2 # Thierry Parmentelat <thierry.parmentelat@inria.fr>
3 # Copyright (C) 2010 INRIA
5 # #################### history
7 # see also Substrate.readme
9 # This is a complete rewrite of TestResources/Tracker/Pool
10 # we don't use trackers anymore and just probe/sense the running
11 # boxes to figure out where we are
12 # in order to implement some fairness in the round-robin allocation scheme
13 # we need an indication of the 'age' of each running entity,
14 # hence the 'timestamp-*' steps in TestPlc
16 # this should be much more flexible:
17 # * supports several plc boxes
18 # * supports several qemu guests per host
19 # * no need to worry about tracker being in sync or not
21 # #################### howto use
23 # each site is to write its own LocalSubstrate.py,
24 # (see e.g. LocalSubstrate.inria.py)
25 # LocalSubstrate.py is expected to be in /root on the testmaster box
28 # . the vserver-capable boxes used for hosting myplcs
29 # . and their admissible load (max # of myplcs)
30 # . the pool of DNS-names and IP-addresses available for myplcs
32 # . the kvm-qemu capable boxes to host qemu instances
33 # . and their admissible load (max # of myplcs)
34 # . the pool of DNS-names and IP-addresses available for nodes
36 # #################### implem. note
38 # this model relies on 'sensing' the substrate,
39 # i.e. probing all the boxes for their running instances of vservers and qemu
40 # this is how we get rid of tracker inconsistencies
41 # however there is a 'black hole' between the time where a given address is
42 # allocated and when it actually gets used/pingable
43 # this is why we still need a shared knowledge among running tests
44 # in a file named /root/starting
45 # this is connected to the Pool class
47 # ####################
56 from optparse import OptionParser
59 from TestSsh import TestSsh
60 from TestMapper import TestMapper
62 def header (message,banner=True):
63 if not message: return
64 if banner: print "===============",
68 def timestamp_sort(o1,o2): return o1.timestamp-o2.timestamp
70 def short_hostname (hostname):
71 return hostname.split('.')[0]
74 # allows to pick an available IP among a pool
75 # input is expressed as a list of tuples (hostname,ip,user_data)
76 # that can be searched iteratively for a free slot
78 # pool = [ (hostname1,user_data1),
79 # (hostname2,user_data2),
80 # (hostname3,user_data2),
81 # (hostname4,user_data4) ]
82 # assuming that ip1 and ip3 are taken (pingable), then we'd get
84 # pool.next_free() -> entry2
85 # pool.next_free() -> entry4
86 # pool.next_free() -> None
87 # that is, even if ip2 is not busy/pingable when the second next_free() is issued
90 def __init__ (self,hostname,userdata):
91 self.hostname=hostname
92 self.userdata=userdata
93 # slot holds 'busy' or 'free' or 'mine' or 'starting' or None
94 # 'mine' is for our own stuff, 'starting' from the concurrent tests
99 return "Pooled %s (%s) -> %s"%(self.hostname,self.userdata, self.status)
102 if self.status==None: return '?'
103 elif self.status=='busy': return '+'
104 elif self.status=='free': return '-'
105 elif self.status=='mine': return 'M'
106 elif self.status=='starting': return 'S'
109 if self.ip: return self.ip
110 ip=socket.gethostbyname(self.hostname)
116 def __init__ (self, tuples,message):
117 self.pool_items= [ PoolItem (hostname,userdata) for (hostname,userdata) in tuples ]
121 for i in self.pool_items: print i.line()
125 for i in self.pool_items: line += ' ' + i.char()
128 def _item (self, hostname):
129 for i in self.pool_items:
130 if i.hostname==hostname: return i
131 raise Exception ("Could not locate hostname %s in pool %s"%(hostname,self.message))
133 def retrieve_userdata (self, hostname):
134 return self._item(hostname).userdata
136 def get_ip (self, hostname):
137 try: return self._item(hostname).get_ip()
138 except: return socket.gethostbyname(hostname)
140 def set_mine (self, hostname):
142 self._item(hostname).status='mine'
144 print 'WARNING: host %s not found in IP pool %s'%(hostname,self.message)
146 def next_free (self):
147 for i in self.pool_items:
148 if i.status == 'free':
150 return (i.hostname,i.userdata)
153 # the place were other test instances tell about their not-yet-started
154 # instances, that go undetected through sensing
155 starting='/root/starting'
156 def add_starting (self, name):
157 try: items=[line.strip() for line in file(Pool.starting).readlines()]
159 if not name in items:
160 file(Pool.starting,'a').write(name+'\n')
161 for i in self.pool_items:
162 if i.hostname==name: i.status='mine'
164 # we load this after actual sensing;
165 def load_starting (self):
166 try: items=[line.strip() for line in file(Pool.starting).readlines()]
168 for i in self.pool_items:
169 if i.hostname in items:
170 if i.status=='free' : i.status='starting'
172 def release_my_starting (self):
173 for i in self.pool_items:
175 self.del_starting(i.hostname)
178 def del_starting (self, name):
179 try: items=[line.strip() for line in file(Pool.starting).readlines()]
182 f=file(Pool.starting,'w')
184 if item != name: f.write(item+'\n')
189 for item in self.pool_items:
190 if item.status is not None:
193 if self.check_ping (item.hostname):
201 print 'Sensing IP pool',self.message,
205 print 'After starting: IP pool'
208 # OS-dependent ping option (support for macos, for convenience)
209 ping_timeout_option = None
210 # returns True when a given hostname/ip responds to ping
211 def check_ping (self,hostname):
212 if not Pool.ping_timeout_option:
213 (status,osname) = commands.getstatusoutput("uname -s")
215 raise Exception, "TestPool: Cannot figure your OS name"
216 if osname == "Linux":
217 Pool.ping_timeout_option="-w"
218 elif osname == "Darwin":
219 Pool.ping_timeout_option="-t"
221 command="ping -c 1 %s 1 %s"%(Pool.ping_timeout_option,hostname)
222 (status,output) = commands.getstatusoutput(command)
227 def __init__ (self,hostname):
228 self.hostname=hostname
229 def shortname (self):
230 return short_hostname(self.hostname)
231 def test_ssh (self): return TestSsh(self.hostname,username='root',unknown_host=False)
232 def reboot (self, options):
233 self.test_ssh().run("shutdown -r now",message="Rebooting %s"%self.hostname,
234 dry_run=options.dry_run)
236 def run(self,argv,message=None,trash_err=False,dry_run=False):
244 return subprocess.call(argv)
246 return subprocess.call(argv,stderr=file('/dev/null','w'))
248 def run_ssh (self, argv, message, trash_err=False, dry_run=False):
249 ssh_argv = self.test_ssh().actual_argv(argv)
250 result=self.run (ssh_argv, message, trash_err, dry_run=dry_run)
252 print "WARNING: failed to run %s on %s"%(" ".join(argv),self.hostname)
255 def backquote (self, argv, trash_err=False):
257 result= subprocess.Popen(argv,stdout=subprocess.PIPE).communicate()[0]
259 result= subprocess.Popen(argv,stdout=subprocess.PIPE,stderr=file('/dev/null','w')).communicate()[0]
262 def backquote_ssh (self, argv, trash_err=False):
263 # first probe the ssh link
264 probe_argv=self.test_ssh().actual_argv(['hostname'])
265 hostname=self.backquote ( probe_argv, trash_err=True )
267 print "root@%s unreachable"%self.hostname
270 return self.backquote( self.test_ssh().actual_argv(argv), trash_err)
272 ############################################################
274 def __init__ (self, buildname, pid, buildbox):
275 self.buildname=buildname
276 self.buildbox=buildbox
279 def add_pid(self,pid):
280 self.pids.append(pid)
283 return "== %s == (pids=%r)"%(self.buildname,self.pids)
285 class BuildBox (Box):
286 def __init__ (self,hostname):
287 Box.__init__(self,hostname)
288 self.build_instances=[]
290 def add_build (self,buildname,pid):
291 for build in self.build_instances:
292 if build.buildname==buildname:
295 self.build_instances.append(BuildInstance(buildname, pid, self))
298 if not self.build_instances:
299 header ('No build process on %s (%s)'%(self.hostname,self.uptime()))
301 header ("Builds on %s (%s)"%(self.hostname,self.uptime()))
302 for b in self.build_instances:
303 header (b.line(),banner=False)
306 if hasattr(self,'_uptime') and self._uptime: return self._uptime
307 return '*undef* uptime'
309 # inspect box and find currently running builds
310 matcher=re.compile("\s*(?P<pid>[0-9]+).*-[bo]\s+(?P<buildname>[^\s]+)(\s|\Z)")
311 def sense(self,options):
316 command=['pkill','vbuild']
317 self.run_ssh(command,"Terminating vbuild processes",dry_run=options.dry_run)
321 self._uptime=self.backquote_ssh(command,trash_err=True).strip()
322 if not self._uptime: self._uptime='unreachable'
323 pids=self.backquote_ssh(['pgrep','vbuild'],trash_err=True)
325 command=['ps','-o','pid,command'] + [ pid for pid in pids.split("\n") if pid]
326 ps_lines=self.backquote_ssh (command).split('\n')
327 for line in ps_lines:
328 if not line.strip() or line.find('PID')>=0: continue
329 m=BuildBox.matcher.match(line)
331 date=time.strftime('%Y-%m-%d',time.localtime(time.time()))
332 buildname=m.group('buildname').replace('@DATE@',date)
333 self.add_build (buildname,m.group('pid'))
334 else: header('command %r returned line that failed to match'%command)
336 ############################################################
338 def __init__ (self, vservername, ctxid, plcbox):
339 self.vservername=vservername
345 def set_timestamp (self,timestamp): self.timestamp=timestamp
346 def set_now (self): self.timestamp=int(time.time())
347 def pretty_timestamp (self): return time.strftime("%Y-%m-%d:%H-%M",time.localtime(self.timestamp))
350 return self.vservername.split('-')[-1]
351 def buildname (self):
352 return self.vservername.rsplit('-',2)[0]
355 msg="== %s =="%(self.vplcname())
356 msg += " [=%s]"%self.vservername
357 if self.ctxid==0: msg+=" not (yet?) running"
358 else: msg+=" (ctx=%s)"%self.ctxid
359 if self.timestamp: msg += " @ %s"%self.pretty_timestamp()
360 else: msg += " *unknown timestamp*"
364 msg="vserver stopping %s on %s"%(self.vservername,self.plc_box.hostname)
365 self.plc_box.run_ssh(['vserver',self.vservername,'stop'],msg)
366 self.plc_box.forget(self)
369 def __init__ (self, hostname, max_plcs):
370 Box.__init__(self,hostname)
371 self.plc_instances=[]
372 self.max_plcs=max_plcs
374 def add_vserver (self,vservername,ctxid):
