Opened 10 years ago
Closed 10 years ago
#195 closed defect (fixed)
Problem at ada when SpaceName=TEST
Reported by: | jgipsl | Owned by: | sdipsl |
---|---|---|---|
Priority: | critical | Milestone: | libIGCM_v2.3 metrics |
Component: | treatment | Version: | |
Keywords: | Cc: |
Description
At ada, setting SpaceName=TEST in config.card makes problem each time when the job is resubmitting. The restart files are not found.
Looking at following section in libIGCM_comp.ksh, beginning at line 826 (rev 1020 on trunk or tag v2.3)
eval Path_BUF=\${R_BUF_${comp}_R}/${generic_restart_file_name_in} eval Path_OUT=\${R_OUT_${comp}_R}/${generic_restart_file_name_in} if [ $( IGCM_sys_TestFileBuffer ${Path_BUF}*.${extension_in} ; echo $? ) = 0 ] ; then IGCM_debug_Print 3 "Buffered restart ${Path_BUF}*.${extension_in}" Buffered=true Archived=false Tared=false nb_restart_file=$(IGCM_sys_CountFileBuffer ${Path_BUF}_????.${extension_in}) elif [ $( IGCM_sys_TestFileArchive ${Path_OUT}*.${extension_in} ; echo $? ) = 0 ] ; then IGCM_debug_Print 3 "Archived restart ${Path_OUT}*.${extension_in}" Buffered=false Archived=true Tared=false nb_restart_file=$(IGCM_sys_CountFileArchive ${Path_OUT}_????.${extension_in}) else IGCM_debug_Print 3 "No restart file in the buffer nor in the archive directory" ....
Path_BUF is not usable since R_BUF_${comp}_R is not set. Path_OUT is correct indiquation the $WORKDIR at ada because of SpaceName=TEST. But as Path_OUT is a path at ada and not archive machine so the
elif [ $( IGCM_sys_TestFileArchive ${Path_OUT}*.${extension_in} ; echo $? ) = 0 ] ; then
cannot work. Using IGCM_sys_TestFileBuffer is appropriate for a path at $WORKDIR at ada. For keeping both cases, I propose changing the elif to the following :
elif [ $( IGCM_sys_TestFileArchive ${Path_OUT}*.${extension_in} ; echo $? ) = 0 ] || [ $( IGCM_sys_TestFileBuffer ${Path_OUT}*.${extension_in} ; echo $? ) = 0 ] ; then
This problem is not seen at curie since a file at $SCRATCHDIR can be accessed in the same way as a file at $STOREDIR.
Another solution might be to add definition of R_BUF_${comp}_R but this seems to have more implications and I havn't tested this.
Change History (4)
comment:1 Changed 10 years ago by jgipsl
comment:2 Changed 10 years ago by sdipsl
- Owner changed from somebody to sdipsl
- Status changed from new to assigned
For consistency it seems better and clearer to define buffered directories.
v2.2 don't suffer this because IGCM_sys_TestFileArchive were also testing buffered part (was not consistent with function name).
comment:3 Changed 10 years ago by sdipsl
- Milestone set to libIGCM_v2.3 metrics
comment:4 Changed 10 years ago by sdipsl
- Resolution set to fixed
- Status changed from assigned to closed
Done see r1022
Note that tag v2.2 do not have this bug.