#
# #################### history
#
+# see also Substrate.readme
+#
# This is a complete rewrite of TestResources/Tracker/Pool
# we don't use trackers anymore and just probe/sense the running
# boxes to figure out where we are
print message
sys.stdout.flush()
-def timestamp_sort(o1,o2):
- if not o1.timestamp: return -1
- elif not o2.timestamp: return 1
- else: return o2.timestamp-o1.timestamp
+def timestamp_sort(o1,o2): return o1.timestamp-o2.timestamp
+
+def short_hostname (hostname):
+ return hostname.split('.')[0]
+
+####################
+# the place were other test instances tell about their not-yet-started
+# instances, that go undetected through sensing
+class Starting:
+
+ location='/root/starting'
+ def __init__ (self):
+ self.tuples=[]
+
+ def load (self):
+ try: self.tuples=[line.strip().split('@')
+ for line in file(Starting.location).readlines()]
+ except: self.tuples=[]
+
+ def vnames (self) :
+ self.load()
+ return [ x for (x,_) in self.tuples ]
+ def add (self, vname, bname):
+ if not vname in self.vnames():
+ file(Starting.location,'a').write("%s@%s\n"%(vname,bname))
+
+ def delete_vname (self, vname):
+ self.load()
+ if vname in self.vnames():
+ f=file(Starting.location,'w')
+ for (v,b) in self.tuples:
+ if v != vname: f.write("%s@%s\n"%(v,b))
+ f.close()
+
####################
# pool class
# allows to pick an available IP among a pool
def line(self):
return "Pooled %s (%s) -> %s"%(self.hostname,self.userdata, self.status)
+
+ def char (self):
+ if self.status==None: return '?'
+ elif self.status=='busy': return '+'
+ elif self.status=='free': return '-'
+ elif self.status=='mine': return 'M'
+ elif self.status=='starting': return 'S'
+
def get_ip(self):
if self.ip: return self.ip
ip=socket.gethostbyname(self.hostname)
class Pool:
- def __init__ (self, tuples,message):
- self.pool= [ PoolItem (h,u) for (h,u) in tuples ]
+ def __init__ (self, tuples,message, substrate):
+ self.pool_items= [ PoolItem (hostname,userdata) for (hostname,userdata) in tuples ]
self.message=message
-
- def sense (self):
- print 'Checking IP pool',self.message,
- for item in self.pool:
- if item.status is not None:
- continue
- if self.check_ping (item.hostname):
- item.status='busy'
- else:
- item.status='free'
- print 'Done'
+ # where to send notifications upon load_starting
+ self.substrate=substrate
def list (self):
- for i in self.pool: print i.line()
+ for i in self.pool_items: print i.line()
- def retrieve_userdata (self, hostname):
- for i in self.pool:
- if i.hostname==hostname: return i.userdata
- return None
+ def line (self):
+ line=self.message
+ for i in self.pool_items: line += ' ' + i.char()
+ return line
+
+ def _item (self, hostname):
+ for i in self.pool_items:
+ if i.hostname==hostname: return i
+ raise Exception ("Could not locate hostname %s in pool %s"%(hostname,self.message))
+
+ def retrieve_userdata (self, hostname):
+ return self._item(hostname).userdata
def get_ip (self, hostname):
- # use cached if in pool
- for i in self.pool:
- if i.hostname==hostname: return i.get_ip()
- # otherwise just ask dns again
- return socket.gethostbyname(hostname)
+ try: return self._item(hostname).get_ip()
+ except: return socket.gethostbyname(hostname)
+
+ def set_mine (self, hostname):
+ try:
+ self._item(hostname).status='mine'
+ except:
+ print 'WARNING: host %s not found in IP pool %s'%(hostname,self.message)
def next_free (self):
- for i in self.pool:
- if i.status in ['busy','mine','starting' ]: continue
- i.status='mine'
- return (i.hostname,i.userdata)
- raise Exception,"No IP address available in pool %s"%self.message
+ for i in self.pool_items:
+ if i.status == 'free':
+ i.status='mine'
+ return (i.hostname,i.userdata)
+ return None
+
+ ####################
+ # we have a starting instance of our own
+ def add_starting (self, vname, bname):
+ Starting().add(vname,bname)
+ for i in self.pool_items:
+ if i.hostname==vname: i.status='mine'
-# OS-dependent ping option (support for macos, for convenience)
+ # load the starting instances from the common file
+ # remember that might be ours
+ # return the list of (vname,bname) that are not ours
+ def load_starting (self):
+ starting=Starting()
+ starting.load()
+ new_tuples=[]
+ for (v,b) in starting.tuples:
+ for i in self.pool_items:
+ if i.hostname==v and i.status=='free':
+ i.status='starting'
+ new_tuples.append( (v,b,) )
+ return new_tuples
+
+ def release_my_starting (self):
