CRAB/python/PhEDExDatasvcInfo.py

from Actor import *
import urllib
from xml.dom.minidom import parse
from crab_exceptions import *
from WorkSpace import *
from urlparse import urlparse 
from LFNBaseName import *

class PhEDExDatasvcInfo:
    def __init__( self , cfg_params ):
 
        ## PhEDEx Data Service URL
        url="https://cmsweb.cern.ch/phedex/datasvc/xml/prod"
        self.datasvc_url = cfg_params.get("USER.datasvc_url",url)

        self.FacOps_savannah = 'https://savannah.cern.ch/support/?func=additem&group=cmscompinfrasup'
        stage_out_faq='https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCrabHowTo#Stageout_and_publication' 
        self.dataPub_faq = 'https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCrabForPublication'

        self.srm_version = cfg_params.get("USER.srm_version",'srmv2')
        self.node = cfg_params.get('USER.storage_element',None)
        

        self.user_lfn = cfg_params.get("USER.lfn",'')
        self.publish_data = cfg_params.get("USER.publish_data",0)
        self.usenamespace = cfg_params.get("USER.usenamespace",0)
        self.user_remote_dir = cfg_params.get("USER.user_remote_dir",'')
        if self.user_remote_dir:
            if ( self.user_remote_dir[-1] != '/' ) : self.user_remote_dir = self.user_remote_dir + '/'
        if self.user_lfn:    
            msg  = 'Warning: lfn has been deprecated, CRAB will ignore it.\n'
            msg += '\t Please use only user_remote_dir removing lfn from your crab.cfg\n'
            msg += '\t For further information please visit : \n\t%s'%stage_out_faq 
            common.logger.info(msg)
          
        self.datasetpath = cfg_params.get("CMSSW.datasetpath")
        self.publish_data_name = cfg_params.get('USER.publish_data_name','')

        self.user_port = cfg_params.get("USER.storage_port",'8443')
        self.user_se_path = cfg_params.get("USER.storage_path",'')
        if self.user_se_path:
            if ( self.user_se_path[-1] != '/' ) : self.user_se_path = self.user_se_path + '/'
                                                    
        #check if using "private" Storage
        self.usePhedex = True 
        if not self.node :
            msg = 'Please specify the storage_element name in your crab.cfg section [USER].\n'
            msg +='\tFor further information please visit : %s'%stage_out_faq 
            raise CrabException(msg)
        if (self.node.find('T1_') + self.node.find('T2_')+self.node.find('T3_')) == -3: self.usePhedex = False 

        if not self.usePhedex and ( self.user_remote_dir == '' or self.user_se_path == '' ):
            msg = 'You are asking to stage out without using CMS Storage Name convention. In this case you \n' 
            msg += '\t must specify both user_remote_dir and storage_path in the crab.cfg section [USER].\n '
            msg += '\t For further information please visit : \n\t%s'%stage_out_faq
            raise CrabException(msg)
        self.sched = common.scheduler.name().upper() 
        self.protocol = self.srm_version

        self.forced_path = '/store/user/'
        if self.sched in ['CAF','LSF']:
            self.protocol = 'direct'
            self.SE = {'CAF':'caf.cern.ch', 'LSF':''}
            if self.sched == 'CAF': self.forced_path = '/store/caf/user/'
            
        if not self.usePhedex: 
            self.forced_path = self.user_remote_dir
        return
 
    def getEndpoint(self):   
        '''
        Return full SE endpoint and related infos
        '''
        self.lfn = self.getLFN()
 
        #extract the PFN for the given node,LFN,protocol
        endpoint = self.getStageoutPFN()
   
