X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=system%2FTestPlc.py;h=8d0a738eafdaaf510fba821065d04339f65e2e53;hb=3f62be2bbaaae30826ee5607e7c06c6e91844c71;hp=714bbdbce83cea28dcdc0baf5463ad5fb6fd3192;hpb=b6581ae8a4409a3783ccb418fec1f07d50528df6;p=tests.git diff --git a/system/TestPlc.py b/system/TestPlc.py index 714bbdb..8d0a738 100644 --- a/system/TestPlc.py +++ b/system/TestPlc.py @@ -1,4 +1,6 @@ -# $Id$ +# Thierry Parmentelat +# Copyright (C) 2010 INRIA +# import os, os.path import datetime import time @@ -14,7 +16,7 @@ from TestUser import TestUser from TestKey import TestKey from TestSlice import TestSlice from TestSliver import TestSliver -from TestBox import TestBox +from TestBoxQemu import TestBoxQemu from TestSsh import TestSsh from TestApiserver import TestApiserver from TestSliceSfa import TestSliceSfa @@ -50,7 +52,7 @@ def node_mapper (method): actual.__doc__=method.__doc__ return actual -def slice_mapper_options (method): +def slice_mapper (method): def actual(self): overall=True slice_method = TestSlice.__dict__[method.__name__] @@ -64,15 +66,14 @@ def slice_mapper_options (method): actual.__doc__=method.__doc__ return actual -def slice_mapper_options_sfa (method): +def slice_sfa_mapper (method): def actual(self): - test_plc=self overall=True slice_method = TestSliceSfa.__dict__[method.__name__] - for slice_spec in self.plc_spec['sfa']['slices_sfa']: + for slice_spec in self.plc_spec['sfa']['sfa_slice_specs']: site_spec = self.locate_site (slice_spec['sitename']) test_site = TestSite(self,site_spec) - test_slice=TestSliceSfa(test_plc,test_site,slice_spec) + test_slice=TestSliceSfa(self,test_site,slice_spec) if not slice_method(test_slice,self.options): overall=False return overall # restore the doc text @@ -80,47 +81,71 @@ def slice_mapper_options_sfa (method): return actual SEP='' +SEPSFA='' class TestPlc: default_steps = [ - 'display', 'local_pre', SEP, - 'delete','create','install', 'configure', 'start', SEP, - 'fetch_keys', 'store_keys', 'clear_known_hosts', SEP, - 'initscripts', 'sites', 'nodes', 'slices', 'nodegroups', SEP, - 'reinstall_node', 'init_node','bootcd', 'configure_qemu', 'export_qemu', - 'kill_all_qemus', 'start_node', SEP, - # better use of time: do this now that the nodes are taking off - 'plcsh_stress_test', SEP, - 'nodes_ssh_debug', 'nodes_ssh_boot', 'check_slice', 'check_initscripts', SEP, - 'install_sfa', 'configure_sfa', 'import_sfa', 'start_sfa', SEP, - 'setup_sfa', 'add_sfa', 'update_sfa', SEP, - 'view_sfa', 'check_slice_sfa', 'delete_sfa', 'stop_sfa', SEP, - 'check_tcp', 'check_hooks', SEP, - 'force_gather_logs', 'force_local_post', + 'show', SEP, + 'vs_delete','timestamp_vs','vs_create', SEP, + 'plc_install', 'plc_configure', 'plc_start', SEP, + 'keys_fetch', 'keys_store', 'keys_clear_known_hosts', SEP, + 'initscripts', 'sites', 'nodes', 'slices', 'nodegroups', 'leases', SEP, + 'nodestate_reinstall', 'qemu_local_init','bootcd', 'qemu_local_config', SEP, + 'qemu_export', 'qemu_kill_mine', 'qemu_start', 'timestamp_qemu', SEP, + 'sfa_install', 'sfa_configure', 'cross_sfa_configure', 'sfa_import', 'sfa_start', SEPSFA, + 'sfi_configure@1', 'sfa_add_user@1', 'sfa_add_slice@1', 'sfa_discover@1', SEPSFA, + 'sfa_create_slice@1', 'sfa_check_slice_plc@1', SEPSFA, + 'sfa_update_user@1', 'sfa_update_slice@1', 'sfa_view@1', 'sfa_utest@1',SEPSFA, + # we used to run plcsh_stress_test, and then ssh_node_debug and ssh_node_boot + # but as the stress test might take a while, we sometimes missed the debug mode.. + 'ssh_node_debug@1', 'plcsh_stress_test@1', SEP, + 'ssh_node_boot@1', 'ssh_slice', 'check_initscripts', SEP, + 'ssh_slice_sfa@1', 'sfa_delete_slice@1', 'sfa_delete_user@1', SEPSFA, + 'check_tcp', SEP, + 'force_gather_logs', SEP, ] other_steps = [ - 'fresh_install', 'stop', 'vs_start', SEP, - 'clean_initscripts', 'clean_nodegroups','clean_all_sites', SEP, - 'clean_sites', 'clean_nodes', 'clean_slices', 'clean_keys', SEP, + 'check_hooks', + 'free_all', + 'show_boxes', 'local_list','local_rel','local_rel_plc','local_rel_qemu',SEP, + 'plc_stop', 'vs_start', 'vs_stop', SEP, + 'delete_initscripts', 'delete_nodegroups','delete_all_sites', SEP, + 'delete_sites', 'delete_nodes', 'delete_slices', 'keys_clean', SEP, + 'delete_leases', 'list_leases', SEP, 'populate' , SEP, - 'show_boxes', 'list_all_qemus', 'list_qemus', 'kill_qemus', SEP, - 'db_dump' , 'db_restore', SEP, - 'local_list','local_cleanup',SEP, - 'standby_1 through 20', + 'nodestate_show','nodestate_safeboot','nodestate_boot', SEP, + 'qemu_list_all', 'qemu_list_mine', 'qemu_kill_all', SEP, + 'sfa_plcclean', 'sfa_dbclean', 'sfa_stop','sfa_uninstall', 'sfi_clean', SEP, + 'plc_db_dump' , 'plc_db_restore', SEP, + 'standby_1 through 20',SEP, ] @staticmethod def printable_steps (list): - return " ".join(list).replace(" "+SEP+" "," \\\n") + single_line=" ".join(list)+" " + return single_line.replace(" "+SEP+" "," \\\n").replace(" "+SEPSFA+" "," \\\n") @staticmethod def valid_step (step): - return step != SEP + return step != SEP and step != SEPSFA + + # turn off the sfa-related steps when build has skipped SFA + # this is originally for centos5 as recent SFAs won't build on this platformb + @staticmethod + def check_whether_build_has_sfa (rpms_url): + # warning, we're now building 'sface' so let's be a bit more picky + retcod=os.system ("curl --silent %s/ | grep -q sfa-"%rpms_url) + # full builds are expected to return with 0 here + if retcod!