<?xml version="1.0" standalone="yes" ?>
<!DOCTYPE bugzilla SYSTEM "http://bugzilla.globus.org/bugzilla/bugzilla.dtd">

<bugzilla version="3.2.3"
          urlbase="http://bugzilla.globus.org/bugzilla/"
          maintainer="bacon@mcs.anl.gov"
>

    <bug>
          <bug_id>5511</bug_id>
          
          <creation_ts>2007-08-30 11:37</creation_ts>
          <short_desc>Review synchronization in MJFS.createManagedJob()</short_desc>
          <delta_ts>2007-11-30 22:55:11</delta_ts>
          <reporter_accessible>1</reporter_accessible>
          <cclist_accessible>1</cclist_accessible>
          <classification_id>1</classification_id>
          <classification>Unclassified</classification>
          <product>GRAM</product>
          <component>wsrf managed job factory service</component>
          <version>unspecified</version>
          <rep_platform>Macintosh</rep_platform>
          <op_sys>All</op_sys>
          <bug_status>CLOSED</bug_status>
          <resolution>FIXED</resolution>
          
          
          
          
          <priority>P3</priority>
          <bug_severity>normal</bug_severity>
          <target_milestone>4.0.6</target_milestone>
          
          
          
          <everconfirmed>1</everconfirmed>
          <reporter name="Martin Feller">feller@mcs.anl.gov</reporter>
          <assigned_to name="Martin Feller">feller@mcs.anl.gov</assigned_to>
          <cc>feller@mcs.anl.gov</cc>
    
    <cc>madduri@mcs.anl.gov</cc>
    
    <cc>smartin@mcs.anl.gov</cc>

      

      
          <long_desc isprivate="0">
            <who name="Martin Feller">feller@mcs.anl.gov</who>
            <bug_when>2007-08-30 11:37:02</bug_when>
            <thetext>The creation of a job resource is synchronized. There had been
some very rare issues where a service thread creating a managed
job resource kept stuck and blocked subsequent threads an caused
all available service threads to be consumed. This finally caused
the container to keep stuck.
This synchronization seems to be unnecessary. However, if we
remove the synchronization here we must take care that access to
the idempotence id map managed in MEJHome is synchronized properly.</thetext>
          </long_desc>
          <long_desc isprivate="0">
            <who name="Martin Feller">feller@mcs.anl.gov</who>
            <bug_when>2007-11-30 22:54:58</bug_when>
            <thetext>Access to the idempotence id map is properly synchronized.
I ran about 4-5 condor-g tests each with 1000 concurrent
jobs with removed synchronization without problems and
committed the change to the 4.0 branch.</thetext>
          </long_desc>
      
      

    </bug>

</bugzilla>