+ for i in self.pool_items:
+ if i.status=='mine':
+ Starting().delete_vname (i.hostname)
+ i.status=None
+
+
+ ##########
+ def _sense (self):
+ for item in self.pool_items:
+ if item.status is not None:
+ print item.char(),
+ continue
+ if self.check_ping (item.hostname):
+ item.status='busy'
+ print '*',
+ else:
+ item.status='free'
+ print '.',
+
+ def sense (self):
+ print 'Sensing IP pool',self.message,
+ self._sense()
+ print 'Done'
+ for (vname,bname) in self.load_starting():
+ self.substrate.add_starting_dummy (bname, vname)
+ print 'After starting: IP pool'
+ print self.line()
+ # OS-dependent ping option (support for macos, for convenience)
ping_timeout_option = None
-# checks whether a given hostname/ip responds to ping
+ # returns True when a given hostname/ip responds to ping
def check_ping (self,hostname):
if not Pool.ping_timeout_option:
(status,osname) = commands.getstatusoutput("uname -s")
command="ping -c 1 %s 1 %s"%(Pool.ping_timeout_option,hostname)
(status,output) = commands.getstatusoutput(command)
- if status==0: print '+',
- else: print '-',
return status == 0
- # the place were other test instances tell about their not-yet-started
- # instances
- starting='/root/starting'
- def add_starting (self, name):
- try: items=[line.strip() for line in file(Pool.starting).readlines()]
- except: items=[]
- if not name in items:
- file(Pool.starting,'a').write(name+'\n')
- for i in self.pool:
- if i.hostname==name: i.status='mine'
-
- def load_starting (self):
- try: items=[line.strip() for line in file(Pool.starting).readlines()]
- except: items=[]
- for item in items:
- for i in self.pool:
- if i.hostname==item: i.status='starting'
-
- def release_my_fakes (self):
- for i in self.pool:
- print 'releasing-scanning','hostname',i.hostname,'status',i.status
- if i.status=='mine':
- self.del_starting(i.hostname)
- i.status=None
-
- def del_starting (self, name):
- try: items=[line.strip() for line in file(Pool.starting).readlines()]
- except: items=[]
- if name in items:
- f=file(Pool.starting,'w')
- for item in items:
- if item != name: f.write(item+'\n')
- f.close()
-
-
-
####################
class Box:
def __init__ (self,hostname):
self.hostname=hostname
- def simple_hostname (self):
- return self.hostname.split('.')[0]
+ self._probed=None
+ def shortname (self):
+ return short_hostname(self.hostname)
def test_ssh (self): return TestSsh(self.hostname,username='root',unknown_host=False)
- def reboot (self):
- self.test_ssh().run("shutdown -r now",message="Rebooting %s"%self.hostname)
+ def reboot (self, options):
+ self.test_ssh().run("shutdown -r now",message="Rebooting %s"%self.hostname,
+ dry_run=options.dry_run)
+
+ def uptime(self):
+ if hasattr(self,'_uptime') and self._uptime: return self._uptime
+ return '*undef* uptime'
+ def sense_uptime (self):
+ command=['uptime']
+ self._uptime=self.backquote_ssh(command,trash_err=True).strip()
+ if not self._uptime: self._uptime='unreachable'
def run(self,argv,message=None,trash_err=False,dry_run=False):
if dry_run:
else:
return subprocess.call(argv,stderr=file('/dev/null','w'))
- def run_ssh (self, argv, message, trash_err=False):
+ def run_ssh (self, argv, message, trash_err=False, dry_run=False):
ssh_argv = self.test_ssh().actual_argv(argv)
- result=self.run (ssh_argv, message, trash_err)
+ result=self.run (ssh_argv, message, trash_err, dry_run=dry_run)
if result!=0:
print "WARNING: failed to run %s on %s"%(" ".join(argv),self.hostname)
return result
def backquote (self, argv, trash_err=False):
+ # print 'running backquote',argv
if not trash_err:
- return subprocess.Popen(argv,stdout=subprocess.PIPE).communicate()[0]
+ result= subprocess.Popen(argv,stdout=subprocess.PIPE).communicate()[0]
else:
- return subprocess.Popen(argv,stdout=subprocess.PIPE,stderr=file('/dev/null','w')).communicate()[0]
+ result= subprocess.Popen(argv,stdout=subprocess.PIPE,stderr=file('/dev/null','w')).communicate()[0]
+ return result
- def backquote_ssh (self, argv, trash_err=False):
+ def probe (self):
+ if self._probed is not None: return self._probed
# first probe the ssh link
probe_argv=self.test_ssh().actual_argv(['hostname'])
- hostname=self.backquote ( probe_argv, trash_err=True )
- if not hostname:
- print "root@%s unreachable"%self.hostname
- return ''
- else:
- return self.backquote( self.test_ssh().actual_argv(argv), trash_err)
+ self._probed=self.backquote ( probe_argv, trash_err=True )