=0: + # move all steps containing 'sfa' from default_steps to other_steps + sfa_steps= [ step for step in TestPlc.default_steps if step.find('sfa')>=0 ] + TestPlc.other_steps += sfa_steps + for step in sfa_steps: TestPlc.default_steps.remove(step) def __init__ (self,plc_spec,options): self.plc_spec=plc_spec self.options=options - self.test_ssh=TestSsh(self.plc_spec['hostname'],self.options.buildname) + self.test_ssh=TestSsh(self.plc_spec['host_box'],self.options.buildname) try: self.vserverip=plc_spec['vserverip'] self.vservername=plc_spec['vservername'] @@ -135,7 +160,7 @@ class TestPlc: return "%s.%s"%(name,self.vservername) def hostname(self): - return self.plc_spec['hostname'] + return self.plc_spec['host_box'] def is_local (self): return self.test_ssh.is_local() @@ -151,6 +176,9 @@ class TestPlc: def start_guest (self): return utils.system(self.test_ssh.actual_command(self.start_guest_in_host())) + def stop_guest (self): + return utils.system(self.test_ssh.actual_command(self.stop_guest_in_host())) + def run_in_guest (self,command): return utils.system(self.actual_command_in_guest(command)) @@ -161,10 +189,13 @@ class TestPlc: def host_to_guest(self,command): return "vserver %s exec %s"%(self.vservername,command) - #command gets run in the vserver + #start/stop the vserver def start_guest_in_host(self): return "vserver %s start"%(self.vservername) + def stop_guest_in_host(self): + return "vserver %s stop"%(self.vservername) + # xxx quick n dirty def run_in_guest_piped (self,local,remote): return utils.system(local+" | "+self.test_ssh.actual_command(self.host_to_guest(remote),keep_stdin=True)) @@ -261,29 +292,32 @@ class TestPlc: # a step for checking this stuff def show_boxes (self): + 'print summary of nodes location' for (box,nodes) in self.gather_hostBoxes().iteritems(): print box,":"," + ".join( [ node.name() for node in nodes ] ) return True # make this a valid step - def kill_all_qemus(self): - "all qemu boxes: kill all running qemus (even of former runs)" + def qemu_kill_all(self): + 'kill all qemu instances on the qemu boxes involved by this setup' # this is the brute force version, kill all qemus on that host box for (box,nodes) in self.gather_hostBoxes().iteritems(): # pass the first nodename, as we don't push template-qemu on testboxes nodedir=nodes[0].nodedir() - TestBox(box,self.options.buildname).kill_all_qemus(nodedir) + TestBoxQemu(box,self.options.buildname).qemu_kill_all(nodedir) return True # make this a valid step - def list_all_qemus(self): + def qemu_list_all(self): + 'list all qemu instances on the qemu boxes involved by this setup' for (box,nodes) in self.gather_hostBoxes().iteritems(): # this is the brute force version, kill all qemus on that host box - TestBox(box,self.options.buildname).list_all_qemus() + TestBoxQemu(box,self.options.buildname).qemu_list_all() return True # kill only the right qemus - def list_qemus(self): + def qemu_list_mine(self): + 'list qemu instances for our nodes' for (box,nodes) in self.gather_hostBoxes().iteritems(): # the fine-grain version for node in nodes: @@ -291,7 +325,8 @@ class TestPlc: return True # kill only the right qemus - def kill_qemus(self): + def qemu_kill_mine(self): + 'kill the qemu instances for our nodes' for (box,nodes) in self.gather_hostBoxes().iteritems(): # the fine-grain version for node in nodes: @@ -299,15 +334,17 @@ class TestPlc: return True #################### display config - def display (self): + def show (self): "show test configuration after localization" self.display_pass (1) self.display_pass (2) return True # entry point + always_display_keys=['PLC_WWW_HOST','nodes','sites',] def display_pass (self,passno): for (key,val) in self.plc_spec.iteritems(): + if not self.options.verbose and key not in TestPlc.always_display_keys: continue if passno == 2: if key == 'sites': for site in val: @@ -324,12 +361,13 @@ class TestPlc: for key in val: self.display_key_spec (key) elif passno == 1: - if key not in ['sites','initscripts','slices','keys']: + if key not in ['sites','initscripts','slices','keys', 'sfa']: print '+ ',key,':',val def display_site_spec (self,site): print '+ ======== site',site['site_fields']['name'] for (k,v) in site.iteritems(): + if not self.options.verbose and k not in TestPlc.always_display_keys: continue if k=='nodes': if v: print '+ ','nodes : ', @@ -347,8 +385,8 @@ class TestPlc: elif k == 'address_fields': pass else: - print '+ ',k, - PrettyPrinter(indent=8,depth=2).pprint(v) + print '+ ', + utils.pprint(k,v) def display_initscript_spec (self,initscript): print '+ ======== initscript',initscript['initscript_fields']['name'] @@ -379,10 +417,11 @@ class TestPlc: print '+ ',k,v def display_node_spec (self,node): - print "+ node",node['name'],"host_box=",node['host_box'], + print "+ node=%s host_box=%s"%(node['name'],node['host_box']), print "hostname=",node['node_fields']['hostname'], print "ip=",node['interface_fields']['ip'] - + if self.options.verbose: + utils.pprint("node details",node,depth=3) # another entry point for just showing the boxes involved def display_mapping (self): @@ -392,7 +431,7 @@ class TestPlc: @staticmethod def display_mapping_plc (plc_spec): print '+ MyPLC',plc_spec['name'] - print '+\tvserver address = root@%s:/vservers/%s'%(plc_spec['hostname'],plc_spec['vservername']) + print '+\tvserver address = root@%s:/vservers/%s'%(plc_spec['host_box'],plc_spec['vservername']) print '+\tIP = %s/%s'%(plc_spec['PLC_API_HOST'],plc_spec['vserverip']) for site_spec in plc_spec['sites']: for node_spec in site_spec['nodes']: @@ -404,6 +443,12 @@ class TestPlc: print '+\tqemu box %s'%node_spec['host_box'] print '+\thostname=%s'%node_spec['node_fields']['hostname'] + # write a timestamp in /vservers/<>.timestamp + # cannot be inside the vserver, that causes vserver .. build to cough + def timestamp_vs (self): + now=int(time.time()) + return utils.system(self.test_ssh.actual_command("echo %d > /vservers/%s.timestamp"%(now,self.vservername)))==0 + def local_pre (self): "run site-dependant pre-test script as defined in LocalTestResources" from LocalTestResources import local_resources @@ -419,19 +464,34 @@ class TestPlc: from LocalTestResources import local_resources return local_resources.step_list(self) - def local_cleanup (self): - "run site-dependant cleanup script as defined in LocalTestResources" + def local_rel (self): + "run site-dependant release script as defined in LocalTestResources" from LocalTestResources import local_resources - return local_resources.step_cleanup(self) + return local_resources.step_release(self) - def delete(self): + def local_rel_plc (self): + "run site-dependant release script as defined in LocalTestResources" + from LocalTestResources import local_resources + return local_resources.step_release_plc(self) + + def local_rel_qemu (self): + "run site-dependant release script as defined in LocalTestResources" + from LocalTestResources import local_resources + return local_resources.step_release_qemu(self) + + def vs_delete(self): "vserver delete the test myplc" self.run_in_host("vserver --silent %s delete"%self.vservername) + self.run_in_host("rm -f /vservers/%s.timestamp"%self.vservername) return True ### install - def create (self): + # historically the build was being fetched by the tests + # now the build pushes itself as a subdir of the tests workdir + # so that the tests do not have to worry about extracting the build (svn, git, or whatever) + def vs_create (self): "vserver creation (no install done)" + # push the local build/ dir to the testplc box if self.is_local(): # a full path for the local calls build_dir=os.path.dirname(sys.argv[0]) @@ -441,10 +501,9 @@ class TestPlc: else: # use a standard name - will be relative to remote buildname build_dir="build" - # run checkout in any case - would do an update if already exists - build_checkout = "svn checkout %s %s"%(self.options.build_url,build_dir) - if self.run_in_host(build_checkout) != 0: - return False + # remove for safety; do *not* mkdir first, otherwise we end up with build/build/ + self.test_ssh.rmdir(build_dir) + self.test_ssh.copy(build_dir,recursive=True) # the repo url is taken from arch-rpms-url # with the last step (i386) removed repo_url = self.options.arch_rpms_url @@ -462,32 +521,47 @@ class TestPlc: vserver_hostname=socket.gethostbyaddr(self.vserverip)[0] vserver_options += " --hostname %s"%vserver_hostname except: - pass + print "Cannot reverse lookup %s"%self.vserverip + print "This is considered fatal, as this might pollute the test results" + return False create_vserver="%(build_dir)s/%(script)s %(test_env_options)s %(vserver_name)s %(repo_url)s -- %(vserver_options)s"%locals() return self.run_in_host(create_vserver) == 0 ### install_rpm - def install(self): + def plc_install(self): "yum install myplc, noderepo, and the plain bootstrapfs" + + # workaround for getting pgsql8.2 on centos5 + if self.options.fcdistro == "centos5": + self.run_in_guest("rpm -Uvh http://download.fedora.redhat.com/pub/epel/5/i386/epel-release-5-3.noarch.rpm") + + # compute nodefamily if self.options.personality == "linux32": arch = "i386" elif self.options.personality == "linux64": arch = "x86_64" else: raise Exception, "Unsupported personality %r"%self.options.personality - return \ - self.run_in_guest("yum -y install myplc")==0 and \ - self.run_in_guest("yum -y install noderepo-%s-%s"%(self.options.pldistro,arch))==0 and \ - self.run_in_guest("yum -y install bootstrapfs-%s-%s-plain"%(self.options.pldistro,arch))==0 + nodefamily="%s-%s-%s"%(self.options.pldistro,self.options.fcdistro,arch) + + pkgs_list=[] + pkgs_list.append ("slicerepo-%s"%nodefamily) + pkgs_list.append ("myplc") + pkgs_list.append ("noderepo-%s"%nodefamily) + pkgs_list.append ("bootstrapfs-%s-plain"%nodefamily) + pkgs_string=" ".join(pkgs_list) + self.run_in_guest("yum -y install %s"%pkgs_string) + return self.run_in_guest("rpm -q %s"%pkgs_string)==0 ### - def configure(self): + def plc_configure(self): "run plc-config-tty" tmpname='%s.plc-config-tty'%(self.name()) fileconf=open(tmpname,'w') for var in [ 'PLC_NAME', - 'PLC_ROOT_PASSWORD', 'PLC_ROOT_USER', + 'PLC_ROOT_PASSWORD', + 'PLC_SLICE_PREFIX', 'PLC_MAIL_ENABLED', 'PLC_MAIL_SUPPORT_ADDRESS', 'PLC_DB_HOST', @@ -497,7 +571,11 @@ class TestPlc: 'PLC_WWW_HOST', 'PLC_BOOT_HOST', 'PLC_NET_DNS1', - 'PLC_NET_DNS2']: + 'PLC_NET_DNS2', + 'PLC_RESERVATION_GRANULARITY', + 'PLC_OMF_ENABLED', + 'PLC_OMF_XMPP_SERVER', + ]: fileconf.write ('e %s\n%s\n'%(var,self.plc_spec[var])) fileconf.write('w\n') fileconf.write('q\n') @@ -507,33 +585,41 @@ class TestPlc: utils.system('rm %s'%tmpname) return True - def start(self): + def plc_start(self): "service plc start" self.run_in_guest('service plc start') return True - def stop(self): + def