        #extract SE name an SE_PATH (needed for publication)
        SE, SE_PATH, User = self.splitEndpoint(endpoint)

        return endpoint, self.lfn , SE, SE_PATH, User         
       
    def splitEndpoint(self, endpoint):
        '''
        Return relevant infos from endpoint  
        '''
        SE = ''
        SE_PATH = ''
        USER = ''
        if self.usePhedex: 
            if self.protocol == 'direct':
                query=endpoint
                SE_PATH = endpoint
                SE = self.SE[self.sched]
            else: 
                url = 'http://'+endpoint.split('://')[1]
                # python > 2.4
                # SE = urlparse(url).hostname 
                scheme, host, path, params, query, fragment = urlparse(url)
                SE = host.split(':')[0]
                SE_PATH = endpoint.split(host)[1]
            USER = (query.split('user')[1]).split('/')[1]
        else:
            #### to test #####
           # url = 'http://'+endpoint.split('://')[1]
           # scheme, host, path, params, query, fragment = urlparse(url)
           # SE = host.split(':')[0]
           # SE_PATH = endpoint.split(host)[1]
            SE = self.node
            SE_PATH = self.user_se_path + self.user_remote_dir 
            try:
                USER = (self.lfn.split('user')[1]).split('/')[1]
            except:
                pass

        return SE, SE_PATH, USER 
   

    def getLFN(self):
        """
        define the LFN composing the needed pieces
        """
        lfn = ''
        l_User = False
        if not self.usePhedex and (int(self.publish_data) == 0 and int(self.usenamespace) == 0) :
            ### add here check if user is trying to force a wrong LFN using a T2  TODO
            ## check if storage_name is a T2 (siteDB query)
            ## if yes :match self.user_lfn with LFNBaseName...
            ##     if NOT : raise (you are using a T2. It's not allowed stage out into self.user_path+self.user_lfn)   
            lfn = self.user_remote_dir
            return lfn
        if self.publish_data_name == '' and int(self.publish_data) == 1:
            msg = "Error. The [USER] section does not have 'publish_data_name'\n"
            msg += '\tFor further information please visit : \n\t%s'%self.dataPub_faq
            raise CrabException(msg)
        if self.publish_data_name == '' and int(self.usenamespace) == 1:
           self.publish_data_name = "DefaultDataset"
        if int(self.publish_data) == 1:
            if self.sched in ['CAF']: l_User=True 
            primaryDataset = self.computePrimaryDataset()
            ### added the case lfn = LFNBase(self.forced_path, primaryDataset, self.publish_data_name, publish=True)
            ### for the publication in order to be able to check the lfn length  
            lfn = LFNBase(self.forced_path, primaryDataset, self.publish_data_name, publish=True)  + '/${PSETHASH}/'    
        elif int(self.usenamespace) == 1:
            if self.sched in ['CAF']: l_User=True 
            primaryDataset = self.computePrimaryDataset()
            lfn = LFNBase(self.forced_path, primaryDataset, self.publish_data_name)  + '/${PSETHASH}/'    
        else:
            if self.sched in ['CAF','LSF']: l_User=True 
            lfn = LFNBase(self.forced_path,self.user_remote_dir)
        return lfn
 
    def computePrimaryDataset(self):
        """
        compute the last part for the LFN in case of publication     
        """
        if (self.datasetpath.upper() != 'NONE'):
            primarydataset = self.datasetpath.split("/")[1]
        else:
            primarydataset = self.publish_data_name
        return primarydataset
    
    def lfn2pfn(self):
        """
        PhEDEx Data Service lfn2pfn call
 
        input:   LFN,node name,protocol
        returns: DOM object with the content of the PhEDEx Data Service call
        """  
        params = {'node' : self.node , 'lfn': self.lfn , 'protocol': self.protocol}
        params = urllib.urlencode(params)
        datasvc_lfn2pfn="%s/lfn2pfn"%self.datasvc_url
        try:
            urlresults = urllib.urlopen(datasvc_lfn2pfn, params)
            urlresults = parse(urlresults)
        except IOError:
            msg="Unable to access PhEDEx Data Service at %s"%datasvc_lfn2pfn
            raise CrabException(msg)
        except:
            urlresults = None

