X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=system%2FSubstrate.py;h=73b66c613799298a144ab6cda6572f2437f4c2fe;hb=a30d5259f12b1783a86f18cf841140550cb223d8;hp=bf8551cda7672b2d52b2a30479a7bb2427499b1b;hpb=4079bffdc14d32e416cb15bb87ca1fcbc5895504;p=tests.git diff --git a/system/Substrate.py b/system/Substrate.py index bf8551c..73b66c6 100644 --- a/system/Substrate.py +++ b/system/Substrate.py @@ -59,6 +59,9 @@ import utils from TestSsh import TestSsh from TestMapper import TestMapper +# too painful to propagate this cleanly +verbose=None + def header (message,banner=True): if not message: return if banner: print "===============", @@ -69,6 +72,37 @@ 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 @@ -113,11 +147,13 @@ class PoolItem: class Pool: - def __init__ (self, tuples,message): + def __init__ (self, tuples,message, substrate): self.pool_items= [ PoolItem (hostname,userdata) for (hostname,userdata) in tuples ] self.message=message + # where to send notifications upon load_starting + self.substrate=substrate - def list (self): + def list (self, verbose=False): for i in self.pool_items: print i.line() def line (self): @@ -150,40 +186,34 @@ class Pool: return (i.hostname,i.userdata) return None - # the place were other test instances tell about their not-yet-started - # instances, that go undetected through sensing - 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') + #################### + # 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==name: i.status='mine' - - # we load this after actual sensing; + if i.hostname==vname: i.status='mine' + + # 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): - try: items=[line.strip() for line in file(Pool.starting).readlines()] - except: items=[] - for i in self.pool_items: - if i.hostname in items: - if i.status=='free' : i.status='starting' + 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': - self.del_starting(i.hostname) + if i.status=='mine': + Starting().delete_vname (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() - + ########## def _sense (self): for item in self.pool_items: @@ -201,10 +231,10 @@ class Pool: print 'Sensing IP pool',self.message, self._sense() print 'Done' - self.load_starting() + 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 # returns True when a given hostname/ip responds to ping @@ -226,6 +256,7 @@ class Pool: class Box: def __init__ (self,hostname): self.hostname=hostname + self._probed=None def shortname (self): return short_hostname(self.hostname) def test_ssh (self): return TestSsh(self.hostname,username='root',unknown_host=False) @@ -233,6 +264,74 @@ class Box: self.test_ssh().run("shutdown -r now",message="Rebooting %s"%self.hostname, dry_run=options.dry_run) + def hostname_fedora (self,virt=None): + result = "%s {"%self.hostname + if virt: result += "%s-"%virt + result += "%s"%self.fedora() + # too painful to propagate this cleanly + global verbose + if verbose: + result += "-%s" % self.uname() + result += "}" + return result + + separator = "===composite===" + + # probe the ssh link + # take this chance to gather useful stuff + def probe (self): + # try it only once + if self._probed is not None: return self._probed + composite_command = [ ] + composite_command += [ "hostname" ] + composite_command += [ ";" , "echo", Box.separator , ";" ] + composite_command += [ "uptime" ] + composite_command += [ ";" , "echo", Box.separator , ";" ] + composite_command += [ "uname", "-r"] + composite_command += [ ";" , "echo", Box.separator , ";" ] + composite_command += [ "cat" , "/etc/fedora-release" ] + + # due to colons and all, this is going wrong on the local box (typically testmaster) + # I am reluctant to change TestSsh as it might break all over the place, so + if self.test_ssh().is_local(): + probe_argv = [ "bash", "-c", " ".join (composite_command) ] + else: + probe_argv=self.test_ssh().actual_argv(composite_command) + composite=self.backquote ( probe_argv, trash_err=True ) + self._hostname = self._uptime = self._uname = self._fedora = "** Unknown **" + if not composite: + print "root@%s unreachable"%self.hostname + self._probed='' + else: + try: + pieces = composite.split(Box.separator) + pieces = [ x.strip() for x in pieces ] + [self._hostname, self._uptime, self._uname, self._fedora] = pieces + # customize + self._uptime = ', '.join([ x.strip() for x in self._uptime.split(',')[2:]]) + self._fedora = self._fedora.replace("Fedora release ","f").split(" ")[0] + except: + import traceback + print 'BEG issue with pieces',pieces + traceback.print_exc() + print 'END issue with pieces',pieces + self._probed=self._hostname + return self._probed + + # use argv=['bash','-c',"the command line"] + def uptime(self): + self.probe() + if hasattr(self,'_uptime') and self._uptime: return self._uptime + return '*unprobed* uptime' + def uname(self): + self.probe() + if hasattr(self,'_uname') and self._uname: return self._uname + return '*unprobed* uname' + def fedora(self): + self.probe() + if hasattr(self,'_fedora') and self._fedora: return self._fedora + return '*unprobed* fedora' + def run(self,argv,message=None,trash_err=False,dry_run=False): if dry_run: print 'DRY_RUN:', @@ -253,21 +352,18 @@ class Box: return result def backquote (self, argv, trash_err=False): + # print 'running backquote',argv if not trash_err: result= subprocess.Popen(argv,stdout=subprocess.PIPE).communicate()[0] else: result= subprocess.Popen(argv,stdout=subprocess.PIPE,stderr=file('/dev/null','w')).communicate()[0] return result + # 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): - # 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) + if not self.probe(): return '' + return self.backquote( self.test_ssh().actual_argv(argv), trash_err) ############################################################ class BuildInstance: @@ -294,58 +390,96 @@ class BuildBox (Box): return self.build_instances.append(BuildInstance(buildname, pid, self)) - def list(self): + def list(self, verbose=False): if not self.build_instances: - header ('No build process on %s (%s)'%(self.hostname,self.uptime())) + header ('No build process on %s (%s)'%(self.hostname_fedora(),self.uptime())) else: - header ("Builds on %s (%s)"%(self.hostname,self.uptime())) + header ("Builds on %s (%s)"%(self.hostname_fedora(),self.uptime())) 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: + Box.reboot(self,options) + else: + self.soft_reboot (options) + +build_matcher=re.compile("\s*(?P[0-9]+).*-[bo]\s+(?P[^\s]+)(\s|\Z)") +build_matcher_initvm=re.compile("\s*(?P[0-9]+).*initvm.*\s+(?P[^\s]+)\s*\Z") + +class BuildVsBox (BuildBox): + def soft_reboot (self, options): + 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[0-9]+).*-[bo]\s+(?P[^\s]+)(\s|\Z)") - def sense(self,options): - if options.reboot: - if not options.soft: - self.reboot(options) - else: - command=['pkill','vbuild'] - self.run_ssh(command,"Terminating vbuild processes",dry_run=options.dry_run) - return - print 'b', - command=['uptime'] - self._uptime=self.backquote_ssh(command,trash_err=True).strip() - if not self._uptime: self._uptime='unreachable' + def sense(self, options): + print 'vb', 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) + m=build_matcher.match(line) 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) + continue + m=build_matcher_initvm.match(line) + if m: + # buildname is expansed here + self.add_build (buildname,m.group('pid')) + continue + header('BuildVsBox.sense: command %r returned line that failed to match'%command) + header(">>%s<<"%line) +class BuildLxcBox (BuildBox): + def soft_reboot (self, options): + command=['pkill','lbuild'] + self.run_ssh(command,"Terminating vbuild processes",dry_run=options.dry_run) + + # inspect box and find currently running builds + def sense(self, options): + print 'xb', + pids=self.backquote_ssh(['pgrep','lbuild'],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=build_matcher.match(line) + 