Missing files generate a warning with htar, failure with tar

[DavidHuber-NOAA]

What is wrong?

Local ROTDIR archiving (LOCALARCH="YES") fails on the first half cycle as the gdasenkf log files for esfc, ecen, eobs, and eupd do not exist. These are requested for archiving on every cycle via

global-workflow/ush/hpssarch_gen.sh

Lines 615 to 617 in 4cb5802

	for fstep in eobs ecen esfc eupd efcs epos ; do
	echo "logs/${PDY}${cyc}/${RUN}${fstep}*.log"
	done

A warning is generated when archiving to HPSS, but tar generates a non-zero status.

What should have happened?

Either archiving of these files should not be attempted on the first half cycle or an error should not be generated by tar.

What machines are impacted?

Orion

Steps to reproduce

1. Generate an experiment with LOCALARCH="YES"
2. Run the first half enkfgdas cycle
3. Check the status of the enkfgdasearc00 job

Additional information

Also an issue for S4. It should also be mentioned that there are possibly other missing files that will cause tar to fail. I have not tested beyond the enkfgdasearc00 job.

Do you have a proposed solution?

Add a test inside of hpssarch_gen.sh similar to the one for eomg logs that checks for the existence of the enkfgdas logs.

[DavidHuber-NOAA]

Per a discussion with @WalterKolczynski-NOAA, missing files should generate errors in most cases and warnings in cases that a missing file could be an expected outcome (e.g. missing observational data resulting in the GSI-monitor not producing files for a given instrument). Thus, a wrapper function may be necessary to check for the existence of files before issuing errors or warnings and not rely on tar or htar to produce the desired effect.

[DavidHuber-NOAA]

The premise of this issue isn't true. htar does issue an error if a file is missing. I seem to have misinterpreted the logs. That said, there are cases where missing files are OK (e.g. #2076), but those special cases can be handled on a case-by-case basis. Closing.