        return urlresults
 
    def parse_error(self,urlresults):
        """
        look for errors in the DOM object returned by PhEDEx Data Service call
        """
        errormsg = None 
        errors=urlresults.getElementsByTagName('error')
        for error in errors:
            errormsg=error.childNodes[0].data
            if len(error.childNodes)>1:
               errormsg+=error.childNodes[1].data
        return errormsg
 
    def parse_lfn2pfn(self,urlresults):
        """
        Parse the content of the result of lfn2pfn PhEDEx Data Service  call
 
        input:    DOM object with the content of the lfn2pfn call
        returns:  PFN  
        """
        result = urlresults.getElementsByTagName('phedex')
               
        if not result:
              return []
        result = result[0]
        pfn = None
        mapping = result.getElementsByTagName('mapping')
        for m in mapping:
            pfn=m.getAttribute("pfn")
            if pfn:
              return pfn
 
    def getStageoutPFN( self ):
        """
        input:   LFN,node name,protocol
        returns: PFN 
        """
        if self.usePhedex:
            fullurl="%s/lfn2pfn?node=%s&lfn=%s&protocol=%s"%(self.datasvc_url,self.node,self.lfn,self.protocol) 
            domlfn2pfn = self.lfn2pfn()
            if not domlfn2pfn :
                msg="Unable to get info from %s"%fullurl
                raise CrabException(msg)
  
            errormsg = self.parse_error(domlfn2pfn)
            if errormsg: 
                msg="Error extracting info from %s due to: %s"%(fullurl,errormsg)
                raise CrabException(msg)
  
            stageoutpfn = self.parse_lfn2pfn(domlfn2pfn)
            if not stageoutpfn:
                msg ='Unable to get stageout path from TFC at Site %s \n'%self.node
                msg+='      Please alert the CompInfraSup group through their savannah %s \n'%self.FacOps_savannah
                msg+='      reporting: \n'
                msg+='       Summary: Unable to get user stageout from TFC at Site %s \n'%self.node
                msg+='       OriginalSubmission: stageout path is not retrieved from %s \n'%fullurl
                raise CrabException(msg)
        else:
            if self.sched in ['CAF','LSF'] :
                stageoutpfn = self.user_se_path+self.lfn 
            else: 
                stageoutpfn = 'srm://'+self.node+':'+self.user_port+self.user_se_path+self.lfn 

        return stageoutpfn 


if __name__ == '__main__':
  """
  Sort of unit testing to check Phedex API for whatever site and/or lfn.
  Usage:
     python PhEDExDatasvcInfo.py --node T2_IT_Bari --lfn /store/maremma

  """
  import getopt,sys
  from crab_util import *
  import common
  klass_name = 'SchedulerGlite'
  klass = importName(klass_name, klass_name)
  common.scheduler = klass()

  lfn="/store/user/"
  node='T2_IT_Bari'
  valid = ['node=','lfn=']
  try:
       opts, args = getopt.getopt(sys.argv[1:], "", valid)
  except getopt.GetoptError, ex:
       print str(ex)
       sys.exit(1)
  for o, a in opts:
        if o == "--node":
            node = a
        if o == "--lfn":
            lfn = a
  
  mycfg_params = { 'USER.storage_element': node }
  dsvc = PhEDExDatasvcInfo(mycfg_params)
  dsvc.lfn = lfn
  print dsvc.getStageoutPFN()

