View Single Post

   
  #2 (permalink)  
Old 02-26-2008, 09:46 AM
Sean McKeough
 
Posts: n/a
Default Re: INITFENCED_JVM

There has never bee a way to avoid jvm init time, because this var was
never enabled. :-)

We considered adding in some more knobs for this, but now that a typical
instance will have only 1 jvm running in it (using threads that are
shared between apps), the jvm startup will only happen 1x. So we didn't
consider it a big deal.

Jean-Marc Blaise wrote:
> Thanks Sean for these precisions.
>
> However, how can I overcome JVM starting time now in V8 ?
>
> I did from the same session:
> db2start
> call SQLJproc (the db2fmp is started with JVM)
> call JDBC1proc (the same db2fmp is used)
> call JDBC2proc ;another db2fmp is started with JVM: ==> why, I would have
> expected the same db2fmp to be used (I am in a multi-partition context on
> Linux with 3 nodes) ?
>
> Rem: all procs have been built with the DecCenter.
>
> Couldn't you have turned the INITFENCED_JVM from a YES/NO value to an
> INTEGER, to overcome this JVM loading time ?
>
> Thanks for your help,
>
> Jean-Marc
>
>
> "Sean McKeough" <mckeough@nospam.ca.ibm.com> a écrit dans le message news:
> bdf0f3$qhq$1@hanover.torolab.ibm.com...
>
>>Jean-Marc,
>>
>>INITFENCED starts a pool of db2fmps that will be used for not
>>threadsafe, non Java routines. In v8 with threading, we started
>>isolating different languaged routines from each other to prevent
>>interop problems (and to keep 'safer' languages isolated), as well as to
>>prevent memory bloat due to proliferation of JVMs. Java routines will
>>only go to fmps that have a JVM started.
>>
>>I think this is documented in the admin guide config parm section.
>>
>>Jean-Marc Blaise wrote:
>>
>>>Hi again, Mick (it's a pleasure !),
>>>
>>>Thanks for confirming if it's deprecated. From my testing, INITFENCED=1

>
> and
>
>>>INITFENCED_JVM=YES does not start db2fmp processes sizing the ones I

>
> find
>
>>>when I launch the 1st time a java proc (ps -elf | grep db2fmp still

>
> about
>
>>>5666 against about 22222 bytes).
>>>
>>>The first Eclipse link works. The 2nd one seems not to be accessible.
>>>
>>>Cheers,
>>>
>>>Jean-Marc
>>>
>>>"Mick Legare" <mlegare@no-spam-ca.ibm.com> a écrit dans le message news:
>>>3EF88D6B.768B5C93@no-spam-ca.ibm.com...
>>>
>>>
>>>>Hi again Jean-Marc,
>>>>I seem to remember that this parameter has been deprecated in V8 and
>>>
>>>should not
>>>
>>>
>>>>really show up anywhere. I'm going to check here and make sure.
>>>>
>>>>Also, regarding out last e-mails: I found the following links to our
>>>
>>>documentation
>>>
>>>
>>>>and they contain excellent search capabilities. I hope they help:
>>>>
>>>>in the Eclipse-based DB2 information center:
>>>> http://publib.boulder.ibm.com/infocenter/db2help
>>>>
>>>>in the HTML-based DB2 information center that gets installed with the

>
> doc
>
>>>CD
>>>
>>>
>>>>visible at:
>>>> http://rossr.torolab.ibm.com/db2doc/...htmlcd/en_US/).
>>>>
>>>>Mick
>>>>
>>>>
>>>>Blair Kenneth Adamache wrote:
>>>>
>>>>
>>>>
>>>>>This is just the new new name for INITDARI_JVM:
>>>>>
>>>>>Initialize DARI Process with JVM (initdari_jvm)
>>>>>
>>>>>Configuration Type
>>>>>Database manager
>>>>>
>>>>>Applies to
>>>>>Database server with local and remote clients
>>>>>Database server with local clients
>>>>>Partitioned database server with local and remote clients
>>>>>Satellite database server with local clients
>>>>>
>>>>>Parameter Type
>>>>>Configurable
>>>>>
>>>>>Default [Range]
>>>>>No [ Yes; No ]
>>>>>
>>>>>Related Parameters
>>>>>Maximum Number of DARI Processes (maxdari)
>>>>>Initial Number of Fenced DARI Processes in Pool (num_initdaris)
>>>>>Keep DARI Process Indicator (keepdari)
>>>>>This parameter indicates whether each fenced DARI process will load the
>>>
>>>Java
>>>
>>>
>>>>>Virtual Machine (JVM) when starting. This parameter will reduce the
>>>
>>>initial
>>>
>>>
>>>>>startup time for fenced Java stored procedures, especially when used in
>>>>>conjunction with the num_initdaris parameter. This parameter could
>>>
>>>increase the
>>>
>>>
>>>>>initial load time for non-Java fenced stored procedures as they do not
>>>
>>>require
>>>
>>>
>>>>>the JVM.
>>>>>
>>>>> Jean-Marc Blaise wrote:
>>>>>
>>>>>
>>>>>
>>>>>>Hi, I would like to know if the INITFENCED_JVM option in the DBM CFG
>>>
>>>is
>>>
>>>
>>>>>>functional, right now if V8 FP2, or if it is a future enhancement,
>>>
>>>because I
>>>
>>>
>>>>>>have not found any documentation in the Information Center (FP2) ?
>>>>>>
>>>>>>Thanks in advance,
>>>>>>
>>>>>>Jean-Marc
>>>>
>>>

>
>


Reply With Quote