X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;ds=sidebyside;f=system%2FTestPlc.py;h=cd9f222d3ab4a8073bb3ea287c8afe8339e23119;hb=5abc2125f8a5aeb58783affd6a3b532f02acd216;hp=25feaf6574603328a6395b65fb44390af65b8295;hpb=124a6036e9068ab1743a8702af219b27003ea686;p=tests.git diff --git a/system/TestPlc.py b/system/TestPlc.py index 25feaf6..cd9f222 100644 --- a/system/TestPlc.py +++ b/system/TestPlc.py @@ -3,9 +3,9 @@ import os, os.path import datetime import time import sys -import datetime import traceback from types import StringTypes +import socket import utils from TestSite import TestSite @@ -63,25 +63,26 @@ SEP='' class TestPlc: default_steps = ['uninstall','install','install_rpm', - 'configure', 'start', SEP, + 'configure', 'start', 'fetch_keys', SEP, 'store_keys', 'clear_known_hosts', 'initscripts', SEP, 'sites', 'nodes', 'slices', 'nodegroups', SEP, 'init_node','bootcd', 'configure_qemu', 'export_qemu', 'kill_all_qemus', 'reinstall_node','start_node', SEP, - 'standby_20', SEP, - 'nodes_booted', 'nodes_ssh', 'check_slice', - 'check_initscripts', 'check_tcp',SEP, - 'force_gather_logs', 'force_kill_qemus', ] - other_steps = [ 'stop_all_vservers','fresh_install', 'cache_rpm', 'stop', SEP, - 'clean_sites', 'clean_nodes', 'clean_slices', 'clean_keys', SEP, + 'nodes_booted', 'nodes_ssh', 'check_slice', 'check_initscripts', SEP, + 'check_sanity', 'check_tcp', 'plcsh_stress_test', SEP, + 'force_gather_logs', 'force_kill_qemus', 'force_record_tracker','force_free_tracker' ] + other_steps = [ 'stop_all_vservers','fresh_install', 'cache_rpm', 'stop', 'vs_start', SEP, + 'clean_initscripts', 'clean_nodegroups','clean_all_sites', SEP, + 'clean_sites', 'clean_nodes', + 'clean_slices', 'clean_keys', SEP, 'show_boxes', 'list_all_qemus', 'list_qemus', SEP, - 'db_dump' , 'db_restore', + 'db_dump' , 'db_restore', ' cleanup_tracker', 'standby_1 through 20' ] @staticmethod def printable_steps (list): - return " ".join(list).replace(" "+SEP+" ","\n") + return " ".join(list).replace(" "+SEP+" "," \\\n") @staticmethod def valid_step (step): return step != SEP @@ -96,17 +97,12 @@ class TestPlc: self.url="https://%s:443/PLCAPI/"%plc_spec['vserverip'] self.vserver=True except: - self.vserver=False - self.url="https://%s:443/PLCAPI/"%plc_spec['hostname'] -# utils.header('Using API url %s'%self.url) + raise Exception,'chroot-based myplc testing is deprecated' self.apiserver=TestApiserver(self.url,options.dry_run) def name(self): name=self.plc_spec['name'] - if self.vserver: - return name+".vserver.%s"%self.vservername - else: - return name+".chroot" + return "%s.%s"%(name,self.vservername) def hostname(self): return self.plc_spec['hostname'] @@ -122,37 +118,23 @@ class TestPlc: def actual_command_in_guest (self,command): return self.test_ssh.actual_command(self.host_to_guest(command)) + def start_guest (self): + return utils.system(self.test_ssh.actual_command(self.start_guest_in_host())) + def run_in_guest (self,command): return utils.system(self.actual_command_in_guest(command)) def run_in_host (self,command): return self.test_ssh.run_in_buildname(command) - #command gets run in the chroot/vserver + #command gets run in the vserver def host_to_guest(self,command): - if self.vserver: - return "vserver %s exec %s"%(self.vservername,command) - else: - return "chroot /plc/root %s"%TestSsh.backslash_shell_specials(command) + return "vserver %s exec %s"%(self.vservername,command) + + #command gets run in the vserver + def start_guest_in_host(self): + return "vserver %s start"%(self.vservername) - # copy a file to the myplc root image - pass in_data=True if the file must go in /plc/data - def copy_in_guest (self, localfile, remotefile, in_data=False): - if in_data: - chroot_dest="/plc/data" - else: - chroot_dest="/plc/root" - if self.is_local(): - if not self.vserver: - utils.system("cp %s %s/%s"%(localfile,chroot_dest,remotefile)) - else: - utils.system("cp %s /vservers/%s/%s"%(localfile,self.vservername,remotefile)) - else: - if not self.vserver: - utils.system("scp %s %s:%s/%s"%(localfile,self.hostname(),chroot_dest,remotefile)) - else: - utils.system("scp %s %s@/vservers/%s/%s"%(localfile,self.hostname(),self.vservername,remotefile)) - - # 