375 for plc in self.plc_instances:
376 if plc.vservername==vservername:
377 header("WARNING, duplicate myplc %s running on %s"%\
378 (vservername,self.hostname),banner=False)
380 self.plc_instances.append(PlcInstance(vservername,ctxid,self))
382 def forget (self, plc_instance):
383 self.plc_instances.remove(plc_instance)
385 # fill one slot even though this one is not started yet
386 def add_dummy (self, plcname):
387 dummy=PlcInstance('dummy_'+plcname,0,self)
389 self.plc_instances.append(dummy)
392 msg="%s [max=%d,%d free] (%s)"%(self.hostname, self.max_plcs,self.free_spots(),self.uname())
396 if not self.plc_instances:
397 header ('No vserver running on %s'%(self.line()))
399 header ("Active plc VMs on %s"%self.line())
400 self.plc_instances.sort(timestamp_sort)
401 for p in self.plc_instances:
402 header (p.line(),banner=False)
404 def free_spots (self):
405 return self.max_plcs - len(self.plc_instances)
408 if hasattr(self,'_uname') and self._uname: return self._uname
409 return '*undef* uname'
411 def plc_instance_by_vservername (self, vservername):
412 for p in self.plc_instances:
413 if p.vservername==vservername: return p
416 def sense (self, options):
422 self.run_ssh(['service','util-vserver','stop'],"Stopping all running vservers",
423 dry_run=options.dry_run)
426 self._uname=self.backquote_ssh(['uname','-r']).strip()
427 # try to find fullname (vserver_stat truncates to a ridiculously short name)
428 # fetch the contexts for all vservers on that box
429 map_command=['grep','.','/etc/vservers/*/context','/dev/null',]
430 context_map=self.backquote_ssh (map_command)
431 # at this point we have a set of lines like
432 # /etc/vservers/2010.01.20--k27-f12-32-vplc03/context:40144
434 for map_line in context_map.split("\n"):
435 if not map_line: continue
436 [path,xid] = map_line.split(':')
437 ctx_dict[xid]=os.path.basename(os.path.dirname(path))
438 # at this point ctx_id maps context id to vservername
440 command=['vserver-stat']
441 vserver_stat = self.backquote_ssh (command)
442 for vserver_line in vserver_stat.split("\n"):
443 if not vserver_line: continue
444 context=vserver_line.split()[0]
445 if context=="CTX": continue
446 longname=ctx_dict[context]
447 self.add_vserver(longname,context)
448 # print self.margin_outline(self.vplcname(longname)),"%(vserver_line)s [=%(longname)s]"%locals()
451 running_vsnames = [ i.vservername for i in self.plc_instances ]
452 command= ['grep','.']
453 command += ['/vservers/%s/timestamp'%vs for vs in running_vsnames]
454 command += ['/dev/null']
455 ts_lines=self.backquote_ssh(command,trash_err=True).split('\n')
456 for ts_line in ts_lines:
457 if not ts_line.strip(): continue
458 # expect /vservers/<vservername>/timestamp:<timestamp>
460 (_,__,vservername,tail)=ts_line.split('/')
461 (_,timestamp)=tail.split(':')
462 timestamp=int(timestamp)
463 p=self.plc_instance_by_vservername(vservername)
465 print 'WARNING unattached plc instance',ts_line
466 print 'was expecting to find',vservername,'in',[i.vservername for i in self.plc_instances]
468 p.set_timestamp(timestamp)
469 except: print 'WARNING, could not parse ts line',ts_line
474 ############################################################
476 def __init__ (self, nodename, pid, qemubox):
477 self.nodename=nodename
479 self.qemu_box=qemubox
484 def set_buildname (self,buildname): self.buildname=buildname
485 def set_timestamp (self,timestamp): self.timestamp=timestamp
486 def set_now (self): self.timestamp=int(time.time())
487 def pretty_timestamp (self): return time.strftime("%Y-%m-%d:%H-%M",time.localtime(self.timestamp))