plc_stop(self): "service plc stop" self.run_in_guest('service plc stop') return True def vs_start (self): + "start the PLC vserver" self.start_guest() return True + def vs_stop (self): + "stop the PLC vserver" + self.stop_guest() + return True + # stores the keys from the config for further use - def store_keys(self): + def keys_store(self): "stores test users ssh keys in keys/" for key_spec in self.plc_spec['keys']: TestKey(self,key_spec).store_key() return True - def clean_keys(self): - utils.system("rm -rf %s/keys/"%os.path(sys.argv[0])) + def keys_clean(self): + "removes keys cached in keys/" + utils.system("rm -rf ./keys") + return True # fetches the ssh keys in the plc's /etc/planetlab and stores them in keys/ # for later direct access to the nodes - def fetch_keys(self): + def keys_fetch(self): "gets ssh keys in /etc/planetlab/ and stores them locally in keys/" dir="./keys" if not os.path.isdir(dir): @@ -551,7 +637,7 @@ class TestPlc: "create sites with PLCAPI" return self.do_sites() - def clean_sites (self): + def delete_sites (self): "delete sites with PLCAPI" return self.do_sites(action="delete") @@ -570,17 +656,19 @@ class TestPlc: test_site.create_users() return True - def clean_all_sites (self): + def delete_all_sites (self): + "Delete all sites in PLC, and related objects" print 'auth_root',self.auth_root() site_ids = [s['site_id'] for s in self.apiserver.GetSites(self.auth_root(), {}, ['site_id'])] for site_id in site_ids: print 'Deleting site_id',site_id self.apiserver.DeleteSite(self.auth_root(),site_id) + return True def nodes (self): "create nodes with PLCAPI" return self.do_nodes() - def clean_nodes (self): + def delete_nodes (self): "delete nodes with PLCAPI" return self.do_nodes(action="delete") @@ -604,10 +692,72 @@ class TestPlc: def nodegroups (self): "create nodegroups with PLCAPI" return self.do_nodegroups("add") - def clean_nodegroups (self): + def delete_nodegroups (self): "delete nodegroups with PLCAPI" return self.do_nodegroups("delete") + YEAR = 365*24*3600 + @staticmethod + def translate_timestamp (start,grain,timestamp): + if timestamp < TestPlc.YEAR: return start+timestamp*grain + else: return timestamp + + @staticmethod + def timestamp_printable (timestamp): + return time.strftime('%m-%d %H:%M:%S UTC',time.gmtime(timestamp)) + + def leases(self): + "create leases (on reservable nodes only, use e.g. run -c default -c resa)" + now=int(time.time()) + grain=self.apiserver.GetLeaseGranularity(self.auth_root()) + print 'API answered grain=',grain + start=(now/grain)*grain + start += grain + # find out all nodes that are reservable + nodes=self.all_reservable_nodenames() + if not nodes: + utils.header ("No reservable node found - proceeding without leases") + return True + ok=True + # attach them to the leases as specified in plc_specs + # this is where the 'leases' field gets interpreted as relative of absolute + for lease_spec in self.plc_spec['leases']: + # skip the ones that come with a null slice id + if not lease_spec['slice']: continue + lease_spec['t_from']=TestPlc.translate_timestamp(start,grain,lease_spec['t_from']) + lease_spec['t_until']=TestPlc.translate_timestamp(start,grain,lease_spec['t_until']) + lease_addition=self.apiserver.AddLeases(self.auth_root(),nodes, + lease_spec['slice'],lease_spec['t_from'],lease_spec['t_until']) + if lease_addition['errors']: + utils.header("Cannot create leases, %s"%lease_addition['errors']) + ok=False + else: + utils.header('Leases on nodes %r for %s from %d (%s) until %d (%s)'%\ + (nodes,lease_spec['slice'], + lease_spec['t_from'],TestPlc.timestamp_printable(lease_spec['t_from']), + lease_spec['t_until'],TestPlc.timestamp_printable(lease_spec['t_until']))) + + return ok + + def delete_leases (self): + "remove all leases in the myplc side" + lease_ids= [ l['lease_id'] for l in self.apiserver.GetLeases(self.auth_root())] + utils.header("Cleaning leases %r"%lease_ids) + self.apiserver.DeleteLeases(self.auth_root(),lease_ids) + return True + + def list_leases (self): + "list all leases known to the myplc" + leases = self.apiserver.GetLeases(self.auth_root()) + now=int(time.time()) + for l in leases: + current=l['t_until']>=now + if self.options.verbose or current: + utils.header("%s %s from %s until %s"%(l['hostname'],l['name'], + TestPlc.timestamp_printable(l['t_from']), + TestPlc.timestamp_printable(l['t_until']))) + return True + # create nodegroups if needed, and populate def do_nodegroups (self, action="add"): # 1st pass to scan contents @@ -637,8 +787,7 @@ class TestPlc: tag_type_id = self.apiserver.AddTagType(auth, {'tagname':nodegroupname, 'description': 'for nodegroup %s'%nodegroupname, - 'category':'test', - 'min_role_id':10}) + 'category':'test'}) print 'located tag (type)',nodegroupname,'as',tag_type_id # create nodegroup nodegroups = self.apiserver.GetNodeGroups (auth, {'groupname':nodegroupname}) @@ -674,12 +823,23 @@ class TestPlc: overall=False return overall - def all_hostnames (self) : - hostnames = [] + # return a list of tuples (nodename,qemuname) + def all_node_infos (self) : + node_infos = [] for site_spec in self.plc_spec['sites']: - hostnames += [ node_spec['node_fields']['hostname'] \ - for node_spec in site_spec['nodes'] ] - return hostnames + node_infos += [ (node_spec['node_fields']['hostname'],node_spec['host_box']) \ + for node_spec in site_spec['nodes'] ] + return node_infos + + def