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)) @@ -197,6 +179,37 @@ class TestPlc: return slice raise Exception,"Cannot locate slice %s"%slicename + def all_sliver_objs (self): + result=[] + for slice_spec in self.plc_spec['slices']: + slicename = slice_spec['slice_fields']['name'] + for nodename in slice_spec['nodenames']: + result.append(self.locate_sliver_obj (nodename,slicename)) + return result + + def locate_sliver_obj (self,nodename,slicename): + (site,node) = self.locate_node(nodename) + slice = self.locate_slice (slicename) + # build objects + test_site = TestSite (self, site) + test_node = TestNode (self, test_site,node) + # xxx the slice site is assumed to be the node site - mhh - probably harmless + test_slice = TestSlice (self, test_site, slice) + return TestSliver (self, test_node, test_slice) + + def locate_first_node(self): + nodename=self.plc_spec['slices'][0]['nodenames'][0] + (site,node) = self.locate_node(nodename) + test_site = TestSite (self, site) + test_node = TestNode (self, test_site,node) + return test_node + + def locate_first_sliver (self): + slice_spec=self.plc_spec['slices'][0] + slicename=slice_spec['slice_fields']['name'] + nodename=slice_spec['nodenames'][0] + return self.locate_sliver_obj(nodename,slicename) + # all different hostboxes used in this plc def gather_hostBoxes(self): # maps on sites and nodes, return [ (host_box,test_node) ] @@ -254,89 +267,94 @@ class TestPlc: node.kill_qemu() return True - #################### step methods - - ### uninstall - def uninstall_chroot(self): - self.run_in_host('service plc safestop') - #####detecting the last myplc version installed and remove it - self.run_in_host('rpm -e myplc') - ##### Clean up the /plc directory - self.run_in_host('rm -rf /plc/data') - ##### stop any running vservers - self.run_in_host('for vserver in $(ls -d /vservers/* | sed -e s,/vservers/,,) ; do case $vserver in vtest*) echo Shutting down vserver $vserver ; vserver $vserver stop ;; esac ; done') + + ### utility methods for handling the pool of IP addresses allocated to plcs + # Logic + # (*) running plcs are recorded in the file named ~/running-test-plcs + # (*) this file contains a line for each running plc, older first + # (*) each line contains the vserver name + the hostname of the (vserver) testbox where it sits + # (*) the free_tracker method performs a vserver stop on the oldest entry + # (*) the record_tracker method adds an entry at the bottom of the file + # (*) the cleanup_tracker method stops all known vservers and removes the tracker file + + TRACKER_FILE="~/running-test-plcs" + + def record_tracker (self): + command="echo %s %s >> %s"%(self.vservername,self.test_ssh.hostname,TestPlc.TRACKER_FILE) + (code,output) = utils.output_of (self.test_ssh.actual_command(command)) + if code != 0: + print "WARNING : COULD NOT record_tracker %s as a running plc on %s"%(self.vservername,self.test_ssh.hostname) + return False + print "Recorded %s in running plcs on host %s"%(self.vservername,self.test_ssh.hostname) return True - def uninstall_vserver(self): - self.run_in_host("vserver --silent %s delete"%self.vservername) + def free_tracker (self): + command="head -1 %s"%TestPlc.TRACKER_FILE + (code,line) = utils.output_of(self.test_ssh.actual_command(command)) + if code != 0: + print "No entry found in %s on %s"%(TestPlc.TRACKER_FILE,self.test_ssh.hostname) + return False + try: + [vserver_to_stop,hostname] = line.split() + except: + print "WARNING: free_tracker: Could not parse %s - skipped"%TestPlc.TRACKER_FILE + return False + stop_command = "vserver --silent %s stop"%vserver_to_stop + utils.system(self.test_ssh.actual_command(stop_command)) + x=TestPlc.TRACKER_FILE + flush_command = "tail --lines=+2 %s > %s.tmp ; mv %s.tmp %s"%(x,x,x,x) + utils.system(self.test_ssh.actual_command(flush_command)) return True + # this should/could stop only the ones in TRACKER_FILE if that turns out to be reliable + def cleanup_tracker (self): + stop_all = "cd /vservers ; for i in * ; do vserver --silent $i stop ; done" + utils.system(self.test_ssh.actual_command(stop_all)) + clean_tracker = "rm -f %s"%TestPlc.TRACKER_FILE + utils.system(self.test_ssh.actual_command(clean_tracker)) + def uninstall(self): - # if there's a