I'm seeing some odd behaviour from our GT 4.0.7 pre-WS GRAM:
About 20% of the time, simple globus-job-run commands return no output
results via globus-job-get-output. When I investigate the gass cache in
the user account on the GRAM service, the time stamp of the problem runs
is always 1am 1st Jan 1970. I thought at first it might be a filesystem
bug, but by attaching strace to the globus-gatekeeper, it seems that
this is being done by both globus-job-manager and globus-gass-cache:
[Globus-jobmanager]
globus.24922:16:36:26.521592
utime("/scratch/pacey/.globus/.gass_cache/global/md5/9f/9077e96a391700f8
ded76c4a49b107/data.lancs2.nw-grid.ac.uk_615a_0", [0, 0]) = 0
globus.24922:16:36:26.698656
utime("/scratch/pacey/.globus/.gass_cache/global/md5/b3/2c4e3d78e3ab4a2b
6625f17bef5c9d/data.lancs2.nw-grid.ac.uk_615a_0", [0, 0]) = 0
[Gass_cache]
globus.24928:16:36:28.212814
utime("/scratch/pacey/.globus/.gass_cache/global/md5/f1/e350ce35d35d24b3
6a1044a3f5f7aa/data.lancs2.nw-grid.ac.uk_6160_0", [0, 0]) = 0
Is this expected behaviour? It seems more than coincidence that the
problem job runs always have this timestamp while successful runs have
normal timestamps.
Regards,
Mike.
-----
Dr Mike Pacey, Email: [EMAIL PROTECTED]
High Performance Systems Support, Phone: 01524 593543
Information Systems Services, Fax: 01524 594459
Lancaster University,
Lancaster LA1 4YW
