Hide Forgot
The condor_job_server should be preferred by users for job submission data and job history (which is otherwise unavailable through the schedd plugin submission publishing). The plugin publisher can only provide full job data for live jobs still in the queue. Thus, we should ensure the the default qmf condor config activates the JS OOTB.
Customer considerations and impact: The current OOTB default is provided in the QMF schedd plug-in code as QMF_PUBLISH_SUBMISSIONS = True Also, the base configuration does not add the JOB_SERVER to the DAEMON_LIST although it is defined. Recommend the following: 1) Code change within the condor_job_server to halt activation if QMF_PUBLISH_SUBMISSIONS = True This will avoid confusion from having two *redundant* job servers active in the same QMF object space. 2) Modify 60condor-qmf.config to add/uncomment: QMF_PUBLISH_SUBMISSIONS = False DAEMON_LIST = $(DAEMON_LIST), JOB_SERVER 3) Release Note "The default configuration of the QMF Job Server used by cumin has been changed. The previous default was to use an embedded QMF Job Server object managed within a schedd plug-in component. The default configuration now launches a standalone daemon (condor_job_server) and disables the submission publishing from within the plug-in's embedded Job Server. Grid users using the old default should note that with the new default configuration, job details will now be available for jobs that have been completed or removed from the queue." 4) Possibly revise section of MCIG <4.1.1. Job Server Configuration> to note or emphasize the new default.
No base-db changes required since there is no feature that specifically activates the current default of QMF_PUBLISH_SUBMISSIONS = True.
Validation techniques: - 1 jobserver object listed via qpid-tool where only 1 schedd has been deployed - jobserver QMF object contains a "jobserver" prefix to schedd host, not "scheduler" - full job ads can be retrieved using QMF from a job that has been committed to history (confirmed using condor_history)
Technical note added. If any revisions are required, please edit the "Technical Notes" field accordingly. All revisions will be proofread by the Engineering Content Services team. New Contents: Release Note: "The default configuration of the QMF Job Server used by cumin has been changed. The previous default was to use an embedded QMF Job Server object managed within a schedd plug-in component. The default configuration now launches a standalone daemon (condor_job_server) and disables the submission publishing from within the plug-in's embedded Job Server. Grid users using the old default should note that with the new default configuration, job details will now be available for jobs that have been completed or removed from the queue."
(In reply to comment #5) > Validation techniques: > - 1 jobserver object listed via qpid-tool where only 1 schedd has been > deployed > - jobserver QMF object contains a "jobserver" prefix to schedd host, not > "scheduler" > - full job ads can be retrieved using QMF from a job that has been committed > to history (confirmed using condor_history) Everything seems ok, only one thing I don't understand, what does it mean jobserver QMF object contains a "jobserver" prefix to schedd host, not "scheduler"? I compare schemas in qpid-tool on old version with schema on new version and it seems the same for me.
Clarification and elaboration of comment #5: The QMF implementation allows the schedd plugin to act as both a "scheduler" and a "jobserver". In this scenario, its QMF object id should contain the string "scheduler" in it anywhere. In the opposite scenario, you will not see that. %qpid-tool Management Tool for QPID qpid: list jobserver Management Object Types: ObjectType Active Deleted ================================== com.redhat.grid:jobserver 1 0 qpid: show com.redhat.grid:jobserver
Tested with: condor-qmf-7.8.8-0.4 Tested on: RHEL6 i386,x86_64 RHEL5 i386,x86_64 second point verification: The scheduler name is not present in the show command SCHEDD_NAME = scheduler@$(FULL_HOSTNAME) schedd plugin: qpid: show 214 ... Name scheduler@host jobserver process: qpid: show 236 ... Name host >>> verified
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. http://rhn.redhat.com/errata/RHSA-2013-0564.html