Revision:	1.26
Committed:	Fri Jul 31 16:21:12 2009 UTC (15 years, 9 months ago) by fanzago
Content type:	text/x-python
Branch:	MAIN
CVS Tags:	CRAB_2_7_0_pre2, CRAB_2_6_3_pre1, test_1, CRAB_2_7_0_pre1, CRAB_2_6_2, CRAB_2_6_2_pre2, CRAB_2_6_2_pre1
Branch point for:	CRAB_2_6_X_br
Changes since 1.25:	+7 -1 lines
Log Message:	changes related publication name too long, savannah task 10664
#	Content
1	from Actor import *
2	import urllib
3	from xml.dom.minidom import parse
4	from crab_exceptions import *
5	from WorkSpace import *
6	from urlparse import urlparse
7	from LFNBaseName import *
8
9	class PhEDExDatasvcInfo:
10	def __init__( self , cfg_params ):
11
12	## PhEDEx Data Service URL
13	url="https://cmsweb.cern.ch/phedex/datasvc/xml/prod"
14	self.datasvc_url = cfg_params.get("USER.datasvc_url",url)
15
16	self.FacOps_savannah = 'https://savannah.cern.ch/support/?func=additem&group=cmscompinfrasup'
17	stage_out_faq='https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCrabHowTo#Stageout_and_publication'
18	self.dataPub_faq = 'https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCrabForPublication'
19
20	self.srm_version = cfg_params.get("USER.srm_version",'srmv2')
21	self.node = cfg_params.get('USER.storage_element',None)
22
23
24
25	self.user_lfn = cfg_params.get("USER.lfn",'')
26	self.publish_data = cfg_params.get("USER.publish_data",0)
27	self.usenamespace = cfg_params.get("USER.usenamespace",0)
28	self.user_remote_dir = cfg_params.get("USER.user_remote_dir",'')
29	if self.user_remote_dir:
30	if ( self.user_remote_dir[-1] != '/' ) : self.user_remote_dir = self.user_remote_dir + '/'
31	if self.user_lfn:
32	msg = 'Warning: lfn has been deprecated, CRAB will ignore it.\n'
33	msg += '\t Please use only user_remote_dir removing lfn from your crab.cfg\n'
34	msg += '\t For further information please visit : \n\t%s'%stage_out_faq
35	common.logger.info(msg)
36
37	self.datasetpath = cfg_params.get("CMSSW.datasetpath")
38	self.publish_data_name = cfg_params.get('USER.publish_data_name','')
39
40	self.user_port = cfg_params.get("USER.storage_port",'8443')
41	self.user_se_path = cfg_params.get("USER.storage_path",'')
42	if self.user_se_path:
43	if ( self.user_se_path[-1] != '/' ) : self.user_se_path = self.user_se_path + '/'
44
45	#check if using "private" Storage
46	self.usePhedex = True
47	if not self.node :
48	msg = 'Please specify the storage_element name in your crab.cfg section [USER].\n'
49	msg +='\tFor further information please visit : %s'%stage_out_faq
50	raise CrabException(msg)
51	if (self.node.find('T1_') + self.node.find('T2_')+self.node.find('T3_')) == -3: self.usePhedex = False
52
53	if not self.usePhedex and ( self.user_remote_dir == '' or self.user_se_path == '' ):
54	msg = 'You are asking to stage out without using CMS Storage Name convention. In this case you \n'
55	msg += '\t must specify both user_remote_dir and storage_path in the crab.cfg section [USER].\n '
56	msg += '\t For further information please visit : \n\t%s'%stage_out_faq
57	raise CrabException(msg)
58	self.sched = common.scheduler.name().upper()
59	self.protocol = self.srm_version
60
61	self.forced_path = '/store/user/'
62	if self.sched in ['CAF','LSF']:
63	self.protocol = 'direct'
64	self.SE = {'CAF':'caf.cern.ch', 'LSF':''}
65	if self.sched == 'CAF': self.forced_path = '/store/caf/user/'
66
67	if not self.usePhedex:
68	self.forced_path = self.user_remote_dir
69	return
70