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')) + continue + m=build_matcher_initvm.match(line) + if m: + # buildname is expansed here + self.add_build (buildname,m.group('pid')) + continue + header('BuildLxcBox.sense: command %r returned line that failed to match'%command) + header(">>%s<<"%line) + ############################################################ class PlcInstance: - def __init__ (self, vservername, ctxid, plcbox): - self.vservername=vservername - self.ctxid=ctxid + def __init__ (self, plcbox): self.plc_box=plcbox # unknown yet 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)) +class PlcVsInstance (PlcInstance): + def __init__ (self, plcbox, vservername, ctxid): + PlcInstance.__init__(self,plcbox) + self.vservername=vservername + self.ctxid=ctxid + def vplcname (self): return self.vservername.split('-')[-1] def buildname (self): @@ -365,65 +499,94 @@ class PlcInstance: self.plc_box.run_ssh(['vserver',self.vservername,'stop'],msg) self.plc_box.forget(self) +class PlcLxcInstance (PlcInstance): + # does lxc have a context id of any kind ? + def __init__ (self, plcbox, lxcname, pid): + PlcInstance.__init__(self, plcbox) + self.lxcname = lxcname + self.pid = pid + + def vplcname (self): + return self.lxcname.split('-')[-1] + def buildname (self): + return self.lxcname.rsplit('-',2)[0] + + def line (self): + msg="== %s =="%(self.vplcname()) + msg += " [=%s]"%self.lxcname + if self.pid==-1: msg+=" not (yet?) running" + else: msg+=" (pid=%s)"%self.pid + if self.timestamp: msg += " @ %s"%self.pretty_timestamp() + else: msg += " *unknown timestamp*" + return msg + + def kill (self): + command="rsync lxc-driver.sh %s:/root"%self.plc_box.hostname + commands.getstatusoutput(command) + msg="lxc container stopping %s on %s"%(self.lxcname,self.plc_box.hostname) + self.plc_box.run_ssh(['/root/lxc-driver.sh','-c','stop_lxc','-n',self.lxcname],msg) + self.plc_box.forget(self) + +########## class PlcBox (Box): def __init__ (self, hostname, max_plcs): Box.__init__(self,hostname) self.plc_instances=[] self.max_plcs=max_plcs - def add_vserver (self,vservername,ctxid): - for plc in self.plc_instances: - if plc.vservername==vservername: - header("WARNING, duplicate myplc %s running on %s"%\ - (vservername,self.hostname),banner=False) - return - self.plc_instances.append(PlcInstance(vservername,ctxid,self)) - - def forget (self, plc_instance): - self.plc_instances.remove(plc_instance) + def free_slots (self): + return self.max_plcs - len(self.plc_instances) # fill one slot even though this one is not started yet def add_dummy (self, plcname): - dummy=PlcInstance('dummy_'+plcname,0,self) + dummy=PlcVsInstance(self,'dummy_'+plcname,0) dummy.set_now() self.plc_instances.append(dummy) - def line(self): - msg="%s [max=%d,%d free] (%s)"%(self.hostname, self.max_plcs,self.free_spots(),self.uname()) - return msg - - def list(self): + def forget (self, plc_instance): + self.plc_instances.remove(plc_instance) + + def reboot (self, options): + if not options.soft: + Box.reboot(self,options) + else: + self.soft_reboot (options) + + def list(self, verbose=False): if not self.plc_instances: - header ('No vserver running on %s'%(self.line())) + header ('No plc 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) - def free_spots (self): - return self.max_plcs - len(self.plc_instances) - - def uname(self): - if hasattr(self,'_uname') and self._uname: return self._uname - return '*undef* uname' +# we do not this at INRIA any more +class PlcVsBox (PlcBox): + def add_vserver (self,vservername,ctxid): + for plc in self.plc_instances: + if plc.vservername==vservername: + header("WARNING, duplicate myplc %s running on %s"%\ + (vservername,self.hostname),banner=False) + return + self.plc_instances.append(PlcVsInstance(self,vservername,ctxid)) + + def line(self): + msg="%s [max=%d,free=%d] (%s)"%(self.hostname_fedora(virt="vs"), self.max_plcs,self.free_slots(),self.uptime()) + return msg + def plc_instance_by_vservername (self, vservername): for p in self.plc_instances: if p.vservername==vservername: return p return None + def soft_reboot (self, options): + self.run_ssh(['service','util-vserver','stop'],"Stopping all running vservers on %s"%(self.hostname,), + dry_run=options.dry_run) + def sense (self, options): - if options.reboot: - if not options.soft: - self.reboot(options) - return - else: - self.run_ssh(['service','util-vserver','stop'],"Stopping all running vservers", - dry_run=options.dry_run) - return - print 'p', - self._uname=self.backquote_ssh(['uname','-r']).strip() + print 'vp', # try to find fullname (vserver_stat truncates to a ridiculously short name) # fetch the contexts for all vservers on that box map_command=['grep','.','/etc/vservers/*/context','/dev/null',] @@ -443,9 +606,11 @@ class PlcBox (Box): 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 running_vsnames = [ i.vservername for i in self.plc_instances ] @@ -455,22 +620,75 @@ class PlcBox (Box): 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//timestamp: + # expect /vservers/.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) p=self.plc_instance_by_vservername(vservername) if not p: - print 'WARNING unattached plc instance',ts_line - print 'was expecting to find',vservername,'in',[i.vservername for i in self.plc_instances] + print 'WARNING zombie plc',self.hostname,ts_line + print '... was expecting',vservername,'in',[i.vservername for i in self.plc_instances] continue p.set_timestamp(timestamp) except: print 'WARNING, could not parse ts line',ts_line +class PlcLxcBox (PlcBox): + + def add_lxc (self,lxcname,pid): + for plc in self.plc_instances: + if plc.lxcname==lxcname: + header("WARNING, duplicate myplc %s running on %s"%\ + (lxcname,self.hostname),banner=False) + return + self.plc_instances.append(PlcLxcInstance(self,lxcname,pid)) + + + # a line describing the box + def line(self): + return "%s [max=%d,free=%d] (%s)"%(self.hostname_fedora(virt="lxc"), + self.max_plcs,self.free_slots(), + self.uptime(), + ) + + def plc_instance_by_lxcname (self, lxcname): + for p in self.plc_instances: + if p.lxcname==lxcname: return p + return None + + # essentially shutdown all running containers + def soft_reboot (self, options): + command="rsync lxc-driver.sh %s:/root"%self.hostname + commands.getstatusoutput(command) + self.run_ssh(['/root/lxc-driver.sh','-c','stop_all'],"Stopping all running lxc containers on %s"%(self.hostname,), + dry_run=options.dry_run) + # sense is expected to fill self.plc_instances with PlcLxcInstance's + # to describe the currently running VM's + def sense (self, options): + print "xp", + command="rsync lxc-driver.sh %s:/root"%self.hostname + commands.getstatusoutput(command) + command=['/root/lxc-driver.sh','-c','sense_all'] + lxc_stat = self.backquote_ssh (command) + for lxc_line in lxc_stat.split("\n"): + if not lxc_line: continue + lxcname=lxc_line.split(";")[0] + pid=lxc_line.split(";")[1] + timestamp=lxc_line.split(";")[2] + self.add_lxc(lxcname,pid) + try: timestamp=int(timestamp) + except: timestamp=0 + p=self.plc_instance_by_lxcname(lxcname) + if not p: + print 'WARNING zombie plc',self.hostname,lxcname + print '... was expecting',lxcname,'in',[i.lxcname for i in self.plc_instances] + continue + p.set_timestamp(timestamp) + ############################################################ class QemuInstance: def __init__ (self, nodename, pid, qemubox): @@ -528,19 +746,20 @@ class QemuBox (Box): self.qemu_instances.append(dummy) def line (self): - msg="%s [max=%d,%d free] (%s)"%(self.hostname, self.max_qemus,self.free_spots(),self.driver()) - return msg + return "%s [max=%d,free=%d] (%s) %s"%( + self.hostname_fedora(virt="qemu"), self.max_qemus,self.free_slots(), + self.uptime(),self.driver()) - def list(self): + def list(self, verbose=False): if not self.qemu_instances: - header ('No qemu process on %s'%(self.line())) + header ('No qemu on %s'%(self.line())) else: - header ("Active qemu processes on %s"%(self.line())) + header ("Qemus on %s"%(self.line())) self.qemu_instances.sort(timestamp_sort) for q in self.qemu_instances: header (q.line(),banner=False) - def free_spots (self): + def free_slots (self): return self.max_qemus - len(self.qemu_instances) def driver(self): @@ -558,24 +777,24 @@ class QemuBox (Box): return q return None + def reboot (self, options): + if not options.soft: + Box.reboot(self,options) + else: + self.run_ssh(['pkill','qemu'],"Killing qemu instances", + dry_run=options.dry_run) + matcher=re.compile("\s*(?P[0-9]+).