all_nodenames (self): return [ x[0] for x in self.all_node_infos() ] + def all_reservable_nodenames (self): + res=[] + for site_spec in self.plc_spec['sites']: + for node_spec in site_spec['nodes']: + node_fields=node_spec['node_fields'] + if 'node_type' in node_fields and node_fields['node_type']=='reservable': + res.append(node_fields['hostname']) + return res # silent_minutes : during the first minutes nothing gets printed def nodes_check_boot_state (self, target_boot_state, timeout_minutes, silent_minutes,period=15): @@ -728,9 +888,9 @@ class TestPlc: return True def nodes_booted(self): - return self.nodes_check_boot_state('boot',timeout_minutes=20,silent_minutes=15) + return self.nodes_check_boot_state('boot',timeout_minutes=30,silent_minutes=20) - def check_nodes_ssh(self,debug,timeout_minutes,silent_minutes,period=20): + def check_nodes_ssh(self,debug,timeout_minutes,silent_minutes,period=15): # compute timeout timeout = datetime.datetime.now()+datetime.timedelta(minutes=timeout_minutes) graceout = datetime.datetime.now()+datetime.timedelta(minutes=silent_minutes) @@ -741,58 +901,50 @@ class TestPlc: else: message="boot" local_key = "keys/key1.rsa" - tocheck = self.all_hostnames() - utils.header("checking ssh access (expected in %s mode) to nodes %r"%(message,tocheck)) + node_infos = self.all_node_infos() + utils.header("checking ssh access (expected in %s mode) to nodes:"%message) + for (nodename,qemuname) in node_infos: + utils.header("hostname=%s -- qemubox=%s"%(nodename,qemuname)) utils.header("max timeout is %d minutes, silent for %d minutes (period is %s)"%\ (timeout_minutes,silent_minutes,period)) - while tocheck: - for hostname in tocheck: + while node_infos: + for node_info in node_infos: + (hostname,qemuname) = node_info # try to run 'hostname' in the node command = TestSsh (hostname,key=local_key).actual_command("hostname;uname -a") # don't spam logs - show the command only after the grace period - if datetime.datetime.now() > graceout: - success=utils.system(command) - else: - # truly silent, just print out a dot to show we're alive - print '.', - sys.stdout.flush() - command += " 2>/dev/null" - if self.options.dry_run: - print 'dry_run',command - success=0 - else: - success=os.system(command) + success = utils.system ( command, silent=datetime.datetime.now() < graceout) if success==0: utils.header('Successfully entered root@%s (%s)'%(hostname,message)) - # refresh tocheck - tocheck.remove(hostname) + # refresh node_infos + node_infos.remove(node_info) else: # we will have tried real nodes once, in case they're up - but if not, just skip (site_spec,node_spec)=self.locate_hostname(hostname) if TestNode.is_real_model(node_spec['node_fields']['model']): utils.header ("WARNING : check ssh access into real node %s - skipped"%hostname) - tocheck.remove(hostname) - if not tocheck: + node_infos.remove(node_info) + if not node_infos: return True if datetime.datetime.now() > timeout: - for hostname in tocheck: - utils.header("FAILURE to ssh into %s"%hostname) + for (hostname,qemuname) in node_infos: + utils.header("FAILURE to ssh into %s (on %s)"%(hostname,qemuname)) return False # otherwise, sleep for a while time.sleep(period) # only useful in empty plcs return True - def nodes_ssh_debug(self): + def ssh_node_debug(self): "Tries to ssh into nodes in debug mode with the debug ssh key" - return self.check_nodes_ssh(debug=True,timeout_minutes=30,silent_minutes=10) + return self.check_nodes_ssh(debug=True,timeout_minutes=10,silent_minutes=5) - def nodes_ssh_boot(self): + def ssh_node_boot(self): "Tries to ssh into nodes in production mode with the root ssh key" - return self.check_nodes_ssh(debug=False,timeout_minutes=30,silent_minutes=10) + return self.check_nodes_ssh(debug=False,timeout_minutes=40,silent_minutes=15) @node_mapper - def init_node (self): + def qemu_local_init (self): "all nodes : init a clean local directory for holding node-dep stuff like iso image..." pass @node_mapper @@ -800,15 +952,27 @@ class TestPlc: "all nodes: invoke GetBootMedium and store result locally" pass @node_mapper - def configure_qemu (self): + def qemu_local_config (self): "all nodes: compute qemu config qemu.conf and store it locally" pass @node_mapper - def reinstall_node (self): + def nodestate_reinstall (self): "all nodes: mark PLCAPI boot_state as reinstall" pass @node_mapper - def export_qemu (self): + def nodestate_safeboot (self): + "all nodes: mark PLCAPI boot_state as safeboot" + pass + @node_mapper + def nodestate_boot (self): + "all nodes: mark PLCAPI boot_state as boot" + pass + @node_mapper + def nodestate_show (self): + "all nodes: show PLCAPI boot_state" + pass + @node_mapper + def qemu_export (self): "all nodes: push local node-dep directory on the qemu box" pass @@ -826,9 +990,9 @@ class TestPlc: def do_check_initscripts(self): overall = True for slice_spec in self.plc_spec['slices']: - if not slice_spec.has_key('initscriptname'): + if not slice_spec.has_key('initscriptstamp'): continue - initscript=slice_spec['initscriptname'] + stamp=slice_spec['initscriptstamp'] for nodename in slice_spec['nodenames']: (site,node) = self.locate_node (nodename) # xxx - passing the wrong site - probably harmless @@ -836,7 +1000,7 @@ class TestPlc: test_slice = TestSlice (self,test_site,slice_spec) test_node = TestNode (self,test_site,node) test_sliver = TestSliver (self, test_node, test_slice) - if