+ if not self._probed: print "root@%s unreachable"%self.hostname
+ return self._probed
+
+ # use argv=['bash','-c',"the command line"]
+ # if you have any shell-expanded arguments like *
+ # and if there's any chance the command is adressed to the local host
+ def backquote_ssh (self, argv, trash_err=False):
+ if not self.probe(): return ''
+ return self.backquote( self.test_ssh().actual_argv(argv), trash_err)
############################################################
class BuildInstance:
for b in self.build_instances:
header (b.line(),banner=False)
- def uptime(self):
- if hasattr(self,'_uptime') and self._uptime: return self._uptime
- return '*undef* uptime'
+ def reboot (self, options):
+ if not options.soft:
+ self.reboot(options)
+ else:
+ command=['pkill','vbuild']
+ self.run_ssh(command,"Terminating vbuild processes",dry_run=options.dry_run)
# inspect box and find currently running builds
matcher=re.compile("\s*(?P<pid>[0-9]+).*-[bo]\s+(?P<buildname>[^\s]+)(\s|\Z)")
- def sense(self,reboot=False,verbose=True):
- if reboot:
- self.reboot(box)
- return
+ def sense(self, options):
print 'b',
- command=['uptime']
- self._uptime=self.backquote_ssh(command,trash_err=True).strip()
- if not self._uptime: self._uptime='unreachable'
- pids=self.backquote_ssh(['pgrep','build'],trash_err=True)
+ self.sense_uptime()
+ pids=self.backquote_ssh(['pgrep','vbuild'],trash_err=True)
if not pids: return
command=['ps','-o','pid,command'] + [ pid for pid in pids.split("\n") if pid]
ps_lines=self.backquote_ssh (command).split('\n')
for line in ps_lines:
if not line.strip() or line.find('PID')>=0: continue
m=BuildBox.matcher.match(line)
- if m: self.add_build (m.group('buildname'),m.group('pid'))
+ if m:
+ date=time.strftime('%Y-%m-%d',time.localtime(time.time()))
+ buildname=m.group('buildname').replace('@DATE@',date)
+ self.add_build (buildname,m.group('pid'))
else: header('command %r returned line that failed to match'%command)
############################################################
self.ctxid=ctxid
self.plc_box=plcbox
# unknown yet
- self.timestamp=None
+ self.timestamp=0
def set_timestamp (self,timestamp): self.timestamp=timestamp
def set_now (self): self.timestamp=int(time.time())
def pretty_timestamp (self): return time.strftime("%Y-%m-%d:%H-%M",time.localtime(self.timestamp))
+ def vplcname (self):
+ return self.vservername.split('-')[-1]
+ def buildname (self):
+ return self.vservername.rsplit('-',2)[0]
+
def line (self):
- msg="== %s == (ctx=%s)"%(self.vservername,self.ctxid)
+ msg="== %s =="%(self.vplcname())
+ msg += " [=%s]"%self.vservername
+ if self.ctxid==0: msg+=" not (yet?) running"
+ else: msg+=" (ctx=%s)"%self.ctxid
if self.timestamp: msg += " @ %s"%self.pretty_timestamp()
else: msg += " *unknown timestamp*"
- if self.ctxid==0: msg+=" not (yet?) running"
return msg
def kill (self):
header ('No vserver running on %s'%(self.line()))
else:
header ("Active plc VMs on %s"%self.line())
+ self.plc_instances.sort(timestamp_sort)
for p in self.plc_instances:
header (p.line(),banner=False)
if p.vservername==vservername: return p
return None
- def sense (self, reboot=False, soft=False):
- if reboot:
- # remove mark for all running servers to avoid resurrection
- stop_command=['rm','-rf','/etc/vservers/*/apps/init/mark']
- self.run_ssh(stop_command,"Removing all vserver marks on %s"%self.hostname)
- if not soft:
- self.reboot()
- return
- else:
- self.run_ssh(['service','util-vserver','stop'],"Stopping all running vservers")
- return
+ def reboot (self, options):
+ if not options.soft:
+ self.reboot(options)
+ else:
+ self.run_ssh(['service','util-vserver','stop'],"Stopping all running vservers",
+ dry_run=options.dry_run)
+
+ def sense (self, options):
print 'p',
self._uname=self.backquote_ssh(['uname','-r']).strip()
# try to find fullname (vserver_stat truncates to a ridiculously short name)
if not vserver_line: continue
context=vserver_line.split()[0]
if context=="CTX": continue
- longname=ctx_dict[context]
- self.add_vserver(longname,context)
-# print self.margin_outline(self.vplcname(longname)),"%(vserver_line)s [=%(longname)s]"%locals()
+ try:
+ longname=ctx_dict[context]
+ self.add_vserver(longname,context)
+ except:
+ print 'WARNING: found ctx %s in vserver_stat but was unable to figure a corresp. vserver'%context
- # scan timestamps
+ # scan timestamps
+ running_vsnames = [ i.vservername for i in self.plc_instances ]
command= ['grep','.']
- command += ['/vservers/%s/timestamp'%b for b in ctx_dict.values()]
+ command += ['/vservers/%s.timestamp'%vs for vs in running_vsnames]
command += ['/dev/null']
ts_lines=self.backquote_ssh(command,trash_err=True).split('\n')
for ts_line in ts_lines:
if not ts_line.strip(): continue
- # expect /vservers/<vservername>/timestamp:<timestamp>
+ # expect /vservers/<vservername>.timestamp:<timestamp>
try:
- (_,__,vservername,tail)=ts_line.split('/')
- (_,timestamp)=tail.split(':')
+ (ts_file,timestamp)=ts_line.split(':')
+ ts_file=os.path.basename(ts_file)
+ (vservername,_)=os.path.splitext(ts_file)
timestamp=int(timestamp)
- q=self.plc_instance_by_vservername(vservername)
- if not q:
- print 'WARNING unattached plc instance',ts_line
+ p=self.plc_instance_by_vservername(vservername)
+ if not p:
+ print 'WARNING zombie plc',self.hostname,ts_line
+ print '... was expecting',vservername,'in',[i.vservername for i in self.plc_instances]
continue
- q.set_timestamp(timestamp)
+ p.set_timestamp(timestamp)
except: print 'WARNING, could not parse ts line',ts_line
self.qemu_box=qemubox
# not known yet
self.buildname=None
- self.timestamp=None
+ self.timestamp=0
def set_buildname (self,buildname): self.buildname=buildname
def set_timestamp (self,timestamp): self.timestamp=timestamp
def pretty_timestamp (self): return time.strftime("%Y-%m-%d:%H-%M",time.localtime(self.timestamp))
def line (self):
- msg = "== %s == (pid=%s)"%(self.nodename,self.pid)
- if self.buildname: msg += " <--> %s"%self.buildname
- else: msg += " *unknown build*"
+ msg = "== %s =="%(short_hostname(self.nodename))
+ msg += " [=%s]"%self.buildname
+ if self.pid: msg += " (pid=%s)"%self.pid
+ else: msg += " not (yet?) running"
if self.timestamp: msg += " @ %s"%self.pretty_timestamp()
else: msg += " *unknown timestamp*"
- if self.pid: msg += "pid=%s"%self.pid
- else: msg += " not (yet?) running"
return msg
def kill(self):
- if self.pid==0: print "cannot kill qemu %s with pid==0"%self.nodename
+ if self.pid==0:
+ print "cannot kill qemu %s with pid==0"%self.nodename
+ return
msg="Killing qemu %s with pid=%s on box %s"%(self.nodename,self.pid,self.qemu_box.hostname)
self.qemu_box.run_ssh(['kill',"%s"%self.pid],msg)
self.qemu_box.forget(self)
header ('No qemu process on %s'%(self.line()))
else:
header ("Active qemu processes on %s"%(self.line()))
+ self.qemu_instances.sort(timestamp_sort)
for q in self.qemu_instances:
header (q.line(),banner=False)
return q
return None
+ def reboot (self, options):
+ if not options.soft:
+ self.reboot(options)
+ else:
+ self.run_ssh(['pkill','qemu'],"Killing qemu instances",
+ dry_run=options.dry_run)
+
matcher=re.compile("\s*(?P<pid>[0-9]+).*-cdrom\s+(?P<nodename>[^\s]+)\.iso")
- def sense(self, reboot=False, soft=False):
- if reboot:
- if not soft:
- self.reboot()
- else:
- self.run_ssh(box,['pkill','qemu'],"Killing qemu instances")
- return
+ def sense(self, options):
print 'q',
modules=self.backquote_ssh(['lsmod']).split('\n')
self._driver='*NO kqemu/kmv_intel MODULE LOADED*'
live_builds.append(buildname)
except: print 'WARNING, could not parse pid line',pid_line
# retrieve timestamps
+ if not live_builds: return
command= ['grep','.']
command += ['%s/*/timestamp'%b for b in live_builds]
command += ['/dev/null']
timestamp=int(timestamp)
q=self.qemu_instance_by_nodename_buildname(nodename,buildname)
if not q:
- print 'WARNING unattached qemu instance',ts_line,nodename,buildname
+ print 'WARNING zombie qemu',self.hostname,ts_line
+ print '... was expecting (',short_hostname(nodename),buildname,') in',\
+ [ (short_hostname(i.nodename),i.buildname) for i in self.qemu_instances ]
continue
q.set_timestamp(timestamp)
except: print 'WARNING, could not parse ts line',ts_line
+####################
+class TestInstance:
+ def __init__ (self, buildname, pid=0):
+ self.pids=[]
+ if pid!=0: self.pid.append(pid)
+ self.buildname=buildname
+ # latest trace line
+ self.trace=''
+ # has a KO test
+ self.broken_steps=[]
+ self.timestamp = 0
+
+ def set_timestamp (self,timestamp): self.timestamp=timestamp
+ def set_now (self): self.timestamp=int(time.time())
+ def pretty_timestamp (self): return time.strftime("%Y-%m-%d:%H-%M",time.localtime(self.timestamp))
+
+
+ def add_pid (self,pid):
+ self.pids.append(pid)
+ def set_broken (self,plcindex, step):
+ self.broken_steps.append ( (plcindex, step,) )
+
+ def line (self):
+ double='=='
+ if self.pids: double='*'+double[1]
+ if self.broken_steps: double=double[0]+'B'
+ msg = " %s %s =="%(double,self.buildname)
+ if not self.pids: pass
+ elif len(self.pids)==1: msg += " (pid=%s)"%self.pids[0]
