Note: This web site is only kept up to date for OSG Software
1.2 (VDT 2.0.0). If you are looking for information for the most recent
release, the RPM-based OSG Software 3.0, please see
the OSG documentation web site
Globus patches in the VDT
We patch Globus when we build it for the VDT.
Building Globus 5.0.0
To build Globus for VDT 5.0.0, we check out Globus 5.0.0 using the
globus_5_0_0 tag then we apply the patches that
You can download the tarball with the complete set of patches
Please note: We know that these patches apply cleanly, but we
have not yet built Globus with them, and certainly haven't yet
Patches for Globus 5.0.0
Enhancement by EDG/LCG: "gridmapdir" functions allow a pool of pre-created
anonymous accounts to be mapped to real users (DNs) at run-time, without
further intervention by a sysadmin. The functionality is disabled and
does not affect any installations unless the environment variable
GRIDMAPDIR points to a directory that has been initialized with the
anonymous accounts (/etc/grid-security/gridmapdir).
||Improvements to the PBS job manager made by TeraGrid.
||In an OSG installation, set a job's GLOBUS_LOCATION based on
the OSG_GRID environment variable. This will not affect a non-OSG
I moved the log_to_graitia() as well. I don't know if that was the
right thing to do.
Three bug fixes:
- Don't remove the Condor user log when reporting the job as
FAILED but instead remove it when the job is being cleaned
up. Otherwise, subsequent polls will report the job as
DONE. This results in the client being told the job
completed when it was in fact removed. Also, return a
failure code instead of just a FAILED status. This tells
the client why the status is FAILED.
- When the perl poll function signals a status of FAILED, it
usually returns a failure code explaining why the job has
FAILED. The Condor grid monitor doesn't report this
failure code. When the jobmanager receives a FAILED status
from the grid monitor, have it run the poll function
itself to get the failure code.
- On restart, the jobmanager may ignore the latest job
status from the grid monitor, instead using the job status
written in the gram state file. It does this because it's
miscalculating when the job status in the gram state file