chroot-based myplc running, and then a native-based myplc is being deployed - # it sounds safer to have the former uninstalled too - # now the vserver method cannot be invoked for chroot instances as vservername is required - if self.vserver: - self.uninstall_vserver() - self.uninstall_chroot() - else: - self.uninstall_chroot() + self.run_in_host("vserver --silent %s delete"%self.vservername) return True ### install - def install_chroot(self): - # nothing to do - return True - - def install_vserver(self): - # we need build dir for vtest-init-vserver + def install(self): if self.is_local(): # a full path for the local calls - build_dir=os.path(sys.argv[0])+"/build" + build_dir=os.path.dirname(sys.argv[0]) + # sometimes this is empty - set to "." in such a case + if not build_dir: build_dir="." + build_dir += "/build" else: - # use a standard name - will be relative to HOME - build_dir="options.buildname" + # 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 - # the repo url is taken from myplc-url - # with the last two steps (i386/myplc...) removed - repo_url = self.options.myplc_url - for level in [ 'rpmname','arch' ]: + # the repo url is taken from arch-rpms-url + # with the last step (i386.) removed + repo_url = self.options.arch_rpms_url + for level in [ 'arch' ]: repo_url = os.path.dirname(repo_url) - create_vserver="%s/vtest-init-vserver.sh %s %s -- --interface eth0:%s"%\ - (build_dir,self.vservername,repo_url,self.vserverip) - return self.run_in_host(create_vserver) == 0 - - def install(self): - if self.vserver: - return self.install_vserver() + if self.options.arch == "i386": + personality_option="-p linux32" else: - return self.install_chroot() - - ### install_rpm - make this an optional step - def cache_rpm(self): - url = self.options.myplc_url - rpm = os.path.basename(url) - cache_fetch="pwd;if [ -f %(rpm)s ] ; then echo Using cached rpm %(rpm)s ; else echo Fetching %(url)s ; curl -O %(url)s; fi"%locals() - return self.run_in_host(cache_fetch)==0 - - def install_rpm_chroot(self): - url = self.options.myplc_url - rpm = os.path.basename(url) - if not self.cache_rpm(): - return False - utils.header('Installing the : %s'%rpm) - return self.run_in_host('rpm -Uvh '+rpm)==0 and self.run_in_host('service plc mount')==0 - - def install_rpm_vserver(self): - return self.run_in_guest("yum -y install myplc-native")==0 + personality_option="-p linux64" + script="vtest-init-vserver.sh" + vserver_name = self.vservername + vserver_options="--netdev eth0 --interface %s"%self.vserverip + try: + vserver_hostname=socket.gethostbyaddr(self.vserverip)[0] + vserver_options += " --hostname %s"%vserver_hostname + except: + pass + create_vserver="%(build_dir)s/%(script)s %(personality_option)s %(vserver_name)s %(repo_url)s -- %(vserver_options)s"%locals() + return self.run_in_host(create_vserver) == 0 + ### install_rpm def install_rpm(self): - if self.vserver: - return self.install_rpm_vserver() - else: - return self.install_rpm_chroot() + return self.run_in_guest("yum -y install myplc-native")==0 ### def configure(self): @@ -362,22 +380,19 @@ class TestPlc: utils.system('rm %s'%tmpname) return True - # the chroot install is slightly different to this respect def start(self): - if self.vserver: - self.run_in_guest('service plc start') - else: - self.run_in_host('service plc start') + self.run_in_guest('service plc start') return True - + def stop(self): - if self.vserver: - self.run_in_guest('service plc stop') - else: - self.run_in_host('service plc stop') + self.run_in_guest('service plc stop') return True - # could use a TestKey class + def vs_start (self): + self.start_guest() + return True + + # stores the keys from the config for further use def store_keys(self): for key_spec in self.plc_spec['keys']: TestKey(self,key_spec).store_key() @@ -386,6 +401,21 @@ class TestPlc: def clean_keys(self): utils.system("rm -rf %s/keys/"%os.path(sys.argv[0])) + # 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): + dir="./keys" + if not os.path.isdir(dir): + os.mkdir(dir) + prefix = 'root_ssh_key' + vservername=self.vservername + overall=True + for ext in [ 'pub', 'rsa' ] : + src="/vservers/%(vservername)s/etc/planetlab/%(prefix)s.%(ext)s"%locals() + dst="keys/%(vservername)s.