+ else: msg += " !!!pids=%s!!!"%self.pids
+ msg += " @%s"%self.pretty_timestamp()
+ if self.broken_steps:
+ msg += "\n BROKEN IN STEPS"
+ for (i,s) in self.broken_steps: msg += " %s@%s"%(s,i)
+ return msg
+
+class TestBox (Box):
+ def __init__ (self,hostname):
+ Box.__init__(self,hostname)
+ self.starting_ips=[]
+ self.test_instances=[]
+
+ def reboot (self, options):
+ # can't reboot a vserver VM
+ self.run_ssh (['pkill','run_log'],"Terminating current runs",
+ dry_run=options.dry_run)
+ self.run_ssh (['rm','-f',Starting.location],"Cleaning %s"%Starting.location,
+ dry_run=options.dry_run)
+
+ def get_test (self, buildname):
+ for i in self.test_instances:
+ if i.buildname==buildname: return i
+
+ # we scan ALL remaining test results, even the ones not running
+ def add_timestamp (self, buildname, timestamp):
+ i=self.get_test(buildname)
+ if i:
+ i.set_timestamp(timestamp)
+ else:
+ i=TestInstance(buildname,0)
+ i.set_timestamp(timestamp)
+ self.test_instances.append(i)
+
+ def add_running_test (self, pid, buildname):
+ i=self.get_test(buildname)
+ if not i:
+ self.test_instances.append (TestInstance (buildname,pid))
+ return
+ if i.pids:
+ print "WARNING: 2 concurrent tests run on same build %s"%buildname
+ i.add_pid (pid)
+
+ def add_broken (self, buildname, plcindex, step):
+ i=self.get_test(buildname)
+ if not i:
+ i=TestInstance(buildname)
+ self.test_instances.append(i)
+ i.set_broken(plcindex, step)
+
+ matcher_proc=re.compile (".*/proc/(?P<pid>[0-9]+)/cwd.*/root/(?P<buildname>[^/]+)$")
+ matcher_grep=re.compile ("/root/(?P<buildname>[^/]+)/logs/trace.*:TRACE:\s*(?P<plcindex>[0-9]+).*step=(?P<step>\S+).*")
+ def sense (self, options):
+ print 't',
+ self.sense_uptime()
+ self.starting_ips=[x for x in self.backquote_ssh(['cat',Starting.location], trash_err=True).strip().split('\n') if x]
+
+ # scan timestamps on all tests
+ # this is likely to not invoke ssh so we need to be a bit smarter to get * expanded
+ # xxx would make sense above too
+ command=['bash','-c',"grep . /root/*/timestamp /dev/null"]
+ ts_lines=self.backquote_ssh(command,trash_err=True).split('\n')
+ for ts_line in ts_lines:
+ if not ts_line.strip(): continue
+ # expect /root/<buildname>/timestamp:<timestamp>
+ try:
+ (ts_file,timestamp)=ts_line.split(':')
+ ts_file=os.path.dirname(ts_file)
+ buildname=os.path.basename(ts_file)
+ timestamp=int(timestamp)
+ t=self.add_timestamp(buildname,timestamp)
+ except: print 'WARNING, could not parse ts line',ts_line
+
+ command=['bash','-c',"grep KO /root/*/logs/trace* /dev/null" ]
+ trace_lines=self.backquote_ssh (command).split('\n')
+ for line in trace_lines:
+ if not line.strip(): continue
+ m=TestBox.matcher_grep.match(line)
+ if m:
+ buildname=m.group('buildname')
+ plcindex=m.group('plcindex')
+ step=m.group('step')
+ self.add_broken(buildname,plcindex, step)
+ else: header("command %r returned line that failed to match\n%s"%(command,line))
+
+ pids = self.backquote_ssh (['pgrep','run_log'],trash_err=True)
+ if not pids: return
+ command=['ls','-ld'] + ["/proc/%s/cwd"%pid for pid in pids.split("\n") if pid]
+ ps_lines=self.backquote_ssh (command).split('\n')
+ for line in ps_lines:
+ if not line.strip(): continue
+ m=TestBox.matcher_proc.match(line)
+ if m:
+ pid=m.group('pid')
+ buildname=m.group('buildname')
+ self.add_running_test(pid, buildname)
+ else: header("command %r returned line that failed to match\n%s"%(command,line))
+
+
+ def line (self):
+ return "%s (%s)"%(self.hostname,self.uptime())
+
+ def list (self):
+ if not self.test_instances:
+ header ("No known tests on %s"%self.line())
+ else:
+ header ("Known tests on %s"%self.line())
+ self.test_instances.sort(timestamp_sort)
+ for i in self.test_instances: print i.line()
+ if self.starting_ips:
+ header ("Starting IP addresses on %s"%self.line())
+ self.starting_ips.sort()
+ for starting in self.starting_ips: print starting
+
############################################################
class Options: pass
class Substrate:
- def test (self):
- self.sense()
-
def __init__ (self):
self.options=Options()
self.options.dry_run=False
self.options.verbose=False
- self.options.probe=True
- self.options.soft=True
+ self.options.reboot=False
+ self.options.soft=False
+ self.test_box = TestBox (self.test_box_spec())
self.build_boxes = [ BuildBox(h) for h in self.build_boxes_spec() ]
self.plc_boxes = [ PlcBox (h,m) for (h,m) in self.plc_boxes_spec ()]
self.qemu_boxes = [ QemuBox (h,m) for (h,m) in self.qemu_boxes_spec ()]
- self.all_boxes = self.build_boxes + self.plc_boxes + self.qemu_boxes
+ self.default_boxes = self.plc_boxes + self.qemu_boxes
+ self.all_boxes = self.build_boxes + [ self.test_box ] + self.plc_boxes + self.qemu_boxes
self._sensed=False
- self.vplc_pool = Pool (self.vplc_ips(),"for vplcs")
- self.vnode_pool = Pool (self.vnode_ips(),"for vnodes")
-
- self.vnode_pool.list()
-
+ self.vplc_pool = Pool (self.vplc_ips(),"for vplcs",self)
+ self.vnode_pool = Pool (self.vnode_ips(),"for vnodes",self)
-# def build_box_names (self):
-# return [ h for h in self.build_boxes_spec() ]
-# def plc_boxes (self):
-# return [ h for (h,m) in self.plc_boxes_spec() ]
-# def qemu_boxes (self):
-# return [ h for (h,m) in self.qemu_boxes_spec() ]
+ def fqdn (self, hostname):
+ if hostname.find('.')<0: return "%s.%s"%(hostname,self.domain())
+ return hostname
+ # return True if actual sensing takes place
def sense (self,force=False):
- if self._sensed and not force: return
+ if self._sensed and not force: return False
print 'Sensing local substrate...',
- for b in self.all_boxes: b.sense()
+ for b in self.default_boxes: b.sense(self.options)
print 'Done'
self._sensed=True
+ return True
+
+ def list (self):
+ for b in self.default_boxes:
+ b.list()
+
+ def add_dummy_plc (self, plc_boxname, plcname):
+ for pb in self.plc_boxes:
+ if pb.hostname==plc_boxname:
+ pb.add_dummy(plcname)
+ return True
+ def add_dummy_qemu (self, qemu_boxname, qemuname):
+ for qb in self.qemu_boxes:
+ if qb.hostname==qemu_boxname:
+ qb.add_dummy(qemuname)
+ return True
+
+ def add_starting_dummy (self, bname, vname):
+ return self.add_dummy_plc (bname, vname) or self.add_dummy_qemu (bname, vname)
##########
def provision (self,plcs,options):
try:
- self.sense()
- self.list_all()
# attach each plc to a plc box and an IP address
plcs = [ self.provision_plc (plc,options) for plc in plcs ]
# attach each node/qemu to a qemu box with an IP address
plcs = [ self.provision_qemus (plc,options) for plc in plcs ]
# update the SFA spec accordingly
plcs = [ self.localize_sfa_rspec(plc,options) for plc in plcs ]
+ self.list()
return plcs
except Exception, e:
print '* Could not provision this test on current substrate','--',e,'--','exiting'
traceback.print_exc()
sys.exit(1)
+ # it is expected that a couple of options like ips_bplc and ips_vplc
+ # are set or unset together
+ @staticmethod
+ def check_options (x,y):
+ if not x and not y: return True
+ return len(x)==len(y)
+
# find an available plc box (or make space)
# and a free IP address (using options if present)
def provision_plc (self, plc, options):
- #### we need to find one plc box that still has a slot
- plc_box=None
- max_free=0
- # use the box that has max free spots for load balancing
- for pb in self.plc_boxes:
- free=pb.free_spots()
- if free>max_free:
- plc_box=pb
- max_free=free
- # everything is already used
- if not plc_box:
- # find the oldest of all our instances
- all_plc_instances=reduce(lambda x, y: x+y,
- [ pb.plc_instances for pb in self.plc_boxes ],
- [])
- all_plc_instances.sort(timestamp_sort)
- plc_instance_to_kill=all_plc_instances[0]
- plc_box=plc_instance_to_kill.plc_box
- plc_instance_to_kill.kill()
- print 'killed oldest = %s on %s'%(plc_instance_to_kill.line(),
- plc_instance_to_kill.plc_box.hostname)
-
- utils.header( 'plc %s -> box %s'%(plc['name'],plc_box.line()))
- plc_box.add_dummy(plc['name'])
- #### OK we have a box to run in, let's find an IP address
- # look in options
+
+ assert Substrate.check_options (options.ips_bplc, options.ips_vplc)
+
+ #### let's find an IP address for that plc
+ # look in options
if options.ips_vplc:
+ # this is a rerun
+ # we don't check anything here,
+ # it is the caller's responsability to cleanup and make sure this makes sense
+ plc_boxname = options.ips_bplc.pop()
vplc_hostname=options.ips_vplc.pop()
else:
- self.vplc_pool.load_starting()
+ if self.sense(): self.list()
+ plc_boxname=None
+ vplc_hostname=None
+ # try to find an available IP
self.vplc_pool.sense()
- (vplc_hostname,unused)=self.vplc_pool.next_free()
+ couple=self.vplc_pool.next_free()
+ if couple:
+ (vplc_hostname,unused)=couple
+ #### we need to find one plc box that still has a slot
+ max_free=0
+ # use the box that has max free spots for load balancing
+ for pb in self.plc_boxes:
+ free=pb.free_spots()
+ if free>max_free:
+ plc_boxname=pb.hostname
+ max_free=free
+ # if there's no available slot in the plc_boxes, or we need a free IP address
+ # make space by killing the oldest running instance
+ if not plc_boxname or not vplc_hostname:
+ # find the oldest of all our instances
+ all_plc_instances=reduce(lambda x, y: x+y,
+ [ pb.plc_instances for pb in self.plc_boxes ],
+ [])
+ all_plc_instances.sort(timestamp_sort)
+ try:
+ plc_instance_to_kill=all_plc_instances[0]
+ except:
+ msg=""
+ if not plc_boxname: msg += " PLC boxes are full"
+ if not vplc_hostname: msg += " vplc IP pool exhausted"
+ raise Exception,"Could not make space for a PLC instance:"+msg
+ freed_plc_boxname=plc_instance_to_kill.plc_box.hostname
+ freed_vplc_hostname=plc_instance_to_kill.vplcname()
+ message='killing oldest plc instance = %s on %s'%(plc_instance_to_kill.line(),
+ freed_plc_boxname)
+ plc_instance_to_kill.kill()
+ # use this new plcbox if that was the problem
+ if not plc_boxname:
+ plc_boxname=freed_plc_boxname
+ # ditto for the IP address
+ if not vplc_hostname:
+ vplc_hostname=freed_vplc_hostname
+ # record in pool as mine
+ self.vplc_pool.set_mine(vplc_hostname)
+
+ #
+ self.add_dummy_plc(plc_boxname,plc['name'])
vplc_ip = self.vplc_pool.get_ip(vplc_hostname)
- self.vplc_pool.add_starting(vplc_hostname)
+ self.vplc_pool.add_starting(vplc_hostname, plc_boxname)
#### compute a helpful vserver name
# remove domain in hostname
- vplc_simple = vplc_hostname.split('.')[0]
- vservername = "%s-%d-%s" % (options.buildname,plc['index'],vplc_simple)
- plc_name = "%s_%s"%(plc['name'],vplc_simple)
+ vplc_short = short_hostname(vplc_hostname)
+ vservername = "%s-%d-%s" % (options.buildname,plc['index'],vplc_short)
+ plc_name = "%s_%s"%(plc['name'],vplc_short)
+
+ utils.header( 'PROVISION plc %s in box %s at IP %s as %s'%\
+ (plc['name'],plc_boxname,vplc_hostname,vservername))
#### apply in the plc_spec
# # informative
# label=options.personality.replace("linux","")
- mapper = {'plc': [ ('*' , {'hostname':plc_box.hostname,
+ mapper = {'plc': [ ('*' , {'host_box':plc_boxname,
# 'name':'%s-'+label,
'name': plc_name,
'vservername':vservername,
} ) ]
}
- utils.header("Attaching %s on IP %s in vserver %s"%(plc['name'],vplc_hostname,vservername))
+
# mappers only work on a list of plcs
return TestMapper([plc],options).map(mapper)[0]
##########
def provision_qemus (self, plc, options):
+
+ assert Substrate.check_options (options.ips_bnode, options.ips_vnode)
+
test_mapper = TestMapper ([plc], options)
nodenames = test_mapper.node_names()
maps=[]
for nodename in nodenames:
- #### similarly we want to find a qemu box that can host us
- qemu_box=None
- max_free=0
- # use the box that has max free spots for load balancing
- for qb in self.qemu_boxes:
- free=qb.free_spots()
- if free>max_free:
- qemu_box=qb
- max_free=free
- # everything is already used
- if not qemu_box:
- # find the oldest of all our instances
- all_qemu_instances=reduce(lambda x, y: x+y,
- [ qb.qemu_instances for qb in self.qemu_boxes ],
- [])
- all_qemu_instances.sort(timestamp_sort)
- qemu_instance_to_kill=all_qemu_instances[0]
- qemu_box=qemu_instance_to_kill.qemu_box
- qemu_instance_to_kill.kill()
- print 'killed oldest = %s on %s'%(qemu_instance_to_kill.line(),
- qemu_instance_to_kill.qemu_box.hostname)
-
- utils.header( 'node %s -> qemu box %s'%(nodename,qemu_box.line()))
- qemu_box.add_dummy(nodename)
- #### OK we have a box to run in, let's find an IP address
- # look in options
+
if options.ips_vnode:
+ # as above, it's a rerun, take it for granted
+ qemu_boxname=options.ips_bnode.pop()
vnode_hostname=options.ips_vnode.pop()
- mac=self.vnode_pool.retrieve_userdata(vnode_hostname)
- print 'case 1 hostname',vnode_hostname,'mac',mac
else:
- self.vnode_pool.load_starting()
+ if self.sense(): self.list()
+ qemu_boxname=None
+ vnode_hostname=None
+ # try to find an available IP
self.vnode_pool.sense()
- (vnode_hostname,mac)=self.vnode_pool.next_free()
- print 'case 2 hostname',vnode_hostname,'mac',mac
+ couple=self.vnode_pool.next_free()
+ if couple:
+ (vnode_hostname,unused)=couple
+ # find a physical box
+ max_free=0
+ # use the box that has max free spots for load balancing
+ for qb in self.qemu_boxes:
+ free=qb.free_spots()
+ if free>max_free:
+ qemu_boxname=qb.hostname
+ max_free=free
+ # if we miss the box or the IP, kill the oldest instance
+ if not qemu_boxname or not vnode_hostname:
+ # find the oldest of all our instances
+ all_qemu_instances=reduce(lambda x, y: x+y,
+ [ qb.qemu_instances for qb in self.qemu_boxes ],
+ [])
+ all_qemu_instances.sort(timestamp_sort)
+ try:
+ qemu_instance_to_kill=all_qemu_instances[0]
+ except:
+ msg=""
+ if not qemu_boxname: msg += " QEMU boxes are full"
+ if not vnode_hostname: msg += " vnode IP pool exhausted"
+ raise Exception,"Could not make space for a QEMU instance:"+msg
+ freed_qemu_boxname=qemu_instance_to_kill.qemu_box.hostname
+ freed_vnode_hostname=short_hostname(qemu_instance_to_kill.nodename)
+ # kill it