490 msg = "== %s =="%(short_hostname(self.nodename))
491 msg += " [=%s]"%self.buildname
492 if self.pid: msg += " (pid=%s)"%self.pid
493 else: msg += " not (yet?) running"
494 if self.timestamp: msg += " @ %s"%self.pretty_timestamp()
495 else: msg += " *unknown timestamp*"
500 print "cannot kill qemu %s with pid==0"%self.nodename
502 msg="Killing qemu %s with pid=%s on box %s"%(self.nodename,self.pid,self.qemu_box.hostname)
503 self.qemu_box.run_ssh(['kill',"%s"%self.pid],msg)
504 self.qemu_box.forget(self)
508 def __init__ (self, hostname, max_qemus):
509 Box.__init__(self,hostname)
510 self.qemu_instances=[]
511 self.max_qemus=max_qemus
513 def add_node (self,nodename,pid):
514 for qemu in self.qemu_instances:
515 if qemu.nodename==nodename:
516 header("WARNING, duplicate qemu %s running on %s"%\
517 (nodename,self.hostname), banner=False)
519 self.qemu_instances.append(QemuInstance(nodename,pid,self))
521 def forget (self, qemu_instance):
522 self.qemu_instances.remove(qemu_instance)
524 # fill one slot even though this one is not started yet
525 def add_dummy (self, nodename):
526 dummy=QemuInstance('dummy_'+nodename,0,self)
528 self.qemu_instances.append(dummy)
531 msg="%s [max=%d,%d free] (%s)"%(self.hostname, self.max_qemus,self.free_spots(),self.driver())
535 if not self.qemu_instances:
536 header ('No qemu process on %s'%(self.line()))
538 header ("Active qemu processes on %s"%(self.line()))
539 self.qemu_instances.sort(timestamp_sort)
540 for q in self.qemu_instances:
541 header (q.line(),banner=False)
543 def free_spots (self):
544 return self.max_qemus - len(self.qemu_instances)
547 if hasattr(self,'_driver') and self._driver: return self._driver
548 return '*undef* driver'
550 def qemu_instance_by_pid (self,pid):
551 for q in self.qemu_instances:
552 if q.pid==pid: return q
555 def qemu_instance_by_nodename_buildname (self,nodename,buildname):
556 for q in self.qemu_instances:
557 if q.nodename==nodename and q.buildname==buildname:
561 matcher=re.compile("\s*(?P<pid>[0-9]+).*-cdrom\s+(?P<nodename>[^\s]+)\.iso")
562 def sense(self, options):
567 self.run_ssh(['pkill','qemu'],"Killing qemu instances",
568 dry_run=options.dry_run)
571 modules=self.backquote_ssh(['lsmod']).split('\n')
572 self._driver='*NO kqemu/kmv_intel MODULE LOADED*'
573 for module in modules:
574 if module.find('kqemu')==0:
575 self._driver='kqemu module loaded'
576 # kvm might be loaded without vkm_intel (we dont have AMD)
577 elif module.find('kvm_intel')==0:
578 self._driver='kvm_intel module loaded'
579 ########## find out running pids
580 pids=self.backquote_ssh(['pgrep','qemu'])
582 command=['ps','-o','pid,command'] + [ pid for pid in pids.split("\n") if pid]
583 ps_lines = self.backquote_ssh (command).split("\n")
584 for line in ps_lines:
585 if not line.strip() or line.find('PID') >=0 : continue
586 m=QemuBox.matcher.match(line)
587 if m: self.add_node (m.group('nodename'),m.group('pid'))
588 else: header('command %r returned line that failed to match'%command)
589 ########## retrieve alive instances and map to build
591 command=['grep','.','*/*/qemu.pid','/dev/null']
592 pid_lines=self.backquote_ssh(command,trash_err=True).split('\n')
593 for pid_line in pid_lines:
594 if not pid_line.strip(): continue
595 # expect <build>/<nodename>/qemu.pid:<pid>pid
597 (buildname,nodename,tail)=pid_line.split('/')
598 (_,pid)=tail.split(':')
599 q=self.qemu_instance_by_pid (pid)
601 q.set_buildname(buildname)
602 live_builds.append(buildname)
603 except: print 'WARNING, could not parse pid line',pid_line
604 # retrieve timestamps
605 command= ['grep','.']
606 command += ['%s/*/timestamp'%b for b in live_builds]
607 command += ['/dev/null']
608 ts_lines=self.backquote_ssh(command,trash_err=True).split('\n')
609 for ts_line in ts_lines:
610 if not ts_line.strip(): continue
611 # expect <build>/<nodename>/timestamp:<timestamp>
613 (buildname,nodename,tail)=ts_line.split('/')
614 nodename=nodename.replace('qemu-','')
615 (_,timestamp)=tail.split(':')
616 timestamp=int(timestamp)
617 q=self.qemu_instance_by_nodename_buildname(nodename,buildname)
619 print 'WARNING unattached qemu instance',ts_line,nodename,buildname
621 q.set_timestamp(timestamp)
622 except: print 'WARNING, could not parse ts line',ts_line
624 ############################################################
630 self.options=Options()
631 self.options.dry_run=False
632 self.options.verbose=False
633 self.options.reboot=False
634 self.options.soft=False
635 self.build_boxes = [ BuildBox(h) for h in self.build_boxes_spec() ]
636 self.plc_boxes = [ PlcBox (h,m) for (h,m) in self.plc_boxes_spec ()]
637 self.qemu_boxes = [ QemuBox (h,m) for (h,m) in self.qemu_boxes_spec ()]
638 self.all_boxes = self.build_boxes + self.plc_boxes + self.qemu_boxes
641 self.vplc_pool = Pool (self.vplc_ips(),"for vplcs")
642 self.vnode_pool = Pool (self.vnode_ips(),"for vnodes")
644 def fqdn (self, hostname):
645 if hostname.find('.')<0: return "%s.%s"%(hostname,self.domain())
648 # return True if actual sensing takes place
649 def sense (self,force=False):
650 if self._sensed and not force: return False
651 print 'Sensing local substrate...',
652 for b in self.all_boxes: b.sense(self.options)
657 def add_dummy_plc (self, plc_boxname, plcname):
658 for pb in self.plc_boxes:
659 if pb.hostname==plc_boxname:
660 pb.add_dummy(plcname)
661 def add_dummy_qemu (self, qemu_boxname, qemuname):
662 for qb in self.qemu_boxes:
663 if qb.hostname==qemu_boxname:
664 qb.add_dummy(qemuname)
667 def provision (self,plcs,options):
669 # attach each plc to a plc box and an IP address
670 plcs = [ self.provision_plc (plc,options) for plc in plcs ]
671 # attach each node/qemu to a qemu box with an IP address
672 plcs = [ self.provision_qemus (plc,options) for plc in plcs ]
673 # update the SFA spec accordingly
674 plcs = [ self.localize_sfa_rspec(plc,options) for plc in plcs ]
677 print '* Could not provision this test on current substrate','--',e,'--','exiting'
678 traceback.print_exc()
681 # it is expected that a couple of options like ips_bplc and ips_vplc
682 # are set or unset together
684 def check_options (x,y):
685 if not x and not y: return True
686 return len(x)==len(y)
688 # find an available plc box (or make space)
689 # and a free IP address (using options if present)
690 def provision_plc (self, plc, options):
692 assert Substrate.check_options (options.ips_bplc, options.ips_vplc)
694 #### let's find an IP address for that plc
698 # we don't check anything here,
699 # it is the caller's responsability to cleanup and make sure this makes sense
700 plc_boxname = options.ips_bplc.pop()
701 vplc_hostname=options.ips_vplc.pop()
703 if self.sense(): self.list_all()
706 # try to find an available IP
707 self.vplc_pool.sense()
708 couple=self.vplc_pool.next_free()
710 (vplc_hostname,unused)=couple
711 #### we need to find one plc box that still has a slot
713 # use the box that has max free spots for load balancing
