Open billsacks opened 7 years ago
I think we need to refactor the short term archiver to not have hard-wired suffix assumptions in the code - and that this needs to be data driven.
On Wed, Sep 27, 2017 at 3:48 PM, Bill Sacks notifications@github.com wrote:
The logic in case_st_archive.py: _archive_history_files does not handle the possibility that we could have an mpas component (which has a different pattern for history file names) in a multi-instance run. @kdraeder https://github.com/kdraeder changed this logic in #1916 https://github.com/ESMCI/cime/pull/1916 but the fundamental problem still remains. In that PR, this is noted explicitly:
if ninst_string: if compname.find('mpas') == 0: # Not correct, but MPAS' multi-instance name format is unknown. newsuffix = compname + '.*' + suffix
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ESMCI/cime/issues/1931, or mute the thread https://github.com/notifications/unsubscribe-auth/AHlxE_388PZ8WxqmX4CtNzawpqyLhNEoks5smsLCgaJpZM4PmceB .
Do we have any requirements for component file name formats? If so they should be clearly stated in the documentation.
I don't think this is a documentation issue. I think if the short term archiver xml files were actually distributed - then it would be up to each component to do this. As a first step - I think we need to move the current hard-wired code into the existing xml file and remove the hard wiring and then proceed to distribute it to the component directories.
On Thu, Sep 28, 2017 at 7:53 AM, jedwards4b notifications@github.com wrote:
Do we have any requirements for component file name formats? If so they should be clearly stated in the documentation.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/ESMCI/cime/issues/1931#issuecomment-332843182, or mute the thread https://github.com/notifications/unsubscribe-auth/AHlxE2AFbot-OAyQIY0OojwdfjNqMi8Kks5sm6TLgaJpZM4PmceB .
I seem to remember @rljacob saying on a recent cime call that mpas was open to changing its file naming conventions to match what's done for the other components, in which case this issue would go away. @rljacob am I remembering that right?
I don't think expecting component models to necessarily always change the file naming convention is the right long term solution. Some minimum standard needs to be kept - but the current implementation needs to be made more general and data driven rather than hard-wired code assumptions.
On Thu, Sep 28, 2017 at 7:59 AM, Bill Sacks notifications@github.com wrote:
I seem to remember @rljacob https://github.com/rljacob saying on a recent cime call that mpas was open to changing its file naming conventions to match what's done for the other components, in which case this issue would go away. @rljacob https://github.com/rljacob am I remembering that right?
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/ESMCI/cime/issues/1931#issuecomment-332844978, or mute the thread https://github.com/notifications/unsubscribe-auth/AHlxExXgBW5NagDsrpUTwILXvKNKsYIAks5sm6YwgaJpZM4PmceB .
Yes MPAS is open to that.
It's not clear to me what the next step on this is?
The short term archiver should use the regex expressions as defined in config_archive.xml as file name keys - is the regex for mpas currently correct in config_archives.xml? I'm working on adding tests for st_archive so I can include this in my branch.
This issue is stale because it has been open 90 days with no activity. Remove stale label or comment or this will be closed in 5 days.
This issue was closed because it has been stalled for 5 days with no activity.
The logic in case_st_archive.py:
_archive_history_files
does not handle the possibility that we could have an mpas component (which has a different pattern for history file names) in a multi-instance run. @kdraeder changed this logic in #1916 but the fundamental problem still remains. In that PR, this is noted explicitly: