LCG-Utils (lcg-cp) has been demonstrated to work successfully at AGLT2 with space tokens (see e.g. job 13714720). For these tests, the following schedconfig fields were used:
copytool: "lcg-cp2"
envsetup: "source /afs/atlas.umich.edu/OSGWN2/setup.sh"
setokens: "ATLASDATADISK,ATLASMCDISK"
se: "token:ATLASMCDISK:srm://head01.aglt2.org:8443/srm/managerv2?SFN="
seprodpath: "/pnfs/aglt2.org/[atlasmcdisk,atlasdatadisk]"
sepath: "/pnfs/aglt2.org/atlasenduserdisk"
The time out mechanism has been replaced by a new and better module (TimedCommand) written by Charles Waldman (see e.g. timed out job 13489501). The 24a version only has this implemented for the copy commands in SiteMover (cp), dCacheSiteMover (dccp), and dCacheLFCSiteMover (dccplfc). UberftpSiteMover (uberftp) was tested for time-outs but the tests were not completed before OU_OSCER_ATLAS went offline, so this site mover will not be added until later. The remaining site movers will get the new time-out module with the next release. The lcg-cp and lcg-cr copy commands used in lcgcpSiteMover and lcgcp2SiteMover understands time-outs natively.
Examples of finished jobs: BNL_ATLAS_test (dccp), TRIUMF_REPRO (dccplfc), BU_ATLAS_Tier2o (cp).
Version info:
Added new error codes for LFC import, 1113-4 (PilotErrors, RunJobUtilities, runJob, pilot, Mover,
lcgcpSiteMover, SiteMover)
Now setting pilotErrorDiag when there are no child processes (pilot)
Created PilotTCPServer class (PilotTCPServer, pilot)
Corrected indentation error in put function (dCacheSiteMover)
Rewrote put function in lcgcp2SiteMover based on dCacheSiteMover (lcgcp2SiteMover)
Corrected SRMv2 -> srmv2 symbol name in get/put functions (lcgcp2SiteMover)
Added protection against overwriting of already set error code by signal handler (runJob)
Created TimedCommand.py module (code from Charles)
Created removeSRMInfo() to remove any SRM info before LRC registration (pUtil)
Added command return message in getFileType (SiteMover)
Added new error code 1115 for NFS/SQLite locking problems (PilotErrors, runJob)
Added filter for "-signs in tolog() (pUtil)
Removed TimeOutModule from Mover
Added timed_command plus handling to get/put function (SiteMover, dCacheSiteMover, BNLdCacheSiteMover)
Protected against single AtlasRelease in multi-trf job def (RunJobUtilities, runJob)
Added timed_command plus handling to get function (only) in dCacheLFCSiteMover
No comments:
Post a Comment