71	def getEndpoint(self):
72	'''
73	Return full SE endpoint and related infos
74	'''
75	self.lfn = self.getLFN()
76
77	#extract the PFN for the given node,LFN,protocol
78	endpoint = self.getStageoutPFN()
79
80	#extract SE name an SE_PATH (needed for publication)
81	SE, SE_PATH, User = self.splitEndpoint(endpoint)
82
83	return endpoint, self.lfn , SE, SE_PATH, User
84
85	def splitEndpoint(self, endpoint):
86	'''
87	Return relevant infos from endpoint
88	'''
89	SE = ''
90	SE_PATH = ''
91	USER = ''
92	if self.usePhedex:
93	if self.protocol == 'direct':
94	query=endpoint
95	SE_PATH = endpoint
96	SE = self.SE[self.sched]
97	else:
98	url = 'http://'+endpoint.split('://')[1]
99	# python > 2.4
100	# SE = urlparse(url).hostname
101	scheme, host, path, params, query, fragment = urlparse(url)
102	SE = host.split(':')[0]
103	SE_PATH = endpoint.split(host)[1]
104	USER = (query.split('user')[1]).split('/')[1]
105	else:
106	#### to test #####
107	# url = 'http://'+endpoint.split('://')[1]
108	# scheme, host, path, params, query, fragment = urlparse(url)
109	# SE = host.split(':')[0]
110	# SE_PATH = endpoint.split(host)[1]
111	SE = self.node
112	SE_PATH = self.user_se_path + self.user_remote_dir
113	try:
114	USER = (self.lfn.split('user')[1]).split('/')[1]
115	except:
116	pass
117
118	return SE, SE_PATH, USER
119
120
121	def getLFN(self):
122	"""
123	define the LFN composing the needed pieces
124	"""
125	lfn = ''
126	l_User = False
127	if not self.usePhedex and (int(self.publish_data) == 0 and int(self.usenamespace) == 0) :
128	### add here check if user is trying to force a wrong LFN using a T2 TODO
129	## check if storage_name is a T2 (siteDB query)
130	## if yes :match self.user_lfn with LFNBaseName...
131	## if NOT : raise (you are using a T2. It's not allowed stage out into self.user_path+self.user_lfn)
132	lfn = self.user_remote_dir
133	return lfn
134	if self.publish_data_name == '' and int(self.publish_data) == 1:
135	msg = "Error. The [USER] section does not have 'publish_data_name'\n"
136	msg += '\tFor further information please visit : \n\t%s'%self.dataPub_faq
137	raise CrabException(msg)
138	if self.publish_data_name == '' and int(self.usenamespace) == 1:
139	self.publish_data_name = "DefaultDataset"
140	if int(self.publish_data) == 1:
141	if self.sched in ['CAF']: l_User=True
142	primaryDataset = self.computePrimaryDataset()
143	### added the case lfn = LFNBase(self.forced_path, primaryDataset, self.publish_data_name, publish=True)
144	### for the publication in order to be able to check the lfn length
145	lfn = LFNBase(self.forced_path, primaryDataset, self.publish_data_name, publish=True) + '/${PSETHASH}/'
146	elif int(self.usenamespace) == 1:
147	if self.sched in ['CAF']: l_User=True
148	primaryDataset = self.computePrimaryDataset()
149	lfn = LFNBase(self.forced_path, primaryDataset, self.publish_data_name) + '/${PSETHASH}/'
150	else:
151	if self.sched in ['CAF','LSF']: l_User=True
152	lfn = LFNBase(self.forced_path,self.user_remote_dir)
153	return lfn
154
155	def computePrimaryDataset(self):
156	"""
157	compute the last part for the LFN in case of publication
158	"""
159	if (self.datasetpath.upper() != 'NONE'):
160	primarydataset = self.datasetpath.split("/")[1]
161	else:
162	primarydataset = self.publish_data_name
163	return primarydataset
164
165	def lfn2pfn(self):
166	"""
167	PhEDEx Data Service lfn2pfn call
168
169	input: LFN,node name,protocol
