CRAB/python/PhEDExDatasvcInfo.py

from Actor import *
import urllib
from xml.dom.minidom import parse
from crab_exceptions import *
from crab_logger import Logger
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/SWGuideCrabFaq#How_to_store_output_with_CRAB_2'
        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.message(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 or 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
        urlresults = urllib.urlopen(datasvc_lfn2pfn, params)
        try:
            urlresults = parse(urlresults)
        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:	1.22
Committed:	Tue Mar 31 23:43:22 2009 UTC (16 years, 1 month ago) by spiga
Content type:	text/x-python
Branch:	MAIN
CVS Tags:	CRAB_2_5_1_pre1
Changes since 1.21:	+4 -2 lines
Log Message:	caf namespace must be /store/caf/user/
#	Content
1	from Actor import *
2	import urllib
3	from xml.dom.minidom import parse
4	from crab_exceptions import *
5	from crab_logger import Logger
6	from WorkSpace import *
7	from urlparse import urlparse
8	from LFNBaseName import *
9
10	class PhEDExDatasvcInfo:
11	def __init__( self , cfg_params ):
12
13	## PhEDEx Data Service URL
14	url="https://cmsweb.cern.ch/phedex/datasvc/xml/prod"
15	self.datasvc_url = cfg_params.get("USER.datasvc_url",url)
16
17	self.FacOps_savannah = 'https://savannah.cern.ch/support/?func=additem&group=cmscompinfrasup'
18	stage_out_faq='https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCrabFaq#How_to_store_output_with_CRAB_2'
19	self.dataPub_faq = 'https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCrabForPublication'
20
21	self.srm_version = cfg_params.get("USER.srm_version",'srmv2')
22	self.node = cfg_params.get('USER.storage_element',None)
23
24
25
26	self.user_lfn = cfg_params.get("USER.lfn",'')
27	self.publish_data = cfg_params.get("USER.publish_data",0)
28	self.usenamespace = cfg_params.get("USER.usenamespace",0)
29	self.user_remote_dir = cfg_params.get("USER.user_remote_dir",'')
30	if self.user_remote_dir:
31	if ( self.user_remote_dir[-1] != '/' ) : self.user_remote_dir = self.user_remote_dir + '/'
32	if self.user_lfn:
33	msg = 'Warning: lfn has been deprecated, CRAB will ignore it.\n'
34	msg += '\t Please use only user_remote_dir removing lfn from your crab.cfg\n'
35	msg += '\t For further information please visit : \n\t%s'%stage_out_faq
36	common.logger.message(msg)
37
38	self.datasetpath = cfg_params.get("CMSSW.datasetpath")
39	self.publish_data_name = cfg_params.get('USER.publish_data_name','')
40
41	self.user_port = cfg_params.get("USER.storage_port",'8443')
42	self.user_se_path = cfg_params.get("USER.storage_path",'')
43	if self.user_se_path:
44	if ( self.user_se_path[-1] != '/' ) : self.user_se_path = self.user_se_path + '/'
45
46	#check if using "private" Storage
47	self.usePhedex = True
48	if not self.node :
49	msg = 'Please specify the storage_element name in your crab.cfg section [USER].\n'
50	msg +='\tFor further information please visit : %s'%stage_out_faq
51	raise CrabException(msg)
52	if (self.node.find('T1_') + self.node.find('T2_')+self.node.find('T3_')) == -3: self.usePhedex = False
53
54	if not self.usePhedex and ( self.user_remote_dir == '' or self.user_se_path == '' ):
55	msg = 'You are asking to stage out without using CMS Storage Name convention. In this case you \n'
56	msg += '\t must specify both user_remote_dir and storage_path in the crab.cfg section [USER].\n '
