Friday, February 27, 2009

UHURA 33a

Summary
* In the case the pilot can not read the athena metadata (for whatever reason, nfs problem etc) the pilot will now fail the job. Debugging is too hard/impossible since the error will only be visible in the batch log which in turn is not kept long enough (troublesome jobs are caught by Solveigs' scripts). The new pilot error code is 1187 (corresponding to proddb error code 440605, EXEPANDA_NOATHENAMETADATA ('Athena metadata is not available').

* It was discovered that the 'voms-proxy-info -exists -valid HH:MM' command is broken (it always returns true). A GGUS ticket was opened (https://gus.fzk.de/ws/ticket_info.php?ticket=45632) and we were told that the problem has been fixed in the version under certification. Awaiting the release of this version, the pilot proxy verification has been updated to instead use the -actimeleft option instead (Graeme).

* The dq2SiteMover using dq2-get/put has been updated and is currently being verified at various sites (Mario).

* The job definition dispatchDBlockToken field can now be used to specify space tokens for the input files (should be a list of equal length to the input file list). If set to a non-NULL and non-empty value, the pilot will sort the replica list (for each guid) accordingly. Requested by Rod to alleviate the March reprocessing - tape residing replicas should in this case be chosen before disk residing replicas since we don't want to delete the existing disk versions. Normally, disk residing replicas are chosen before any tape replicas.


Version info
Now using option -actimeleft with voms-proxy-info since command is otherwise broken (code from Graeme, SiteMover)
Removed command output in case voms-proxy-info is not available (SiteMover)
Printing LD_LIBRARY_PATH prior to job execution (runJob)
Out-commented useless code for Nordugrid in updatePandaServer (pilot)
Added new pilot error code 1187 (PilotErrors)
Now failing job if athena metadata is not available (skipped for analysis and panda mover jobs, requested by Solveig, pilot)
Added handling of dispatchDBlockToken (Job)
dq2-get/put now setting client id to pilot, dsname properly handled by put function (dq2SiteMover)
get_data() is now using dispatchDBlockToken (RunJobUtilities)
Token handling now enabled in mover_get_data() and getPoolFileCatalog() (Mover)
sortReplicas() now matches space token to replica if necessary/possible (Mover)
Created pullMatchedAreas() used to pull forward a given area in the se list (SiteMover)
Some cleanup of old debugging messages in lfc_mkdir() (SiteMover)
subDir has been removed from CMTPATH for trfs of the form AnalysisTransforms-([^/]+) (RunJobUtilities)
Now only using a short proxy limit for pandadev test jobs (pilot, pUtil)

No comments: