# Thierry Parmentelat <thierry.parmentelat@inria.fr>
# Copyright (C) 2010 INRIA
#
-import os, os.path
-import datetime
-import time
import sys
+import time
+import os, os.path
import traceback
-from types import StringTypes
import socket
+from datetime import datetime, timedelta
+from types import StringTypes
import utils
from TestSite import TestSite
from TestSsh import TestSsh
from TestApiserver import TestApiserver
from TestAuthSfa import TestAuthSfa
+from PlcapiUrlScanner import PlcapiUrlScanner
+from Completer import Completer, CompleterTask
# step methods must take (self) and return a boolean (options is a member of the class)
default_steps = [
'show', SEP,
'vs_delete','timestamp_vs','vs_create', SEP,
+# 'plc_install', 'mod_python', 'plc_configure', 'plc_start', SEP,
'plc_install', 'plc_configure', 'plc_start', SEP,
- 'keys_fetch', 'keys_store', 'keys_clear_known_hosts', 'speed_up_slices', SEP,
+ 'keys_fetch', 'keys_store', 'keys_clear_known_hosts', SEP,
+ 'plcapi_urls','speed_up_slices', SEP,
'initscripts', 'sites', 'nodes', 'slices', 'nodegroups', 'leases', SEP,
+# slices created under plcsh interactively seem to be fine but these ones don't have the tags
+# keep this our of the way for now
+# 'check_vsys_defaults', SEP,
'nodestate_reinstall', 'qemu_local_init','bootcd', 'qemu_local_config', SEP,
- 'qemu_export', 'qemu_kill_mine', 'qemu_start', 'timestamp_qemu', SEP,
+ 'qemu_kill_mine','qemu_clean_mine', 'qemu_export', 'qemu_start', 'timestamp_qemu', SEP,
'sfa_install_all', 'sfa_configure', 'cross_sfa_configure', 'sfa_start', 'sfa_import', SEPSFA,
'sfi_configure@1', 'sfa_add_site@1','sfa_add_pi@1', SEPSFA,
'sfa_add_user@1', 'sfa_update_user@1', 'sfa_add_slice@1', 'sfa_renew_slice@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_node_boot@1', 'node_bmlogs@1', 'ssh_slice', 'ssh_slice_basics', 'check_initscripts', SEP,
'ssh_slice_sfa@1', 'sfa_delete_slice@1', 'sfa_delete_user@1', SEPSFA,
'cross_check_tcp@1', 'check_system_slice', SEP,
'empty_slices', 'ssh_slice_off', 'fill_slices', SEP,
self.vservername=plc_spec['vservername']
self.url="https://%s:443/PLCAPI/"%plc_spec['vserverip']
self.apiserver=TestApiserver(self.url,options.dry_run)
+ (self.ssh_node_boot_timeout,self.ssh_node_boot_silent)=plc_spec['ssh_node_boot_timers']
+ (self.ssh_node_debug_timeout,self.ssh_node_debug_silent)=plc_spec['ssh_node_debug_timers']
def has_addresses_api (self):
return self.apiserver.has_method('AddIpAddress')
pass
def actual_command_in_guest (self,command):
- return self.test_ssh.actual_command(self.host_to_guest(command))
+ return self.test_ssh.actual_command(self.host_to_guest(command),dry_run=self.options.dry_run)
def start_guest (self):
- return utils.system(self.test_ssh.actual_command(self.start_guest_in_host()))
+ return utils.system(self.test_ssh.actual_command(self.start_guest_in_host(),dry_run=self.options.dry_run))
def stop_guest (self):
- return utils.system(self.test_ssh.actual_command(self.stop_guest_in_host()))
+ return utils.system(self.test_ssh.actual_command(self.stop_guest_in_host(),dry_run=self.options.dry_run))
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)
+ return self.test_ssh.run_in_buildname(command, dry_run=self.options.dry_run)
#command gets run in the plc's vm
def host_to_guest(self,command):
if self.options.plcs_use_lxc:
- return "ssh -o StrictHostKeyChecking=no %s %s"%(self.vserverip,command)
+ return "ssh -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null %s %s"%(self.vserverip,command)
else:
return "vserver %s exec %s"%(self.vservername,command)
def vm_root_in_host(self):
if self.options.plcs_use_lxc:
- return "/var/lib/lxc/%s/rootfs/"%(self.vservername)
+ return "/vservers/%s/rootfs/"%(self.vservername)
else:
return "/vservers/%s"%(self.vservername)
def vm_timestamp_path (self):
if self.options.plcs_use_lxc:
- return "/var/lib/lxc/%s/%s.timestamp"%(self.vservername,self.vservername)
+ return "/vservers/%s/%s.timestamp"%(self.vservername,self.vservername)
else:
return "/vservers/%s.timestamp"%(self.vservername)
#start/stop the vserver
def start_guest_in_host(self):
if self.options.plcs_use_lxc:
- return "lxc-start --daemon --name=%s"%(self.vservername)
+ return "virsh -c lxc:// start %s"%(self.vservername)
else:
return "vserver %s start"%(self.vservername)
def stop_guest_in_host(self):
if self.options.plcs_use_lxc:
- return "lxc-stop --name=%s"%(self.vservername)
+ return "virsh -c lxc:// destroy %s"%(self.vservername)
else:
return "vserver %s stop"%(self.vservername)
return self.locate_sliver_obj(nodename,slicename)
# all different hostboxes used in this plc
- def gather_hostBoxes(self):
+ def get_BoxNodes(self):
# maps on sites and nodes, return [ (host_box,test_node) ]
tuples=[]
for site_spec in self.plc_spec['sites']:
# a step for checking this stuff
def show_boxes (self):
'print summary of nodes location'
- for (box,nodes) in self.gather_hostBoxes().iteritems():
+ for (box,nodes) in self.get_BoxNodes().iteritems():
print box,":"," + ".join( [ node.name() for node in nodes ] )
return True
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():
+ for (box,nodes) in self.get_BoxNodes().iteritems():
# pass the first nodename, as we don't push template-qemu on testboxes
nodedir=nodes[0].nodedir()
TestBoxQemu(box,self.options.buildname).qemu_kill_all(nodedir)
# make this a valid step
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():
+ for (box,nodes) in self.get_BoxNodes().iteritems():
# this is the brute force version, kill all qemus on that host box
TestBoxQemu(box,self.options.buildname).qemu_list_all()
return True
- # kill only the right qemus
+ # kill only the qemus related to this test
def qemu_list_mine(self):
'list qemu instances for our nodes'
- for (box,nodes) in self.gather_hostBoxes().iteritems():
+ for (box,nodes) in self.get_BoxNodes().iteritems():
# the fine-grain version
for node in nodes:
node.list_qemu()
return True
+ # kill only the qemus related to this test
+ def qemu_clean_mine(self):
+ 'cleanup (rm -rf) qemu instances for our nodes'
+ for (box,nodes) in self.get_BoxNodes().iteritems():
+ # the fine-grain version
+ for node in nodes:
+ node.qemu_clean()
+ return True
+
# kill only the right qemus
def qemu_kill_mine(self):
'kill the qemu instances for our nodes'
- for (box,nodes) in self.gather_hostBoxes().iteritems():
+ for (box,nodes) in self.get_BoxNodes().iteritems():
# the fine-grain version
for node in nodes:
node.kill_qemu()
# guess local domain from hostname
if TestPlc.exported_id>1:
print "export GUESTHOSTNAME%d=%s"%(TestPlc.exported_id,self.plc_spec['vservername'])
- return
+ return True
TestPlc.exported_id+=1
domain=socket.gethostname().split('.',1)[1]
fqdn="%s.%s"%(self.plc_spec['host_box'],domain)
stamp_path=self.vm_timestamp_path()
self.run_in_host("rm -f %s"%stamp_path)
if self.options.plcs_use_lxc:
- self.run_in_host("lxc-stop --name %s"%self.vservername)
- self.run_in_host("lxc-destroy --name %s"%self.vservername)
+ self.run_in_host("virsh -c lxc:// destroy %s"%self.vservername)
+ self.run_in_host("virsh -c lxc:// undefine %s"%self.vservername)
+ self.run_in_host("rm -fr /vservers/%s"%self.vservername)
return True
else:
self.run_in_host("vserver --silent %s delete"%self.vservername)
pkgs_string=" ".join(pkgs_list)
return self.yum_install (pkgs_list)
+ ###
+ def mod_python(self):
+ """yum install mod_python, useful on f18 and above so as to avoid broken wsgi"""
+ return self.yum_install ( [ 'mod_python' ] )
+
###
def plc_configure(self):
"run plc-config-tty"
'PLC_RESERVATION_GRANULARITY',
'PLC_OMF_ENABLED',
'PLC_OMF_XMPP_SERVER',
+ 'PLC_VSYS_DEFAULTS',
]:
fileconf.write ('e %s\n%s\n'%(var,self.plc_spec[var]))
fileconf.write('w\n')
def delete_all_sites (self):
"Delete all sites in PLC, and related objects"
print 'auth_root',self.auth_root()
- sites = self.apiserver.GetSites(self.auth_root(), {}, ['site_id'])
+ sites = self.apiserver.GetSites(self.auth_root(), {}, ['site_id','login_base'])
for site in sites:
# keep automatic site - otherwise we shoot in our own foot, root_auth is not valid anymore
if site['login_base']==self.plc_spec['PLC_SLICE_PREFIX']: continue
return res
# silent_minutes : during the first <silent_minutes> minutes nothing gets printed
- def nodes_check_boot_state (self, target_boot_state, timeout_minutes, silent_minutes,period=15):
+ def nodes_check_boot_state (self, target_boot_state, timeout_minutes, silent_minutes,period_seconds=15):
if self.options.dry_run:
print 'dry_run'
return True
- # compute timeout
- timeout = datetime.datetime.now()+datetime.timedelta(minutes=timeout_minutes)
- graceout = datetime.datetime.now()+datetime.timedelta(minutes=silent_minutes)
+
+ class CompleterTaskBootState (CompleterTask):
+ def __init__ (self, test_plc,hostname):
+ self.test_plc=test_plc
+ self.hostname=hostname
+ self.last_boot_state='undef'
+ def actual_run (self):
+ try:
+ node = self.test_plc.apiserver.GetNodes(self.test_plc.auth_root(), [ self.hostname ],
+ ['boot_state'])[0]
+ self.last_boot_state = node['boot_state']
+ return self.last_boot_state == target_boot_state
+ except:
+ return False
+ def message (self):
+ return "CompleterTaskBootState with node %s"%self.hostname
+ def failure_message (self):
+ return "node %s in state %s - expected %s"%(self.hostname,self.last_boot_state,target_boot_state)
+
+ timeout = timedelta(minutes=timeout_minutes)
+ graceout = timedelta(minutes=silent_minutes)
+ period = timedelta(seconds=period_seconds)
# the nodes that haven't checked yet - start with a full list and shrink over time
- tocheck = self.all_hostnames()
- utils.header("checking nodes %r"%tocheck)
- # create a dict hostname -> status
- status = dict ( [ (hostname,'undef') for hostname in tocheck ] )
- while tocheck:
- # get their status
- tocheck_status=self.apiserver.GetNodes(self.auth_root(), tocheck, ['hostname','boot_state' ] )
- # update status
- for array in tocheck_status:
- hostname=array['hostname']
- boot_state=array['boot_state']
- if boot_state == target_boot_state:
- utils.header ("%s has reached the %s state"%(hostname,target_boot_state))
- else:
- # if it's a real node, never mind
- (site_spec,node_spec)=self.locate_hostname(hostname)
- if TestNode.is_real_model(node_spec['node_fields']['model']):
- utils.header("WARNING - Real node %s in %s - ignored"%(hostname,boot_state))
- # let's cheat
- boot_state = target_boot_state
- 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
- # refresh tocheck
- tocheck = [ hostname for (hostname,boot_state) in status.iteritems() if boot_state != target_boot_state ]
- if not tocheck:
- return True
- if datetime.datetime.now() > timeout:
- for hostname in tocheck:
- utils.header("FAILURE due to %s in '%s' state"%(hostname,status[hostname]))
- return False
- # otherwise, sleep for a while
- time.sleep(period)
- # only useful in empty plcs
- return True
+ utils.header("checking nodes boot state (expected %s)"%target_boot_state)
+ tasks = [ CompleterTaskBootState (self,hostname) \
+ for (hostname,_) in self.all_node_infos() ]
+ return Completer (tasks).run (timeout, graceout, period)
def nodes_booted(self):
return self.nodes_check_boot_state('boot',timeout_minutes=30,silent_minutes=28)
- 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)
+ def check_nodes_ssh(self,debug,timeout_minutes,silent_minutes,period_seconds=15):
+ class CompleterTaskNodeSsh (CompleterTask):
+ def __init__ (self, hostname, qemuname, boot_state, local_key):
+ self.hostname=hostname
+ self.qemuname=qemuname
+ self.boot_state=boot_state
+ self.local_key=local_key
+ def run (self, silent):
+ command = TestSsh (self.hostname,key=self.local_key).actual_command("hostname;uname -a")
+ return utils.system (command, silent=silent)==0
+ def failure_message (self):
+ return "Cannot reach %s @ %s in %s mode"%(self.hostname, self.qemuname, self.boot_state)
+
+ # various delays
+ timeout = timedelta(minutes=timeout_minutes)
+ graceout = timedelta(minutes=silent_minutes)
+ period = timedelta(seconds=period_seconds)
vservername=self.vservername
if debug:
message="debug"
else:
message="boot"
local_key = "keys/key_admin.rsa"
+ utils.header("checking ssh access to nodes (expected in %s mode)"%message)
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 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
- success = utils.system ( command, silent=datetime.datetime.now() < graceout)
- if success==0:
- utils.header('Successfully entered root@%s (%s)'%(hostname,message))
- # 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)
- node_infos.remove(node_info)
- if not node_infos:
- return True
- if datetime.datetime.now() > timeout:
- 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
+ tasks = [ CompleterTaskNodeSsh (nodename, qemuname, message, local_key) \
+ for (nodename,qemuname) in node_infos ]
+ return Completer (tasks).run (timeout, graceout, period)
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=10,silent_minutes=8)
+ return self.check_nodes_ssh(debug=True,
+ timeout_minutes=self.ssh_node_debug_timeout,
+ silent_minutes=self.ssh_node_debug_silent)
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=40,silent_minutes=38)
+ return self.check_nodes_ssh(debug=False,
+ timeout_minutes=self.ssh_node_boot_timeout,
+ silent_minutes=self.ssh_node_boot_silent)
+
+ def node_bmlogs(self):
+ "Checks that there's a non-empty dir. /var/log/bm/raw"
+ return utils.system(self.actual_command_in_guest("ls /var/log/bm/raw"))==0
@node_mapper
def qemu_local_init (self): pass
### initscripts
def do_check_initscripts(self):
- overall = True
+ class CompleterTaskInitscript (CompleterTask):
+ def __init__ (self, test_sliver, stamp):
+ self.test_sliver=test_sliver
+ self.stamp=stamp
+ def actual_run (self):
+ return self.test_sliver.check_initscript_stamp (self.stamp)
+ def message (self):
+ return "initscript checker for %s"%self.test_sliver.name()
+ def failure_message (self):
+ return "initscript stamp %s not found in sliver %s"%(self.stamp,self.test_sliver.name())
+
+ tasks=[]
for slice_spec in self.plc_spec['slices']:
if not slice_spec.has_key('initscriptstamp'):
continue
stamp=slice_spec['initscriptstamp']
+ slicename=slice_spec['slice_fields']['name']
for nodename in slice_spec['nodenames']:
+ print 'nodename',nodename,'slicename',slicename,'stamp',stamp
(site,node) = self.locate_node (nodename)
# xxx - passing the wrong site - probably harmless
test_site = TestSite (self,site)
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_stamp(stamp):
- overall = False
- return overall
+ tasks.append ( CompleterTaskInitscript (test_sliver, stamp))
+ return Completer (tasks).run (timedelta(minutes=5), timedelta(minutes=4), timedelta(seconds=10))
def check_initscripts(self):
"check that the initscripts have triggered"
def ssh_slice(self): pass
@slice_mapper
def ssh_slice_off (self): pass
+ @slice_mapper
+ def ssh_slice_basics(self): pass
+
+ @slice_mapper
+ def check_vsys_defaults(self): pass
@node_mapper
def keys_clear_known_hosts (self): pass
+ def plcapi_urls (self):
+ return PlcapiUrlScanner (self.auth_root(),ip=self.vserverip).scan()
+
def speed_up_slices (self):
"tweak nodemanager settings on all nodes using a conf file"
# create the template on the server-side
def check_drl (self): return self._check_system_slice ('drl')
# we have the slices up already here, so it should not take too long
- def _check_system_slice (self, slicename, timeout_minutes=5, period=15):
- timeout = datetime.datetime.now()+datetime.timedelta(minutes=timeout_minutes)
- test_nodes=self.all_nodes()
- while test_nodes:
- for test_node in test_nodes:
- if test_node._check_system_slice (slicename,dry_run=self.options.dry_run):
- utils.header ("ok")
- test_nodes.remove(test_node)
- else:
- print '.',
- if not test_nodes:
- return True
- if datetime.datetime.now () > timeout:
- for test_node in test_nodes:
- utils.header ("can't find system slice %s in %s"%(slicename,test_node.name()))
- return False
- time.sleep(period)
- return True
+ def _check_system_slice (self, slicename, timeout_minutes=5, period_seconds=15):
+ class CompleterTaskSystemSlice (CompleterTask):
+ def __init__ (self, test_node, dry_run):
+ self.test_node=test_node
+ self.dry_run=dry_run
+ def actual_run (self):
+ return self.test_node._check_system_slice (slicename, dry_run=self.dry_run)
+ def message (self):
+ return "System slice %s @ %s"%(slicename, self.test_node.name())
+ def failure_message (self):
+ return "COULD not find system slice %s @ %s"%(slicename, self.test_node.name())
+ timeout = timedelta(minutes=timeout_minutes)
+ silent = timedelta (0)
+ period = timedelta (seconds=period_seconds)
+ tasks = [ CompleterTaskSystemSlice (test_node, self.options.dry_run) \
+ for test_node in self.all_nodes() ]
+ return Completer (tasks) . run (timeout, silent, period)
def plcsh_stress_test (self):
"runs PLCAPI stress test, that checks Add/Update/Delete on all types - preserves contents"
self.run_in_guest("sfa-nuke.py")==0 or \
self.run_in_guest("sfa-nuke-plc.py")==0
+ def sfa_fsclean(self):
+ "cleanup /etc/sfa/trusted_roots and /var/lib/sfa"
+ self.run_in_guest("rm -rf /etc/sfa/trusted_roots /var/lib/sfa/authorities")
+ return True
+
def sfa_plcclean(self):
"cleans the PLC entries that were created as a side effect of running the script"
# ignore result
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()
+ raise Exception,"Cannot create config dir for plc %s"%self.name()
return dirname
def conffile(self,filename):
and self.test_ssh.copy_abs(reg_fname,'/%s/etc/sfa/registries.xml'%self.vm_root_in_host())==0
def sfa_import(self):
- "sfa-import-plc"
+ "use sfaadmin to import from plc"
auth=self.plc_spec['sfa']['SFA_REGISTRY_ROOT_AUTH']
return \
self.run_in_guest('sfaadmin reg import_registry')==0
if not isinstance(name,StringTypes):
raise Exception
except:
- t=datetime.datetime.now()
+ t=datetime.now()
d=t.date()
name=str(d)
return "/root/%s-%s.sql"%(database,name)
git://git.onelab.eu / tests.git / blobdiff