714 for pb in self.plc_boxes:
717 plc_boxname=pb.hostname
719 # if there's no available slot in the plc_boxes, or we need a free IP address
720 # make space by killing the oldest running instance
721 if not plc_boxname or not vplc_hostname:
722 # find the oldest of all our instances
723 all_plc_instances=reduce(lambda x, y: x+y,
724 [ pb.plc_instances for pb in self.plc_boxes ],
726 all_plc_instances.sort(timestamp_sort)
728 plc_instance_to_kill=all_plc_instances[0]
731 if not plc_boxname: msg += " PLC boxes are full"
732 if not vplc_hostname: msg += " vplc IP pool exhausted"
733 raise Exception,"Could not make space for a PLC instance:"+msg
734 freed_plc_boxname=plc_instance_to_kill.plc_box.hostname
735 freed_vplc_hostname=plc_instance_to_kill.vplcname()
736 message='killing oldest plc instance = %s on %s'%(plc_instance_to_kill.line(),
738 plc_instance_to_kill.kill()
739 # use this new plcbox if that was the problem
741 plc_boxname=freed_plc_boxname
742 # ditto for the IP address
743 if not vplc_hostname:
744 vplc_hostname=freed_vplc_hostname
745 # record in pool as mine
746 self.vplc_pool.set_mine(vplc_hostname)
749 self.add_dummy_plc(plc_boxname,plc['name'])
750 vplc_ip = self.vplc_pool.get_ip(vplc_hostname)
751 self.vplc_pool.add_starting(vplc_hostname)
753 #### compute a helpful vserver name
754 # remove domain in hostname
755 vplc_short = short_hostname(vplc_hostname)
756 vservername = "%s-%d-%s" % (options.buildname,plc['index'],vplc_short)
757 plc_name = "%s_%s"%(plc['name'],vplc_short)
759 utils.header( 'PROVISION plc %s in box %s at IP %s as %s'%\
760 (plc['name'],plc_boxname,vplc_hostname,vservername))
762 #### apply in the plc_spec
764 # label=options.personality.replace("linux","")
765 mapper = {'plc': [ ('*' , {'host_box':plc_boxname,
766 # 'name':'%s-'+label,
768 'vservername':vservername,
770 'PLC_DB_HOST':vplc_hostname,
771 'PLC_API_HOST':vplc_hostname,
772 'PLC_BOOT_HOST':vplc_hostname,
773 'PLC_WWW_HOST':vplc_hostname,
774 'PLC_NET_DNS1' : self.network_settings() [ 'interface_fields:dns1' ],
775 'PLC_NET_DNS2' : self.network_settings() [ 'interface_fields:dns2' ],
780 # mappers only work on a list of plcs
781 return TestMapper([plc],options).map(mapper)[0]
784 def provision_qemus (self, plc, options):
786 assert Substrate.check_options (options.ips_bnode, options.ips_vnode)
788 test_mapper = TestMapper ([plc], options)
789 nodenames = test_mapper.node_names()
791 for nodename in nodenames:
793 if options.ips_vnode:
794 # as above, it's a rerun, take it for granted
795 qemu_boxname=options.ips_bnode.pop()
796 vnode_hostname=options.ips_vnode.pop()
798 if self.sense(): self.list_all()
801 # try to find an available IP
802 self.vnode_pool.sense()
803 couple=self.vnode_pool.next_free()
805 (vnode_hostname,unused)=couple
806 # find a physical box
808 # use the box that has max free spots for load balancing
809 for qb in self.qemu_boxes:
812 qemu_boxname=qb.hostname
814 # if we miss the box or the IP, kill the oldest instance
815 if not qemu_boxname or not vnode_hostname:
816 # find the oldest of all our instances
817 all_qemu_instances=reduce(lambda x, y: x+y,
818 [ qb.qemu_instances for qb in self.qemu_boxes ],
820 all_qemu_instances.sort(timestamp_sort)
822 qemu_instance_to_kill=all_qemu_instances[0]
825 if not qemu_boxname: msg += " QEMU boxes are full"
826 if not vnode_hostname: msg += " vnode IP pool exhausted"
