - Glassfish/DB2    
                apache.org                                                 
                                                                           
                                                                           
                                                                           
                                                                           
                                                                           




Hi Nathan -

Thanks for the suggestion...as you suspected, it works fine in the test env
with the SIMPLE datasource that comes with ibatis.  While this does give us
more information, at this point I still have to use the Glassfish
connection pools in production, as ours is one of many apps that reside on
these (cluster) servers, and we must conform to the standard Connection
Pools.

Any additional insight into these questions would be greatly
appreciated....

1. Why are the iBATIS queries keeping idle threads open on DB2 while the
straight JDBC coded queries are not.
2. Why does this only appear to happen with Glassfish Connection Pools?
Again, thanks for you help!  Jeff

>>> Nathan Maves <nathan.maves@gmail.com> 4/3/2009 10:29 AM >>>
Jeff,

Not to state the obvious but it does not seem like an ibatis issue. It
sounds more like the connection pool implementation of Glassfish. As a test
you might try creating your own connection pool using the SIMPLE datasource
type in ibatis just to ensure that it is an issue with Glassfish.

Nathan

On Fri, Apr 3, 2009 at 8:15 AM, Jeff Hibbs <jhibbs@bop.gov> wrote:
   Hello All -
   Any help will be greatly appreciated...Thanks!!!
   Our iBatis-based application was running on Sun1 Server/DB2 Version 8 -
   z/OS with no problems. When we migrated to Glassfish V2, the DB folks
   noticed many idle threads coming from our application which uses iBATIS
   2.1.5 (July 2005 Build). Other (non-iBATIS) applications that use
   straight JDBC (no ORM) on the same server, using the same connection
   pool, were not causing idle threads. Below is a sample what the DBA is
   seeing:
   —---------------------------------------------------------------------------------------------------------------------
   Primauth Planname name ID Status elapsed time CPU time
   xxxxxxxxx DISTSERV SYSLN100 SERVER *DB2 5:23.78195 0.000969
   xxxxxxxxx DISTSERV SYSLN100 SERVER *DB2 5:23.67919 0.001146
   xxxxxxxxx DISTSERV SYSLN100 SERVER *DB2 5:23.59251 0.000896
   xxxxxxxxx DISTSERV SYSLN100 SERVER *DB2 5:18.40476 0.001567
   xxxxxxxxx DISTSERV SYSLN100 SERVER *DB2 5:18.38349 0.001066

   14.46.15 STC12568 DSNL028I #J3P1 GAD00841.K6FE.C3F92EF69C21=157421 914
   914 ACCESSING DATA FOR
   914 LOCATION xx.xxx.x.xx
   914 IPADDR xx.xxx.x.xx
   14.48.14 STC12568 DSNL027I #J3P1 SERVER DISTRIBUTED AGENT WITH 561
   561 LUWID=GAD00840.PC1B.C3F92F10E401=157523
   561 THREAD-INFO=xxxxxx:genie4:xxxxxxx:db2jcc_applic
   561 RECEIVED ABEND=04E
   561 FOR REASON=00D3003B
   14.48.14 STC12568 DSNL027I #J3P1 SERVER DISTRIBUTED AGENT WITH 562
   562 LUWID=GAD00840.PC20.C3F92F1B5DDF=157544
   562 THREAD-INFO=xxxxxxx:genie4:xxxxxxx:db2jcc_applic
   562 RECEIVED ABEND=04E
   562 FOR REASON=00D3003B
   —-----------------------------------------------------------------------------------------------------------------------
   I'm not going to pretend to know what all this means, but apparently
   iBATIS/Glassfish is not releasing the threads after the SQL completes.
   Again, other non-iBATIS applications using the same connection pool are
   not generating these ilde threads. From a user's perspective the system
   is running fine - the queries are returning quickly. Also, we are not
   exhausting the connections in the connection pool, but apparently some
   resources in DB2 are incorrectly being left open. I guess I'm not sure
   of the difference between a "connection" and a "thread" from the DB2
   perspective.
   We have been able to replicate this in the Test env. Here's what we know
   so far:
   - Tried iBATIS 2.3.3.720: same results
   - Used replaced glassfish with Tomcat and the problem went away
   Obvious questions:
   1. Why are the iBATIS queries keeping idle threads open on DB2 while the
   straight JDBC coded queries are not.
   2. Why does this only appear to happen with Glassfish?
   Here's our iBATIS config:
   <settings
   useStatementNamespaces="false"
   cacheModelsEnabled="true"
   enhancementEnabled="true"
   />

   <transactionManager type="JDBC" >
   <dataSource type="JNDI">
   <property name="DataSource" value="java:comp/env/@isds.datasource.name
   @"/>
   </dataSource>
   </transactionManager>
   .......
   TEST Connection Pool Info:
   Datasource Classname: com.ibm.db2.jcc.DB2DataSource (prod same)
   Resource Type:javax.sql.DataSource (prod same)

   Pool Settings:
   Initial and Minimum Pool Size:8 (prod = 0)
   Maximum Pool Size: 32 (prod = 300)
   Pool Resize Quantity: 2 (prod = 5)
   Idle Timeout: 300 (prod = 15)
   Max Wait Time:60000 (prod = 60000)

Mime
View raw message