*-cdrom\s+(?P[^\s]+)\.iso") def sense(self, options): - if options.reboot: - if not options.soft: - self.reboot(options) - else: - self.run_ssh(['pkill','qemu'],"Killing qemu instances", - dry_run=options.dry_run) - return - print 'q', + print 'qn', modules=self.backquote_ssh(['lsmod']).split('\n') - self._driver='*NO kqemu/kmv_intel MODULE LOADED*' + self._driver='*NO kqemu/kvm_intel MODULE LOADED*' for module in modules: if module.find('kqemu')==0: self._driver='kqemu module loaded' - # kvm might be loaded without vkm_intel (we dont have AMD) + # kvm might be loaded without kvm_intel (we dont have AMD) elif module.find('kvm_intel')==0: - self._driver='kvm_intel module loaded' + self._driver='kvm_intel OK' ########## find out running pids pids=self.backquote_ssh(['pgrep','qemu']) if not pids: return @@ -584,17 +803,20 @@ class QemuBox (Box): for line in ps_lines: if not line.strip() or line.find('PID') >=0 : continue m=QemuBox.matcher.match(line) - if m: self.add_node (m.group('nodename'),m.group('pid')) - else: header('command %r returned line that failed to match'%command) + if m: + self.add_node (m.group('nodename'),m.group('pid')) + continue + header('QemuBox.sense: command %r returned line that failed to match'%command) + header(">>%s<<"%line) ########## retrieve alive instances and map to build live_builds=[] - command=['grep','.','*/*/qemu.pid','/dev/null'] + command=['grep','.','/vservers/*/*/qemu.pid','/dev/null'] pid_lines=self.backquote_ssh(command,trash_err=True).split('\n') for pid_line in pid_lines: if not pid_line.strip(): continue # expect //qemu.pid:pid try: - (buildname,nodename,tail)=pid_line.split('/') + (_,__,buildname,nodename,tail)=pid_line.split('/') (_,pid)=tail.split(':') q=self.qemu_instance_by_pid (pid) if not q: continue @@ -602,25 +824,208 @@ class QemuBox (Box): 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 += ['/vservers/%s/*/timestamp'%b for b in live_builds] 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 //timestamp: try: - (buildname,nodename,tail)=ts_line.split('/') + (_,__,buildname,nodename,tail)=ts_line.split('/') nodename=nodename.replace('qemu-','') (_,timestamp)=tail.split(':') 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 is_running (self): return len(self.pids) != 0 + + def add_pid (self,pid): + self.pids.append(pid) + def set_broken (self, plcindex, step): + self.broken_steps.append ( (plcindex, step,) ) + + def second_letter (self): + if not self.broken_steps: return '=' + else: + really_broken = [ step for (i,step) in self.broken_steps if '_ignore' not in step ] + # W is for warning like what's in the build mail + if len(really_broken)==0: return 'W' + else: return 'B' + + def line (self): + # make up a 2-letter sign + # first letter : '=', unless build is running : '*' + double = '*' if self.pids else '=' + # second letter : '=' if fine, 'W' for warnings (only ignored steps) 'B' for broken + letter2 = self.second_letter() + double += letter2 + 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 letter2 != '=': + msg2 = ( ' BROKEN' if letter2 == 'B' else ' WARNING' ) + # sometimes we have an empty plcindex + msg += " [%s="%msg2 + " ".join( [ "%s@%s"%(s,i) if i else s for (i,s) in self.broken_steps ] ) + "]" + 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[0-9]+)/cwd.*/root/(?P[^/]+)$") + matcher_grep=re.compile ("/root/(?P[^/]+)/logs/trace.*:TRACE:\s*(?P[0-9]+).*step=(?P\S+).*") + matcher_grep_missing=re.compile ("grep: /root/(?P[^/]+)/logs/trace: No such file or directory") + def sense (self, options): + print 'tm', + 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//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 + + # let's try to be robust here -- tests that fail very early like e.g. + # "Cannot make space for a PLC instance: vplc IP pool exhausted", that occurs as part of provision + # will result in a 'trace' symlink to an inexisting 'trace-<>.txt' because no step has gone through + # simple 'trace' should exist though as it is created by run_log + command=['bash','-c',"grep KO /root/*/logs/trace /dev/null 2>&1" ] + trace_lines=self.backquote_ssh (command).split('\n') + for line in trace_lines: + if not line.strip(): continue + m=TestBox.matcher_grep_missing.match(line) + if m: + buildname=m.group('buildname') + self.add_broken(buildname,'','NO STEP DONE') + 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) + continue + header("TestBox.sense: command %r returned line that failed to match\n%s"%(command,line)) + header(">>%s<<"%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) + continue + header("TestBox.sense: command %r returned line that failed to match\n%s"%(command,line)) + header(">>%s<<"%line) + + + def line (self): + return self.hostname_fedora() + + def list (self, verbose=False): + # verbose shows all tests + if verbose: + instances = self.test_instances + msg="tests" + else: + instances = [ i for i in self.test_instances if i.is_running() ] + msg="running tests" + + if not instances: + header ("No %s on %s"%(msg,self.line())) + else: + header ("%s on %s"%(msg,self.line())) + instances.sort(timestamp_sort) + for i in instances: print i.line() + # show 'starting' regardless of verbose + if self.starting_ips: + header ("Starting IP addresses on %s"%self.line()) + self.starting_ips.sort() + for starting in self.starting_ips: print starting + else: + header ("Empty 'starting' on %s"%self.line()) + ############################################################ class Options: pass @@ -632,14 +1037,26 @@ class Substrate: self.options.verbose=False self.options.reboot=False self.options.soft=False - 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.test_box = TestBox (self.test_box_spec()) + self.build_lxc_boxes = [ BuildLxcBox(h) for h in self.build_lxc_boxes_spec() ] + self.plc_lxc_boxes = [ PlcLxcBox (h,m) for (h,m) in self.plc_lxc_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._sensed=False - self.vplc_pool = Pool (self.vplc_ips(),"for vplcs") - self.vnode_pool = Pool (self.vnode_ips(),"for vnodes") + self.vplc_pool = Pool (self.vplc_ips(),"for vplcs",self) + self.vnode_pool = Pool (self.vnode_ips(),"for vnodes",self) + + self.build_boxes = self.build_lxc_boxes + self.plc_boxes = self.plc_lxc_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 + + def summary_line (self): + msg = "[" + msg += " %d xp"%len(self.plc_lxc_boxes) + msg += " %d tried plc boxes"%len(self.plc_boxes) + msg += "]" + return msg def fqdn (self, hostname): if hostname.find('.')<0: return "%s.%s"%(hostname,self.domain()) @@ -649,19 +1066,28 @@ class Substrate: def sense (self,force=False): if self._sensed and not force: return False print 'Sensing local substrate...', - for b in self.all_boxes: b.sense(self.options) + for b in self.default_boxes: b.sense(self.options) print 'Done' self._sensed=True return True + def list (self, verbose=False): + 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): @@ -672,6 +1098,7 @@ class Substrate: 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' @@ -700,7 +1127,7 @@ class Substrate: plc_boxname = options.ips_bplc.pop() vplc_hostname=options.ips_vplc.pop() else: - if self.sense(): self.list_all() + if self.sense(): self.list() plc_boxname=None vplc_hostname=None # try to find an available IP @@ -712,7 +1139,7 @@ class Substrate: max_free=0 # use the box that has max free spots for load balancing for pb