57	msg += '\t For further information please visit : \n\t%s'%stage_out_faq
58	raise CrabException(msg)
59	self.sched = common.scheduler.name().upper()
60	self.protocol = self.srm_version
61
62	self.forced_path = '/store/user/'
63	if self.sched in ['CAF','LSF']:
64	self.protocol = 'direct'
65	self.SE = {'CAF':'caf.cern.ch', 'LSF':''}
66	if self.sched == 'CAF': self.forced_path = '/store/caf/user/'
67
68	if not self.usePhedex:
69	self.forced_path = self.user_remote_dir
70	return
71
72	def getEndpoint(self):
73	'''
74	Return full SE endpoint and related infos
75	'''
76	self.lfn = self.getLFN()
77
78	#extract the PFN for the given node,LFN,protocol
79	endpoint = self.getStageoutPFN()
80
81	#extract SE name an SE_PATH (needed for publication)
82	SE, SE_PATH, User = self.splitEndpoint(endpoint)
83
84	return endpoint, self.lfn , SE, SE_PATH, User
85
86	def splitEndpoint(self, endpoint):
87	'''
88	Return relevant infos from endpoint
89	'''
90	SE = ''
91	SE_PATH = ''
92	USER = ''
93	if self.usePhedex:
94	if self.protocol == 'direct':
95	query=endpoint
96	SE_PATH = endpoint
97	SE = self.SE[self.sched]
98	else:
99	url = 'http://'+endpoint.split('://')[1]
100	# python > 2.4
101	# SE = urlparse(url).hostname
102	scheme, host, path, params, query, fragment = urlparse(url)
103	SE = host.split(':')[0]
104	SE_PATH = endpoint.split(host)[1]
105	USER = (query.split('user')[1]).split('/')[1]
106	else:
107	#### to test #####
108	# url = 'http://'+endpoint.split('://')[1]
109	# scheme, host, path, params, query, fragment = urlparse(url)
110	# SE = host.split(':')[0]
111	# SE_PATH = endpoint.split(host)[1]
112	SE = self.node
113	SE_PATH = self.user_se_path + self.user_remote_dir
114	try:
115	USER = (self.lfn.split('user')[1]).split('/')[1]
116	except:
117	pass
118
119	return SE, SE_PATH, USER
120
121
122	def getLFN(self):
123	"""
124	define the LFN composing the needed pieces
125	"""
126	lfn = ''
127	l_User = False
128	if not self.usePhedex and (int(self.publish_data) == 0 and int(self.usenamespace) == 0) :
129	### add here check if user is trying to force a wrong LFN using a T2 TODO
130	## check if storage_name is a T2 (siteDB query)
131	## if yes :match self.user_lfn with LFNBaseName...
132	## if NOT : raise (you are using a T2. It's not allowed stage out into self.user_path+self.user_lfn)
133	lfn = self.user_remote_dir
134	return lfn
135	if self.publish_data_name == '' and int(self.publish_data) == 1:
136	msg = "Error. The [USER] section does not have 'publish_data_name'\n"
137	msg += '\tFor further information please visit : \n\t%s'%self.dataPub_faq
138	raise CrabException(msg)
139	if self.publish_data_name == '' and int(self.usenamespace) == 1:
140	self.publish_data_name = "DefaultDataset"
141	if int(self.publish_data) == 1 or int(self.usenamespace) == 1:
142	if self.sched in ['CAF']: l_User=True
143	primaryDataset = self.computePrimaryDataset()
144	lfn = LFNBase(self.forced_path, primaryDataset, self.publish_data_name) + '/${PSETHASH}/'
145	else:
146	if self.sched in ['CAF','LSF']: l_User=True
147	lfn = LFNBase(self.forced_path,self.user_remote_dir)
148	return lfn
149
150	def computePrimaryDataset(self):
151	"""
152	compute the last part for the LFN in case of publication
153	"""
154	if (self.datasetpath.upper() != 'NONE'):
155	primarydataset = self.datasetpath.split("/")[1]
156	else:
157	primarydataset = self.publish_data_name
158	return primarydataset
