Retry reading of S5p data, sometimes fails for unknown reason. (f7c048ed) · Commits · CAMS / CSO

py/cso_s5p.py

+82 −23

Original line number	Diff line number	Diff line
		@@ -189,6 +189,7 @@ class S5p_File( object ) :
		# modules:
		import os
		import xarray
		import time

		# split:
		if '/' in path :
		@@ -199,6 +200,14 @@ class S5p_File( object ) :
		group,vname = None,path
		#endif

		# somtimes fails ..
		maxtry = 5
		ntry = 0
		while True :
		# increase counter:
		ntry += 1
		# try to read:
		try :
		# access group:
		with xarray.open_dataset( self.filename, group=group ) as ds :
		# check ..
		@@ -210,7 +219,24 @@ class S5p_File( object ) :
		da = ds[vname]
		# not lazy ...
		da.load()
		#endwith # xarray
		#endwith # ds
		# nothing wrong, leave try loop
		break
		except :
		# again?
		if ntry <= maxtry :
		# info ...
		logging.warning( 'attempt %i of %i to read group "%s" from "%s", try again ...' % (ntry,maxtry,group,self.filename) )
		# wait ..
		time.sleep(5)
		# try again:
		continue
		else :
		logging.error( 'tried to read %i times, break ..' % maxtry )
		raise Exception
		#endif
		#endtry # read group
		#endwhile # try to read

		# unit stuff?
		if units is not None :
		@@ -1574,12 +1600,6 @@ class CSO_S5p_File( cso_file.CSO_File ) :
		avk[:,k,iretr] = pdat['avk'][:,k] * pdat['amf'] / pdat['amft']
		#endfor

		# check ...
		if sfile.nlayer is None :
		logging.error( 'no layer dimension defined yet' )
		raise Exception
		#endif

		# is troposphere layer defined ?
		key = 'troplayer'
		opath = rcf.get( vkey+'.'+key )
		@@ -2030,8 +2050,46 @@ class CSO_S5p_Convert( utopya.UtopyaRc ) :
		# combine:
		output_filename = os.path.join( outdir, outfile )

		# create file?
		if not os.path.isfile(output_filename) :
		# info ...
		logging.info( ' create new file ...' )
		# always create:
		create = True
		#~ renew?
		elif renew :
		# info ...
		logging.info( ' renew file ...' )
		# always create:
		create = True
		#~ already a version present:
		elif os.path.isfile(output_filename) :
		# there might be duplicated processings, with different processing times;
		# filenames are sorted, so current is probably newer
		# open existing file:
		csf = CSO_S5p_File( output_filename )
		# original orbit as attribute?
		orbit_file = csf.GetAttr( 'orbit_file', quiet=True )
		if orbit_file is None :
		# info ...
		logging.info( ' replace existing file, no "orbit_file" attribute yet ...' )
		# attribute not found, re-create to have this attribute:
		create = True
		elif fname <= orbit_file :
		# info ...
		logging.info( ' keep existing file, based on same or newer source file ...' )
		# same file, no need to re-create:
		create = False
		else :
		# info ...
		logging.info( ' replace existing file, based on older processing: %s' % orbit_file )
		# replace:
		create = True
		#endif
		#endif

		# create?
		if (not os.path.isfile(output_filename)) or renew :
		if create :

		# initialize list of temporary files, removed at end:
		tmpfiles = []
		@@ -2083,6 +2141,7 @@ class CSO_S5p_Convert( utopya.UtopyaRc ) :
		history.append( 'added %i pixels from %s' % (nselected,os.path.basename(fname)) )
		# update attributes:
		attrs['orbit' ] = orbit
		attrs['orbit_file'] = fname
		# write:
		csf.Write( filename=output_filename, attrs=attrs, history=history )

		@@ -2143,7 +2202,7 @@ class CSO_S5p_Listing( utopya.UtopyaRc ) :
		! - orbit number
		<rcbase>.file : /Scratch/CSO/S5p/RPRO/NO2/Europe/listing.csv

		An existing listing files is not replaced,
		An existing listing file is not replaced,
		unless the following flag is set::

		! renew table?
		@@ -2239,7 +2298,7 @@ class CSO_S5p_Listing( utopya.UtopyaRc ) :
		# remove entries that do not exist anymore:
		listing.Cleanup( indent=indent+' ' )

		# patternectory templates:
		# filename pattern templates:
		pattern_templates = self.GetSetting( 'patterns' ).split()
		# collection of scanned patterns:
		patterns = []
		@@ -2279,7 +2338,7 @@ class CSO_S5p_Listing( utopya.UtopyaRc ) :
		fnames.sort()
		# loop over files:
		for fname in fnames :
		# relative path to basedir:
		# absolute path:
		filename = os.path.join(bdir,fname)
		# already in table?
		if fname in listing :