%(ext)s"%locals() + if self.test_ssh.fetch(src,dst) != 0: overall=False + return overall + def sites (self): return self.do_sites() @@ -407,6 +437,13 @@ class TestPlc: test_site.create_users() return True + def clean_all_sites (self): + 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) + def nodes (self): return self.do_nodes() def clean_nodes (self): @@ -429,9 +466,13 @@ class TestPlc: test_node.create_node () return True - # create nodegroups if needed, and populate - # no need for a clean_nodegroups if we are careful enough def nodegroups (self): + return self.do_nodegroups("add") + def clean_nodegroups (self): + return self.do_nodegroups("delete") + + # create nodegroups if needed, and populate + def do_nodegroups (self, action="add"): # 1st pass to scan contents groups_dict = {} for site_spec in self.plc_spec['sites']: @@ -447,14 +488,54 @@ class TestPlc: groups_dict[nodegroupname]=[] groups_dict[nodegroupname].append(test_node.name()) auth=self.auth_root() + overall = True for (nodegroupname,group_nodes) in groups_dict.iteritems(): - try: - self.apiserver.GetNodeGroups(auth,{'name':nodegroupname})[0] - except: - self.apiserver.AddNodeGroup(auth,{'name':nodegroupname}) - for node in group_nodes: - self.apiserver.AddNodeToNodeGroup(auth,node,nodegroupname) - return True + if action == "add": + print 'nodegroups:','dealing with nodegroup',nodegroupname,'on nodes',group_nodes + # first, check if the nodetagtype is here + tag_types = self.apiserver.GetTagTypes(auth,{'tagname':nodegroupname}) + if tag_types: + tag_type_id = tag_types[0]['tag_type_id'] + else: + tag_type_id = self.apiserver.AddTagType(auth, + {'tagname':nodegroupname, + 'description': 'for nodegroup %s'%nodegroupname, + 'category':'test', + 'min_role_id':10}) + print 'located tag (type)',nodegroupname,'as',tag_type_id + # create nodegroup + nodegroups = self.apiserver.GetNodeGroups (auth, {'groupname':nodegroupname}) + if not nodegroups: + self.apiserver.AddNodeGroup(auth, nodegroupname, tag_type_id, 'yes') + print 'created nodegroup',nodegroupname,'from tagname',nodegroupname,'and value','yes' + # set node tag on all nodes, value='yes' + for nodename in group_nodes: + try: + self.apiserver.AddNodeTag(auth, nodename, nodegroupname, "yes") + except: + traceback.print_exc() + print 'node',nodename,'seems to already have tag',nodegroupname + # check anyway + try: + expect_yes = self.apiserver.GetNodeTags(auth, + {'hostname':nodename, + 'tagname':nodegroupname}, + ['tagvalue'])[0]['tagvalue'] + if expect_yes != "yes": + print 'Mismatch node tag on node',nodename,'got',expect_yes + overall=False + except: + if not self.options.dry_run: + print 'Cannot find tag',nodegroupname,'on node',nodename + overall = False + else: + try: + print 'cleaning nodegroup',nodegroupname + self.apiserver.DeleteNodeGroup(auth,nodegroupname) + except: + traceback.print_exc() + overall=False + return overall def all_hostnames (self) : hostnames = [] @@ -464,7 +545,7 @@ class TestPlc: return hostnames # gracetime : during the first minutes nothing gets printed - def do_nodes_booted (self, minutes, gracetime=2): + def do_nodes_booted (self, minutes, gracetime,period=15): if self.options.dry_run: print 'dry_run' return True @@ -492,7 +573,7 @@ class TestPlc: utils.header("WARNING - Real node %s in %s - ignored"%(hostname,boot_state)) # let's cheat boot_state = 'boot' - if datetime.datetime.now() > graceout: + elif datetime.datetime.now() > graceout: utils.header ("%s still in '%s' state"%(hostname,boot_state)) graceout=datetime.datetime.now()+datetime.timedelta(1) status[hostname] = boot_state @@ -505,17 +586,17 @@ class TestPlc: utils.header("FAILURE due to %s in '%s' state"%(hostname,status[hostname])) return False # otherwise, sleep for a while - time.sleep(15) + time.sleep(period) # only useful in empty plcs return True def nodes_booted(self): - return self.do_nodes_booted(minutes=0) - + return self.do_nodes_booted(minutes=20,gracetime=15) - def do_nodes_ssh(self,minutes): + def do_nodes_ssh(self,minutes,gracetime,period=15): # compute timeout timeout = datetime.datetime.now()+datetime.timedelta(minutes=minutes) + graceout = datetime.datetime.now()+datetime.timedelta(minutes=gracetime) tocheck = self.all_hostnames() # self.scan_publicKeys(tocheck) utils.header("checking Connectivity on nodes %r"%tocheck) @@ -523,8 +604,8 @@ class TestPlc: for hostname in tocheck: # try to ssh in nodes node_test_ssh = TestSsh (hostname,key="/etc/planetlab/root_ssh_key.rsa") - access=self.run_in_guest(node_test_ssh.actual_command("date")) - if not access: + success=self.run_in_guest(node_test_ssh.actual_command("hostname"))==0 + if success: utils.header('The node %s is sshable -->'%hostname) # refresh tocheck tocheck.remove(hostname) @@ -534,6 +615,8 @@ class TestPlc: 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) + elif datetime.datetime.now() > graceout: + utils.header("Could not ssh-enter root context on %s"%hostname) if not tocheck: return True if datetime.datetime.now() > timeout: @@ -541,12 +624,12 @@ class TestPlc: utils.header("FAILURE to ssh into %s"%hostname) return False # otherwise, sleep for a while - time.sleep(15) + time.sleep(period) # only useful in empty plcs return True def nodes_ssh(self): - return self.do_nodes_ssh(minutes=2) + return self.do_nodes_ssh(minutes=10,gracetime=5) @node_mapper def init_node (self): pass @@ -559,6 +642,16 @@ class TestPlc: @node_mapper def export_qemu (self): pass + ### check sanity : invoke scripts from qaapi/qa/tests/{node,slice} + def check_sanity_node (self): + return self.locate_first_node().check_sanity() + def check_sanity_sliver (self) : + return self.locate_first_sliver().check_sanity() + + def check_sanity (self): + return self.check_sanity_node() and self.check_sanity_sliver() + + ### initscripts def do_check_initscripts(self): overall = True for slice_spec in self.plc_spec['slices']: @@ -585,6 +678,18 @@ class TestPlc: self.apiserver.AddInitScript(self.auth_root(),initscript['initscript_fields']) return True + def clean_initscripts (self): + for initscript in self.plc_spec['initscripts']: + initscript_name = initscript['initscript_fields']['name'] + print('Attempting to delete %s in plc %s'%(initscript_name,self.plc_spec['name'])) + try: + self.apiserver.DeleteInitScript(self.auth_root(),initscript_name) + print initscript_name,'deleted' + except: + print 'deletion went wrong - probably did not exist' + return True + + ### manage slices def slices (self): return self.do_slices() @@ -614,24 +719,6 @@ class TestPlc: @node_mapper def start_node (self) : pass - def all_sliver_objs (self): - result=[] - for slice_spec in self.plc_spec['slices']: - slicename = slice_spec['slice_fields']['name'] - for nodename in slice_spec['nodenames']: - result.append(self.locate_sliver_obj (nodename,slicename)) - return result - - def locate_sliver_obj (self,nodename,slicename): - (site,node) = self.locate_node(nodename) - slice = self.locate_slice (slicename) - # build objects - test_site = TestSite (self, site) - test_node = TestNode (self, test_site,node) - # xxx the slice site is assumed to be the node site - mhh - probably harmless - test_slice = TestSlice (self, test_site, slice) - return TestSliver (self, test_node, test_slice) - def check_tcp (self): specs = self.plc_spec['tcp_test'] overall=True @@ -648,7 +735,17 @@ class TestPlc: if not c_test_sliver.run_tcp_client(s_test_sliver.test_node.name(),port): overall=False return overall - + + def plcsh_stress_test (self): + # install the stress-test in the plc image + location = "/usr/share/plc_api/plcsh-stress-test.py" + remote="/vservers/%s/%s"%(self.vservername,location) + self.test_ssh.copy_abs("plcsh-stress-test.py",remote) + command = location + command += " -- --check" + if self.options.small_test: + command += " --tiny" + return ( self.run_in_guest(command) == 0) def gather_logs (self): # (1) get the plc's /var/log and store it locally in logs/myplc.var-log./* @@ -682,9 +779,11 @@ class TestPlc: return True def gather_var_logs (self): + utils.system("mkdir -p logs/myplc.var-log.%s"%self.name()) to_plc = self.actual_command_in_guest("tar -C /var/log/ -cf - .") command = to_plc + "| tar -C logs/myplc.var-log.%s -xf -"%self.name() - utils.system("mkdir -p logs/myplc.var-log.%s"%self.name()) + utils.system(command) + command = "chmod a+r,a+x logs/myplc.var-log.%s/httpd"%self.name() utils.system(command) def gather_nodes_var_logs (self):