827 raise Exception,"Could not make space for a QEMU instance:"+msg
828 freed_qemu_boxname=qemu_instance_to_kill.qemu_box.hostname
829 freed_vnode_hostname=short_hostname(qemu_instance_to_kill.nodename)
831 message='killing oldest qemu node = %s on %s'%(qemu_instance_to_kill.line(),
833 qemu_instance_to_kill.kill()
834 # use these freed resources where needed
836 qemu_boxname=freed_qemu_boxname
837 if not vnode_hostname:
838 vnode_hostname=freed_vnode_hostname
839 self.vnode_pool.set_mine(vnode_hostname)
841 self.add_dummy_qemu (qemu_boxname,nodename)
842 mac=self.vnode_pool.retrieve_userdata(vnode_hostname)
843 ip=self.vnode_pool.get_ip (vnode_hostname)
844 self.vnode_pool.add_starting(vnode_hostname)
846 vnode_fqdn = self.fqdn(vnode_hostname)
847 nodemap={'host_box':qemu_boxname,
848 'node_fields:hostname':vnode_fqdn,
849 'interface_fields:ip':ip,
850 'interface_fields:mac':mac,
852 nodemap.update(self.network_settings())
853 maps.append ( (nodename, nodemap) )
855 utils.header("PROVISION node %s in box %s at IP %s with MAC %s"%\
856 (nodename,qemu_boxname,vnode_hostname,mac))
858 return test_mapper.map({'node':maps})[0]
860 def localize_sfa_rspec (self,plc,options):
862 plc['sfa']['SFA_REGISTRY_HOST'] = plc['PLC_DB_HOST']
863 plc['sfa']['SFA_AGGREGATE_HOST'] = plc['PLC_DB_HOST']
864 plc['sfa']['SFA_SM_HOST'] = plc['PLC_DB_HOST']
865 plc['sfa']['SFA_PLC_DB_HOST'] = plc['PLC_DB_HOST']
866 plc['sfa']['SFA_PLC_URL'] = 'https://' + plc['PLC_API_HOST'] + ':443/PLCAPI/'
867 for site in plc['sites']:
868 for node in site['nodes']:
869 plc['sfa']['sfa_slice_rspec']['part4'] = node['node_fields']['hostname']
872 #################### release:
873 def release (self,options):
874 self.vplc_pool.release_my_starting()
875 self.vnode_pool.release_my_starting()
878 #################### show results for interactive mode
879 def get_box (self,box):
880 for b in self.build_boxes + self.plc_boxes + self.qemu_boxes:
881 if b.shortname()==box:
883 print "Could not find box %s"%box
888 for b in self.all_boxes: b.list()
890 def list_boxes(self,boxes):
891 print 'Partial Sensing',
895 b.sense(self.options)
903 # can be run as a utility to manage the local infrastructure
905 parser=OptionParser()
906 parser.add_option ('-r',"--reboot",action='store_true',dest='reboot',default=False,
907 help='reboot mode (use shutdown -r)')
908 parser.add_option ('-s',"--soft",action='store_true',dest='soft',default=False,
909 help='soft mode for reboot (vserver stop or kill qemus)')
910 parser.add_option ('-b',"--build",action='store_true',dest='builds',default=False,
911 help='add build boxes')
912 parser.add_option ('-p',"--plc",action='store_true',dest='plcs',default=False,
913 help='add plc boxes')
914 parser.add_option ('-q',"--qemu",action='store_true',dest='qemus',default=False,
915 help='add qemu boxes')
916 parser.add_option ('-v',"--verbose",action='store_true',dest='verbose',default=False,
918 parser.add_option ('-n',"--dry_run",action='store_true',dest='dry_run',default=False,
920 (self.options,args)=parser.parse_args()
923 if self.options.builds: boxes += [b.hostname for b in self.build_boxes]
924 if self.options.plcs: boxes += [b.hostname for b in self.plc_boxes]
925 if self.options.qemus: boxes += [b.hostname for b in self.qemu_boxes]
926 boxes=list(set(boxes))
931 self.list_boxes (boxes)