not test_sliver.check_initscript(initscript): + if not test_sliver.check_initscript_stamp(stamp): overall = False return overall @@ -851,7 +1015,7 @@ class TestPlc: self.apiserver.AddInitScript(self.auth_root(),initscript['initscript_fields']) return True - def clean_initscripts (self): + def delete_initscripts (self): "delete initscripts with PLCAPI" for initscript in self.plc_spec['initscripts']: initscript_name = initscript['initscript_fields']['name'] @@ -868,7 +1032,7 @@ class TestPlc: "create slices with PLCAPI" return self.do_slices() - def clean_slices (self): + def delete_slices (self): "delete slices with PLCAPI" return self.do_slices("delete") @@ -886,18 +1050,23 @@ class TestPlc: utils.header('Created Slice %s'%slice['slice_fields']['name']) return True - @slice_mapper_options - def check_slice(self): + @slice_mapper + def ssh_slice(self): "tries to ssh-enter the slice with the user key, to ensure slice creation" pass @node_mapper - def clear_known_hosts (self): + def keys_clear_known_hosts (self): "remove test nodes entries from the local known_hosts file" pass @node_mapper - def start_node (self) : + def qemu_start (self) : + "all nodes: start the qemu instance (also runs qemu-bridge-init start)" + pass + + @node_mapper + def timestamp_qemu (self) : "all nodes: start the qemu instance (also runs qemu-bridge-init start)" pass @@ -935,27 +1104,98 @@ class TestPlc: # in particular runs with --preserve (dont cleanup) and without --check # also it gets run twice, once with the --foreign option for creating fake foreign entries - ### install_sfa_rpm - def install_sfa(self): + ### sfa_install_rpm + def sfa_install(self): "yum install sfa, sfa-plc and sfa-client" - if self.options.personality == "linux32": - arch = "i386" - elif self.options.personality == "linux64": - arch = "x86_64" - else: - raise Exception, "Unsupported personality %r"%self.options.personality - return \ - self.run_in_guest("yum -y install sfa")==0 and \ - self.run_in_guest("yum -y install sfa-client")==0 and \ - self.run_in_guest("yum -y install sfa-plc")==0 and \ - self.run_in_guest("yum -y install sfa-sfatables")==0 + # ignore yum retcod + self.run_in_guest("yum -y install sfa sfa-client sfa-plc sfa-sfatables") + return self.run_in_guest("rpm -q sfa sfa-client sfa-plc sfa-sfatables")==0 + + + def sfa_dbclean(self): + "thoroughly wipes off the SFA database" + self.run_in_guest("sfa-nuke-plc.py")==0 + return True + + def sfa_plcclean(self): + "cleans the PLC entries that were created as a side effect of running the script" + # ignore result + sfa_spec=self.plc_spec['sfa'] + + for sfa_slice_spec in sfa_spec['sfa_slice_specs']: + slicename='%s_%s'%(sfa_spec['login_base'],sfa_slice_spec['slicename']) + try: self.apiserver.DeleteSlice(self.auth_root(),slicename) + except: print "Slice %s already absent from PLC db"%slicename + + username="%s@%s"%(sfa_spec['regularuser'],sfa_spec['domain']) + try: self.apiserver.DeletePerson(self.auth_root(),username) + except: print "User %s already absent from PLC db"%username + + print "REMEMBER TO RUN sfa_import AGAIN" + return True + + def sfa_uninstall(self): + "uses rpm to uninstall sfa - ignore result" + self.run_in_guest("rpm -e sfa sfa-sfatables sfa-client sfa-plc") + self.run_in_guest("rm -rf /var/lib/sfa") + self.run_in_guest("rm -rf /etc/sfa") + self.run_in_guest("rm -rf /var/log/sfa_access.log /var/log/sfa_import_plc.log /var/log/sfa.daemon") + # xxx tmp + self.run_in_guest("rpm -e --noscripts sfa-plc") + return True + + ### run unit tests for SFA + # NOTE: for some reason on f14/i386, yum install sfa-tests fails for no reason + # Running Transaction + # Transaction couldn't start: + # installing package sfa-tests-1.0-21.onelab.i686 needs 204KB on the / filesystem + # [('installing package sfa-tests-1.0-21.onelab.i686 needs 204KB on the / filesystem', (9, '/', 208896L))] + # no matter how many Gbs are available on the testplc + # could not figure out what's wrong, so... + # if the yum install phase fails, consider the test is successful + # other combinations will eventually run it hopefully + def sfa_utest(self): + "yum install sfa-tests and run SFA unittests" + self.run_in_guest("yum -y install sfa-tests") + # failed to install - forget it + if self.run_in_guest("rpm -q sfa-tests")!=0: + utils.header("WARNING: SFA unit tests failed to install, ignoring") + return True + return self.run_in_guest("/usr/share/sfa/tests/testAll.py")==0 + ### - def configure_sfa(self): + def confdir(self): + dirname="conf.%s"%self.plc_spec['name'] + if not os.path.isdir(dirname): + utils.system("mkdir -p %s"%dirname) + if not os.path.isdir(dirname): + raise "Cannot create config dir for plc %s"%self.name() + return dirname + + def conffile(self,filename): + return "%s/%s"%(self.confdir(),filename) + def confsubdir(self,dirname,clean,dry_run=False): + subdirname="%s/%s"%(self.confdir(),dirname) + if clean: + utils.system("rm -rf %s"%subdirname) + if not os.path.isdir(subdirname): + utils.system("mkdir -p %s"%subdirname) + if not dry_run and not os.path.isdir(subdirname): + raise "Cannot create config subdir %s for plc %s"%(dirname,self.name()) + return subdirname + + def conffile_clean (self,filename): + filename=self.conffile(filename) + return utils.system("rm -rf %s"%filename)==0 + + ### + def sfa_configure(self): "run