170	returns: DOM object with the content of the PhEDEx Data Service call
171	"""
172	params = {'node' : self.node , 'lfn': self.lfn , 'protocol': self.protocol}
173	params = urllib.urlencode(params)
174	datasvc_lfn2pfn="%s/lfn2pfn"%self.datasvc_url
175	try:
176	urlresults = urllib.urlopen(datasvc_lfn2pfn, params)
177	urlresults = parse(urlresults)
178	except IOError:
179	msg="Unable to access PhEDEx Data Service at %s"%datasvc_lfn2pfn
180	raise CrabException(msg)
181	except:
182	urlresults = None
183
184	return urlresults
185
186	def parse_error(self,urlresults):
187	"""
188	look for errors in the DOM object returned by PhEDEx Data Service call
189	"""
190	errormsg = None
191	errors=urlresults.getElementsByTagName('error')
192	for error in errors:
193	errormsg=error.childNodes[0].data
194	if len(error.childNodes)>1:
195	errormsg+=error.childNodes[1].data
196	return errormsg
197
198	def parse_lfn2pfn(self,urlresults):
199	"""
200	Parse the content of the result of lfn2pfn PhEDEx Data Service call
201
202	input: DOM object with the content of the lfn2pfn call
203	returns: PFN
204	"""
205	result = urlresults.getElementsByTagName('phedex')
206
207	if not result:
208	return []
209	result = result[0]
210	pfn = None
211	mapping = result.getElementsByTagName('mapping')
212	for m in mapping:
213	pfn=m.getAttribute("pfn")
214	if pfn:
215	return pfn
216
217	def getStageoutPFN( self ):
218	"""
219	input: LFN,node name,protocol
220	returns: PFN
221	"""
222	if self.usePhedex:
223	fullurl="%s/lfn2pfn?node=%s&lfn=%s&protocol=%s"%(self.datasvc_url,self.node,self.lfn,self.protocol)
224	domlfn2pfn = self.lfn2pfn()
225	if not domlfn2pfn :
226	msg="Unable to get info from %s"%fullurl
227	raise CrabException(msg)
228
229	errormsg = self.parse_error(domlfn2pfn)
230	if errormsg:
231	msg="Error extracting info from %s due to: %s"%(fullurl,errormsg)
232	raise CrabException(msg)
233
234	stageoutpfn = self.parse_lfn2pfn(domlfn2pfn)
235	if not stageoutpfn:
236	msg ='Unable to get stageout path from TFC at Site %s \n'%self.node
237	msg+=' Please alert the CompInfraSup group through their savannah %s \n'%self.FacOps_savannah
238	msg+=' reporting: \n'
239	msg+=' Summary: Unable to get user stageout from TFC at Site %s \n'%self.node
240	msg+=' OriginalSubmission: stageout path is not retrieved from %s \n'%fullurl
241	raise CrabException(msg)
242	else:
243	if self.sched in ['CAF','LSF'] :
244	stageoutpfn = self.user_se_path+self.lfn
245	else:
246	stageoutpfn = 'srm://'+self.node+':'+self.user_port+self.user_se_path+self.lfn
247
248	return stageoutpfn
249
250
251
252	if __name__ == '__main__':
253	"""
254	Sort of unit testing to check Phedex API for whatever site and/or lfn.
255	Usage:
256	python PhEDExDatasvcInfo.py --node T2_IT_Bari --lfn /store/maremma
257
258	"""
259	import getopt,sys
260	from crab_util import *
261	import common
262	klass_name = 'SchedulerGlite'
263	klass = importName(klass_name, klass_name)
264	common.scheduler = klass()
265
266	lfn="/store/user/"
267	node='T2_IT_Bari'
268	valid = ['node=','lfn=']
269	try:
270	opts, args = getopt.getopt(sys.argv[1:], "", valid)
271	except getopt.GetoptError, ex:
272	print str(ex)
273	sys.exit(1)
274	for o, a in opts:
275	if o == "--node":
276	node = a
277	if o == "--lfn":
278	lfn = a
279
280	mycfg_params = { 'USER.storage_element': node }
281	dsvc = PhEDExDatasvcInfo(mycfg_params)
282	dsvc.lfn = lfn
283	print dsvc.getStageoutPFN()
284