159
160	def lfn2pfn(self):
161	"""
162	PhEDEx Data Service lfn2pfn call
163
164	input: LFN,node name,protocol
165	returns: DOM object with the content of the PhEDEx Data Service call
166	"""
167	params = {'node' : self.node , 'lfn': self.lfn , 'protocol': self.protocol}
168	params = urllib.urlencode(params)
169	datasvc_lfn2pfn="%s/lfn2pfn"%self.datasvc_url
170	urlresults = urllib.urlopen(datasvc_lfn2pfn, params)
171	try:
172	urlresults = parse(urlresults)
173	except:
174	urlresults = None
175
176	return urlresults
177
178	def parse_error(self,urlresults):
179	"""
180	look for errors in the DOM object returned by PhEDEx Data Service call
181	"""
182	errormsg = None
183	errors=urlresults.getElementsByTagName('error')
184	for error in errors:
185	errormsg=error.childNodes[0].data
186	if len(error.childNodes)>1:
187	errormsg+=error.childNodes[1].data
188	return errormsg
189
190	def parse_lfn2pfn(self,urlresults):
191	"""
192	Parse the content of the result of lfn2pfn PhEDEx Data Service call
193
194	input: DOM object with the content of the lfn2pfn call
195	returns: PFN
196	"""
197	result = urlresults.getElementsByTagName('phedex')
198
199	if not result:
200	return []
201	result = result[0]
202	pfn = None
203	mapping = result.getElementsByTagName('mapping')
204	for m in mapping:
205	pfn=m.getAttribute("pfn")
206	if pfn:
207	return pfn
208
209	def getStageoutPFN( self ):
210	"""
211	input: LFN,node name,protocol
212	returns: PFN
213	"""
214	if self.usePhedex:
215	fullurl="%s/lfn2pfn?node=%s&lfn=%s&protocol=%s"%(self.datasvc_url,self.node,self.lfn,self.protocol)
216	domlfn2pfn = self.lfn2pfn()
217	if not domlfn2pfn :
218	msg="Unable to get info from %s"%fullurl
219	raise CrabException(msg)
220
221	errormsg = self.parse_error(domlfn2pfn)
222	if errormsg:
223	msg="Error extracting info from %s due to: %s"%(fullurl,errormsg)
224	raise CrabException(msg)
225
226	stageoutpfn = self.parse_lfn2pfn(domlfn2pfn)
227	if not stageoutpfn:
228	msg ='Unable to get stageout path from TFC at Site %s \n'%self.node
229	msg+=' Please alert the CompInfraSup group through their savannah %s \n'%self.FacOps_savannah
230	msg+=' reporting: \n'
231	msg+=' Summary: Unable to get user stageout from TFC at Site %s \n'%self.node
232	msg+=' OriginalSubmission: stageout path is not retrieved from %s \n'%fullurl
233	raise CrabException(msg)
234	else:
235	if self.sched in ['CAF','LSF'] :
236	stageoutpfn = self.user_se_path+self.lfn
237	else:
238	stageoutpfn = 'srm://'+self.node+':'+self.user_port+self.user_se_path+self.lfn
239
240	return stageoutpfn
241
242
243
244	if __name__ == '__main__':
245	"""
246	Sort of unit testing to check Phedex API for whatever site and/or lfn.
247	Usage:
248	python PhEDExDatasvcInfo.py --node T2_IT_Bari --lfn /store/maremma
249
250	"""
251	import getopt,sys
252	from crab_util import *
253	import common
254	klass_name = 'SchedulerGlite'
255	klass = importName(klass_name, klass_name)
256	common.scheduler = klass()
257
258	lfn="/store/user/"
259	node='T2_IT_Bari'
260	valid = ['node=','lfn=']
261	try:
262	opts, args = getopt.getopt(sys.argv[1:], "", valid)
263	except getopt.GetoptError, ex:
264	print str(ex)
265	sys.exit(1)
266	for o, a in opts:
267	if o == "--node":
268	node = a
269	if o == "--lfn":
270	lfn = a
271
272	mycfg_params = { 'USER.storage_element': node }
273	dsvc = PhEDExDatasvcInfo(mycfg_params)
274	dsvc.lfn = lfn
275	print dsvc.getStageoutPFN()
276