sfa-config-tty" - tmpname='%s.sfa-config-tty'%(self.name()) + tmpname=self.conffile("sfa-config-tty") fileconf=open(tmpname,'w') for var in [ 'SFA_REGISTRY_ROOT_AUTH', - 'SFA_REGISTRY_LEVEL1_AUTH', + 'SFA_INTERFACE_HRN', +# 'SFA_REGISTRY_LEVEL1_AUTH', 'SFA_REGISTRY_HOST', 'SFA_AGGREGATE_HOST', 'SFA_SM_HOST', @@ -964,146 +1204,178 @@ class TestPlc: 'SFA_PLC_DB_HOST', 'SFA_PLC_DB_USER', 'SFA_PLC_DB_PASSWORD', - 'SFA_PLC_URL']: + 'SFA_PLC_URL', + ]: fileconf.write ('e %s\n%s\n'%(var,self.plc_spec['sfa'][var])) + # the way plc_config handles booleans just sucks.. + for var in ['SFA_API_DEBUG']: + val='false' + if self.plc_spec['sfa'][var]: val='true' + fileconf.write ('e %s\n%s\n'%(var,val)) fileconf.write('w\n') fileconf.write('R\n') fileconf.write('q\n') fileconf.close() utils.system('cat %s'%tmpname) self.run_in_guest_piped('cat %s'%tmpname,'sfa-config-tty') - utils.system('rm %s'%tmpname) return True - def import_sfa(self): + def aggregate_xml_line(self): + port=self.plc_spec['sfa']['neighbours-port'] + return '' % \ + (self.vserverip,self.plc_spec['sfa']['SFA_REGISTRY_ROOT_AUTH'],port) + + def registry_xml_line(self): + return '' % \ + (self.vserverip,self.plc_spec['sfa']['SFA_REGISTRY_ROOT_AUTH']) + + + # a cross step that takes all other plcs in argument + def cross_sfa_configure(self, other_plcs): + "writes aggregates.xml and registries.xml that point to all other PLCs in the test" + # of course with a single plc, other_plcs is an empty list + if not other_plcs: + return True + agg_fname=self.conffile("agg.xml") + file(agg_fname,"w").write("%s\n" % \ + " ".join([ plc.aggregate_xml_line() for plc in other_plcs ])) + utils.header ("(Over)wrote %s"%agg_fname) + reg_fname=self.conffile("reg.xml") + file(reg_fname,"w").write("%s\n" % \ + " ".join([ plc.registry_xml_line() for plc in other_plcs ])) + utils.header ("(Over)wrote %s"%reg_fname) + return self.test_ssh.copy_abs(agg_fname,'/vservers/%s/etc/sfa/aggregates.xml'%self.vservername)==0 \ + and self.test_ssh.copy_abs(reg_fname,'/vservers/%s/etc/sfa/registries.xml'%self.vservername)==0 + + def sfa_import(self): "sfa-import-plc" auth=self.plc_spec['sfa']['SFA_REGISTRY_ROOT_AUTH'] - self.run_in_guest('sfa-import-plc.py') - self.run_in_guest('cp /etc/sfa/authorities/%s/%s.pkey /etc/sfa/authorities/server.key'%(auth,auth)) - return True + return self.run_in_guest('sfa-import-plc.py')==0 +# not needed anymore +# self.run_in_guest('cp /etc/sfa/authorities/%s/%s.pkey /etc/sfa/authorities/server.key'%(auth,auth)) - def start_sfa(self): + def sfa_start(self): "service sfa start" - self.run_in_guest('service sfa start') - return True - - def setup_sfa(self): - "sfi client configuration" - dir_name=".sfi" - if os.path.exists(dir_name): - utils.system('rm -rf %s'%dir_name) - utils.system('mkdir %s'%dir_name) - file_name=dir_name + os.sep + 'fake-pi1.pkey' + return self.run_in_guest('service sfa start')==0 + + def sfi_configure(self): + "Create /root/.sfi on the plc side for sfi client configuration" + sfa_spec=self.plc_spec['sfa'] + dir_name=self.confsubdir("dot-sfi",clean=True,dry_run=self.options.dry_run) + if self.options.dry_run: return True + file_name=dir_name + os.sep + sfa_spec['piuser'] + '.pkey' fileconf=open(file_name,'w') fileconf.write (self.plc_spec['keys'][0]['private']) fileconf.close() + utils.header ("(Over)wrote %s"%file_name) file_name=dir_name + os.sep + 'sfi_config' fileconf=open(file_name,'w') - SFI_AUTH=self.plc_spec['sfa']['SFA_REGISTRY_ROOT_AUTH']+".main" + SFI_AUTH="%s.%s"%(sfa_spec['SFA_REGISTRY_ROOT_AUTH'],sfa_spec['login_base']) fileconf.write ("SFI_AUTH='%s'"%SFI_AUTH) fileconf.write('\n') - SFI_USER=SFI_AUTH+'.fake-pi1' + SFI_USER=SFI_AUTH + '.' + sfa_spec['piuser'] fileconf.write ("SFI_USER='%s'"%SFI_USER) fileconf.write('\n') - SFI_REGISTRY='http://' + self.plc_spec['sfa']['SFA_PLC_DB_HOST'] + ':12345/' + SFI_REGISTRY='http://' + sfa_spec['SFA_PLC_DB_HOST'] + ':12345/' fileconf.write ("SFI_REGISTRY='%s'"%SFI_REGISTRY) fileconf.write('\n') - SFI_SM='http://' + self.plc_spec['sfa']['SFA_PLC_DB_HOST'] + ':12347/' + SFI_SM='http://' + sfa_spec['SFA_PLC_DB_HOST'] + ':12347/' fileconf.write ("SFI_SM='%s'"%SFI_SM) fileconf.write('\n') fileconf.close() + utils.header ("(Over)wrote %s"%file_name) - file_name=dir_name + os.sep + 'person.xml' - fileconf=open(file_name,'w') - for record in self.plc_spec['sfa']['sfa_person_xml']: - person_record=record - fileconf.write(person_record) - fileconf.write('\n') - fileconf.close() - - file_name=dir_name + os.sep + 'slice.xml' - fileconf=open(file_name,'w') - for record in self.plc_spec['sfa']['sfa_slice_xml']: - slice_record=record - #slice_record=self.plc_spec['sfa']['sfa_slice_xml'] - fileconf.write(slice_record) - fileconf.write('\n') - fileconf.close() + # cannot use sfa_slice_mapper to pass dir_name + for slice_spec in self.plc_spec['sfa']['sfa_slice_specs']: + site_spec = self.locate_site (slice_spec['sitename']) + test_site = TestSite(self,site_spec) + test_slice=TestSliceSfa(self,test_site,slice_spec) + test_slice.sfi_config(dir_name) file_name=dir_name + os.sep + 'slice.rspec' fileconf=open(file_name,'w') slice_rspec='' - for (key, value) in self.plc_spec['sfa']['sfa_slice_rspec'].items(): + for (key, value) in sfa_spec['sfa_slice_rspec'].items(): slice_rspec +=value fileconf.write(slice_rspec) fileconf.write('\n') fileconf.close() - location = "root/" + utils.header ("(Over)wrote %s"%file_name) + + # push to the remote root's .sfi + location = "root/.sfi" remote="/vservers/%s/%s"%(self.vservername,location) self.test_ssh.copy_abs(dir_name, remote, recursive=True) - #utils.system('cat %s'%tmpname) - utils.system('rm -rf %s'%dir_name) return True - def add_sfa(self): - "run sfi.py add (on Registry) and sfi.py create (on SM) to form new objects" - test_plc=self - test_user_sfa=TestUserSfa(test_plc,self.plc_spec['sfa']) - success=test_user_sfa.add_user() + def sfi_clean (self): + "clean up /root/.sfi on the plc side" + self.run_in_guest("rm -rf /root/.sfi") + return True - for slice_spec in self.plc_spec['sfa']['slices_sfa']: - site_spec = self.locate_site (slice_spec['sitename']) - test_site = TestSite(self,site_spec) - test_slice_sfa=TestSliceSfa(test_plc,test_site,slice_spec) - success1=test_slice_sfa.add_slice() - success2=test_slice_sfa.create_slice() - return success and success1 and success2 - - def update_sfa(self): - "run sfi.py update (on Registry) and sfi.py create (on SM) on existing objects" - test_plc=self - test_user_sfa=TestUserSfa(test_plc,self.plc_spec['sfa']) - success1=test_user_sfa.update_user() - - for slice_spec in self.plc_spec['sfa']['slices_sfa']: - site_spec = self.locate_site (slice_spec['sitename']) - test_site = TestSite(self,site_spec) - test_slice_sfa=TestSliceSfa(test_plc,test_site,slice_spec) - success2=test_slice_sfa.update_slice() - return success1 and success2 - - def view_sfa(self): + def sfa_add_user(self): + "run sfi.py add using person.xml" + return TestUserSfa(self).add_user() + + def sfa_update_user(self): + "run sfi.py update using person.xml" + return TestUserSfa(self).update_user() + + @slice_sfa_mapper + def sfa_add_slice(self): + "run sfi.py add (on Registry) from slice.xml" + pass + + @slice_sfa_mapper + def sfa_discover(self): + "discover resources into resouces_in.rspec" + pass + + @slice_sfa_mapper + def sfa_create_slice(self): + "run sfi.py create (on SM) - 1st time" + pass + + @slice_sfa_mapper + def sfa_check_slice_plc(self): + "check sfa_create_slice at the plcs - all local nodes should be in slice" + pass + + @slice_sfa_mapper + def sfa_update_slice(self): + "run sfi.py create (on SM) on existing object" + pass + + def sfa_view(self): "run sfi.py list and sfi.py show (both on Registry) and sfi.py slices and sfi.py resources (both on SM)" - auth=self.plc_spec['sfa']['SFA_REGISTRY_ROOT_AUTH'] + sfa_spec=self.plc_spec['sfa'] + auth=sfa_spec['SFA_REGISTRY_ROOT_AUTH'] return \ - self.run_in_guest("sfi.py -d /root/.sfi/ list %s.main"%auth)==0 and \ - self.run_in_guest("sfi.py -d /root/.sfi/ show %s.main"%auth)==0 and \ + self.run_in_guest("sfi.py -d /root/.sfi/ list %s.%s"%(auth,sfa_spec['login_base']))==0 and \ + self.run_in_guest("sfi.py -d /root/.sfi/ show %s.%s"%(auth,sfa_spec['login_base']))==0 and \ self.run_in_guest("sfi.py -d /root/.sfi/ slices")==0 and \ - self.run_in_guest("sfi.py -d /root/.sfi/ resources")==0 + self.run_in_guest("sfi.py -d /root/.sfi/ resources -o resources")==0 - @slice_mapper_options_sfa - def check_slice_sfa(self): + @slice_sfa_mapper + def ssh_slice_sfa(self): "tries to ssh-enter the SFA slice" pass - def delete_sfa(self): - "run sfi.py delete (on SM), sfi.py remove (on Registry)" - test_plc=self - test_user_sfa=TestUserSfa(test_plc,self.plc_spec['sfa']) - success1=test_user_sfa.delete_user() - for slice_spec in self.plc_spec['sfa']['slices_sfa']: - site_spec = self.locate_site (slice_spec['sitename']) - test_site = TestSite(self,site_spec) - test_slice_sfa=TestSliceSfa(test_plc,test_site,slice_spec) - success2=test_slice_sfa.delete_slice() + def sfa_delete_user(self): + "run sfi.py delete (on SM) for user" + test_user_sfa=TestUserSfa(self) + return test_user_sfa.delete_user() - return success1 and success2 + @slice_sfa_mapper + def sfa_delete_slice(self): + "run sfi.py delete (on SM), sfi.py remove (on Registry) to clean slices" + pass - def stop_sfa(self): + def sfa_stop(self): "service sfa stop" - self.run_in_guest('service sfa stop') + self.run_in_guest('service sfa stop')==0 return True def populate (self): @@ -1195,20 +1467,22 @@ class TestPlc: name=str(d) return "/root/%s-%s.sql"%(database,name) - def db_dump(self): - dump=self.dbfile("planetab4") - self.run_in_guest('pg_dump -U pgsqluser planetlab4 -f '+ dump) - utils.header('Dumped planetlab4 database in %s'%dump) + def plc_db_dump(self): + 'dump the planetlab5 DB in /root in the PLC - filename has time' + dump=self.dbfile("planetab5") + self.run_in_guest('pg_dump -U pgsqluser planetlab5 -f '+ dump) + utils.header('Dumped planetlab5 database in %s'%dump) return True - def db_restore(self): - dump=self.dbfile("planetab4") + def plc_db_restore(self): + 'restore the planetlab5 DB - looks broken, but run -n might help' + dump=self.dbfile("planetab5") ##stop httpd service self.run_in_guest('service httpd stop') # xxx - need another wrapper - self.run_in_guest_piped('echo drop database planetlab4','psql --user=pgsqluser template1') - self.run_in_guest('createdb -U postgres --encoding=UNICODE --owner=pgsqluser planetlab4') - self.run_in_guest('psql -U pgsqluser planetlab4 -f '+dump) + self.run_in_guest_piped('echo drop database planetlab5','psql --user=pgsqluser template1') + self.run_in_guest('createdb -U postgres --encoding=UNICODE --owner=pgsqluser planetlab5') + self.run_in_guest('psql -U pgsqluser planetlab5 -f '+dump) ##starting httpd service self.run_in_guest('service httpd start')