in self.plc_boxes: - free=pb.free_spots() + free=pb.free_slots() if free>max_free: plc_boxname=pb.hostname max_free=free @@ -729,8 +1156,9 @@ class Substrate: 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 + if not vplc_hostname: msg += " vplc IP pool exhausted" + msg += " %s"%self.summary_line() + raise Exception,"Cannot 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(), @@ -748,7 +1176,7 @@ class Substrate: # 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 @@ -795,7 +1223,7 @@ class Substrate: qemu_boxname=options.ips_bnode.pop() vnode_hostname=options.ips_vnode.pop() else: - if self.sense(): self.list_all() + if self.sense(): self.list() qemu_boxname=None vnode_hostname=None # try to find an available IP @@ -807,7 +1235,7 @@ class Substrate: max_free=0 # use the box that has max free spots for load balancing for qb in self.qemu_boxes: - free=qb.free_spots() + free=qb.free_slots() if free>max_free: qemu_boxname=qb.hostname max_free=free @@ -824,7 +1252,8 @@ class Substrate: 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 + msg += " %s"%self.summary_line() + raise Exception,"Cannot 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 @@ -838,15 +1267,16 @@ class Substrate: vnode_hostname=freed_vnode_hostname self.vnode_pool.set_mine(vnode_hostname) - self.add_dummy_qemu (qemu_boxname,nodename) + 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) vnode_fqdn = self.fqdn(vnode_hostname) nodemap={'host_box':qemu_boxname, 'node_fields:hostname':vnode_fqdn, 'interface_fields:ip':ip, + 'ipaddress_fields:ip_addr':ip, 'interface_fields:mac':mac, } nodemap.update(self.network_settings()) @@ -862,11 +1292,8 @@ class Substrate: plc['sfa']['SFA_REGISTRY_HOST'] = plc['PLC_DB_HOST'] plc['sfa']['SFA_AGGREGATE_HOST'] = plc['PLC_DB_HOST'] plc['sfa']['SFA_SM_HOST'] = plc['PLC_DB_HOST'] - plc['sfa']['SFA_PLC_DB_HOST'] = plc['PLC_DB_HOST'] + plc['sfa']['SFA_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: @@ -876,43 +1303,51 @@ class Substrate: pass #################### show results for interactive mode - def get_box (self,box): - for b in self.build_boxes + self.plc_boxes + self.qemu_boxes: - if b.shortname()==box: - return b - print "Could not find box %s"%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 + try: + if b.shortname()==boxname.split('.')[0]: return b + except: pass + print "Could not find box %s"%boxname return None - def list_all (self): - self.sense() - for b in self.all_boxes: b.list() - - def list_boxes(self,boxes): - print 'Partial Sensing', - for box in boxes: - b=self.get_box(box) - if not b: continue - b.sense(self.options) + 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 boxes: - b=self.get_box(box) - if not b: continue - b.list() + for box in box_or_names: + if not isinstance(box,Box): box=self.get_box(box) + if not box: continue + box.list(self.options.verbose) + + 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 + # can be run as a utility to probe/display/manage the local infrastructure def main (self): parser=OptionParser() 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)') + help='soft mode for reboot (terminates processes)') + 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') + 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, @@ -920,12 +1355,17 @@ class Substrate: (self.options,args)=parser.parse_args() boxes=args - if self.options.builds: boxes += [b.hostname for b in self.build_boxes] - if self.options.plcs: boxes += [b.hostname for b in self.plc_boxes] - if self.options.qemus: boxes += [b.hostname for b in self.qemu_boxes] - boxes=list(set(boxes)) + 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 + global verbose + verbose=self.options.verbose + # default scope is -b -p -q -t if not boxes: - self.list_all() - else: - self.list_boxes (boxes) + boxes = self.build_boxes + self.plc_boxes + self.qemu_boxes + [self.test_box] + + if self.options.reboot: self.reboot_boxes (boxes) + else: self.list_boxes (boxes)