+ message='killing oldest qemu node = %s on %s'%(qemu_instance_to_kill.line(),
+ freed_qemu_boxname)
+ qemu_instance_to_kill.kill()
+ # use these freed resources where needed
+ if not qemu_boxname:
+ qemu_boxname=freed_qemu_boxname
+ if not vnode_hostname:
+ vnode_hostname=freed_vnode_hostname
+ self.vnode_pool.set_mine(vnode_hostname)
+
+ self.add_dummy_qemu (qemu_boxname,vnode_hostname)
+ mac=self.vnode_pool.retrieve_userdata(vnode_hostname)
ip=self.vnode_pool.get_ip (vnode_hostname)
- self.vnode_pool.add_starting(vnode_hostname)
+ self.vnode_pool.add_starting(vnode_hostname,qemu_boxname)
- if vnode_hostname.find('.')<0:
- vnode_hostname += "."+self.domain()
- nodemap={'host_box':qemu_box.hostname,
- 'node_fields:hostname':vnode_hostname,
+ vnode_fqdn = self.fqdn(vnode_hostname)
+ nodemap={'host_box':qemu_boxname,
+ 'node_fields:hostname':vnode_fqdn,
'interface_fields:ip':ip,
'interface_fields:mac':mac,
}
nodemap.update(self.network_settings())
maps.append ( (nodename, nodemap) )
- utils.header("Attaching %s on IP %s MAC %s"%(plc['name'],vnode_hostname,mac))
+ utils.header("PROVISION node %s in box %s at IP %s with MAC %s"%\
+ (nodename,qemu_boxname,vnode_hostname,mac))
return test_mapper.map({'node':maps})[0]
plc['sfa']['SFA_SM_HOST'] = plc['PLC_DB_HOST']
plc['sfa']['SFA_PLC_DB_HOST'] = plc['PLC_DB_HOST']
plc['sfa']['SFA_PLC_URL'] = 'https://' + plc['PLC_API_HOST'] + ':443/PLCAPI/'
- for site in plc['sites']:
- for node in site['nodes']:
- plc['sfa']['sfa_slice_rspec']['part4'] = node['node_fields']['hostname']
return plc
#################### release:
def release (self,options):
- self.vplc_pool.release_my_fakes()
- self.vnode_pool.release_my_fakes()
+ self.vplc_pool.release_my_starting()
+ self.vnode_pool.release_my_starting()
pass
#################### show results for interactive mode
- def list_all (self):
- self.sense()
- for b in self.all_boxes: b.list()
-
- def get_box (self,box):
- for b in self.build_boxes + self.plc_boxes + self.qemu_boxes:
- if b.simple_hostname()==box:
+ def get_box (self,boxname):
+ for b in self.build_boxes + self.plc_boxes + self.qemu_boxes + [self.test_box] :
+ if b.shortname()==boxname:
return b
- print "Could not find box %s"%box
+ print "Could not find box %s"%boxname
return None
- def list_box(self,box):
- b=self.get_box(box)
- if not b: return
- b.sense()
- b.list()
-
+ def list_boxes(self,box_or_names):
+ print 'Sensing',
+ for box in box_or_names:
+ if not isinstance(box,Box): box=self.get_box(box)
+ if not box: continue
+ box.sense(self.options)
+ print 'Done'
+ for box in box_or_names:
+ if not isinstance(box,Box): box=self.get_box(box)
+ if not box: continue
+ box.list()
+
+ def reboot_boxes(self,box_or_names):
+ for box in box_or_names:
+ if not isinstance(box,Box): box=self.get_box(box)
+ if not box: continue
+ box.reboot(self.options)
+
+ ####################
# can be run as a utility to manage the local infrastructure
def main (self):
parser=OptionParser()
- (options,args)=parser.parse_args()
- if not args:
- self.list_all()
- else:
- for box in args:
- self.list_box(box)
+ parser.add_option ('-r',"--reboot",action='store_true',dest='reboot',default=False,
+ help='reboot mode (use shutdown -r)')
+ parser.add_option ('-s',"--soft",action='store_true',dest='soft',default=False,
+ help='soft mode for reboot (vserver stop or kill qemus)')
+ parser.add_option ('-t',"--testbox",action='store_true',dest='testbox',default=False,
+ help='add test box')
+ parser.add_option ('-b',"--build",action='store_true',dest='builds',default=False,
+ help='add build boxes')
+ parser.add_option ('-p',"--plc",action='store_true',dest='plcs',default=False,
+ help='add plc boxes')
+ parser.add_option ('-q',"--qemu",action='store_true',dest='qemus',default=False,
+ help='add qemu boxes')
+ parser.add_option ('-a',"--all",action='store_true',dest='all',default=False,
+ help='address all known boxes, like -b -t -p -q')
+ parser.add_option ('-v',"--verbose",action='store_true',dest='verbose',default=False,
+ help='verbose mode')
+ parser.add_option ('-n',"--dry_run",action='store_true',dest='dry_run',default=False,
+ help='dry run mode')
+ (self.options,args)=parser.parse_args()
+
+ boxes=args
+ if self.options.testbox: boxes += [self.test_box]
+ if self.options.builds: boxes += self.build_boxes
+ if self.options.plcs: boxes += self.plc_boxes
+ if self.options.qemus: boxes += self.qemu_boxes
+ if self.options.all: boxes += self.all_boxes
+
+ # default scope is -b -p -q
+ if not boxes:
+ boxes = self.build_boxes + self.plc_boxes + self.qemu_boxes
+
+ if self.options.reboot: self.reboot_boxes (boxes)
+ else: self.list_boxes (boxes)