Difference: CreamTestsP2748 (1 vs. 8)

• All the tests of the CREAM testsuite have been successfully executed except the following ones:

• Bugs #46024: Sometimes the AbstractJobExecutor throws the exception "Too many open files" HOPEFULLY FIXED

• • Kill the BLParserPBS process (e.g. kill -15 28232), and after some seconds you should see a new process

Automatic tests:

• All the tests of the CREAM testsuite have been successfully executed except the following ones:
  • test #4 which reports a 'missing cancellation' over 1000 jobs for unreproducible reason.

• Bugs #47655: Failure updating the main authN proxy FIXED
  • Verified with test #13 of the CREAM testsuite.

• Bugs #47700: glexec failures in CREAM because /tmp is used as working directory HOPEFULLY FIXED
  • Not easy to reproduce. The bug occured only on IN2P3 installations, the reasons are still unknown. This should be a blocking bug for every register operation.

• Bugs #47655: Failure updating the main authN proxy

• Bugs #47700: glexec failures in CREAM because /tmp is used as working directory

• Bugs #45983: BLAH can leave children processes behind. HOPEFULLY FIXED
  • Not easy to reproduce

• Bugs #46024: Sometimes the AbstractJobExecutor throws the exception "Too many open files" HOPEFULLY FIXED
  • This bug is not easy to reproduce, it happened only when the load of the CE was very high. In all the tests of heavy submissions through ICE we haven't see it.

• Bugs #46283: Possible memory leak in strtoken function for BLParser HOPEFULLY FIXED
  • Not easy to reproduce

• Bugs #46684: CREAM should call glexec directly (not via a wrapper script) HOPEFULLY FIXED
  • Changes inside the code.

• Bugs #47375: BLParser could fail in sending notification to cream HOPEFULLY FIXED
  • Not easy to reproduce

• Bugs #45718: Some check on log lines should be added on BLParser code HOPEFULLY FIXED
  • Changes inside the code.

• Bugs #45736: Problems in case of resubmissions in the same CREAM CE FIXED
  • Submit a job that will be resubmitted (e.g. Set Prologue = "/bin/false" in the jdl) and that will match only one resource (e.g. use option --resource, -r <ce_id> of the glite-wms-job-submit command)
  • When the job terminated check if the creamID changes after every resubmission using the command glite-job-logging-info -v (e.g. look at the filed Dest jobid in the Transfer event logged by LogMonitor)

• Bugs #45913: Proxy renewal not done for CREAM jobs not yet in IDLE status HOPEFULLY FIXED
  • You can verify it only with a large number of jobs, because it is a very rare condition. In all the tests of heavy submissions through ICE we haven't see it.

• Bugs #45437: Sometimes the jobPurger throws the exception "Too many open files" HOPEFULLY FIXED
  • This bug is not easy to reproduce, it happened only when the load of the CE was very high and there were a lot of jobs to purge. In all the tests of heavy submissions through ICE we haven't see it.

• Bugs #46916: blparser startup daemon FIXED
  • Log on a CE where the Blparser is supposed to run.
  • Using the command "ps ax | grep -i parser" you should see two processes:

    23221  /opt/glite/bin/blparser_master
    28232  /opt/glite/bin/BLParserPBS -d 1 -l /opt/glite/var/log/glite-pbsparser.log -s /var/spool/pbs -p 33333 -m 56565

  • Kill the BLParserPBS process (e.g. kill -15 28232), and after some seconds you should see a new process

• Bugs #47209: BL parser startup script assumes ${GLITE_LOCATION_VAR} exists FIXED
  • Check the file /opt/glite/etc/init.d/glite-ce-blparser to see is these lines have been removed:

    pidfile=$GLITE_LOCATION_VAR/glite-blparser${lrms}.pid
    echo `pidofproc ${exe}` > $pidfile 
    

Check bugs:

• Bugs #45437: Sometimes the jobPurger throws the exception "Too many open files"

• Bugs #45718: Some check on log lines should be added on BLParser code

• Bugs #45736: Problems in case of resubmissions in the same CREAM CE

• Bugs #45913: Proxy renewal not done for CREAM jobs not yet in IDLE status

• Bugs #45983: BLAH can leave children processes behind.

• Bugs #46024: Sometimes the AbstractJobExecutor throws the exception "Too many open files"

• Bugs #46283: Possible memory leak in strtoken function for BLParser

• Bugs #46684: CREAM should call glexec directly (not via a wrapper script)

• Bugs #46916: blparser startup daemon

• Bugs #47209: BL parser startup script assumes ${GLITE_LOCATION_VAR} exists

-- AlessioGianelle - 24 Feb 2009