X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=system%2FSubstrate.py;h=2706286f171e4cdda6f8faca3f8f7b066106c146;hb=4e0d3d28756fd6e622e66526d6d15770373faba5;hp=9656251451c9c9b9b8839b3ba402291c9d369505;hpb=8b60a0d9dc6dd5a7e48c1fc41b9bf4349ea96f9c;p=tests.git diff --git a/system/Substrate.py b/system/Substrate.py index 9656251..2706286 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 "===============", @@ -261,7 +264,16 @@ class Box: self.test_ssh().run("shutdown -r now",message="Rebooting %s"%self.hostname, dry_run=options.dry_run) - def hostname_fedora (self): return "%s [%s]"%(self.hostname,self.fedora()) + 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===" @@ -286,21 +298,24 @@ class Box: else: probe_argv=self.test_ssh().actual_argv(composite_command) composite=self.backquote ( probe_argv, trash_err=True ) - if not composite: print "root@%s unreachable"%self.hostname self._hostname = self._uptime = self._uname = self._fedora = "** Unknown **" - 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 + 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"] @@ -387,34 +402,67 @@ class BuildBox (Box): 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)") - matcher_building_vm=re.compile("\s*(?P[0-9]+).*init-vserver.*\s+(?P[^\s]+)\s*\Z") def sense(self, options): - print 'bb', + 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')) continue - m=BuildBox.matcher_building_vm.match(line) + m=build_matcher_initvm.match(line) if m: # buildname is expansed here self.add_build (buildname,m.group('pid')) continue - header('BuildBox.sense: command %r returned line that failed to match'%command) + 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, plcbox): @@ -513,7 +561,7 @@ class PlcBox (Box): for p in self.plc_instances: header (p.line(),banner=False) - +# we do not this at INRIA any more class PlcVsBox (PlcBox): def add_vserver (self,vservername,ctxid): @@ -525,7 +573,7 @@ class PlcVsBox (PlcBox): self.plc_instances.append(PlcVsInstance(self,vservername,ctxid)) def line(self): - msg="%s [max=%d,free=%d, VS-based] (%s)"%(self.hostname_fedora(), self.max_plcs,self.free_slots(),self.uname()) + 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): @@ -600,8 +648,10 @@ class PlcLxcBox (PlcBox): # a line describing the box def line(self): - return "%s [max=%d,free=%d, LXC-based] (%s)"%(self.hostname_fedora(), self.max_plcs,self.free_slots(), - self.uname()) + 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: @@ -618,7 +668,6 @@ class PlcLxcBox (PlcBox): # sense is expected to fill self.plc_instances with PlcLxcInstance's # to describe the currently running VM's - # as well as to call self.get_uname() once def sense (self, options): print "xp", command="rsync lxc-driver.sh %s:/root"%self.hostname @@ -631,7 +680,8 @@ class PlcLxcBox (PlcBox): pid=lxc_line.split(";")[1] timestamp=lxc_line.split(";")[2] self.add_lxc(lxcname,pid) - timestamp=int(timestamp) + try: timestamp=int(timestamp) + except: timestamp=0 p=self.plc_instance_by_lxcname(lxcname) if not p: print 'WARNING zombie plc',self.hostname,lxcname @@ -696,14 +746,15 @@ class QemuBox (Box): self.qemu_instances.append(dummy) def line (self): - msg="%s [max=%d,free=%d] (%s)"%(self.hostname_fedora(), self.max_qemus,self.free_slots(),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, verbose=False): if not self.qemu_instances: header ('No qemu on %s'%(self.line())) else: - header ("Active qemus 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) @@ -818,18 +869,30 @@ class TestInstance: 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): - double='==' - if self.pids: double='*'+double[1] - if self.broken_steps: double=double[0]+'B' + # 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 self.broken_steps: + if letter2 != '=': + msg += ( 'BROKEN' if letter2 == 'B' else 'WARNING' ) # sometimes we have an empty plcindex - msg += " [BROKEN=" + " ".join( [ "%s@%s"%(s,i) if i else s for (i,s) in self.broken_steps ] ) + "]" + msg += " [%s="%msg + " ".join( [ "%s@%s"%(s,i) if i else s for (i,s) in self.broken_steps ] ) + "]" return msg class TestBox (Box): @@ -901,7 +964,7 @@ class TestBox (Box): # 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' sohuld exist though as it is created by run_log + # 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: @@ -968,40 +1031,28 @@ class Options: pass class Substrate: - def __init__ (self, plcs_on_vs=True, plcs_on_lxc=False): + def __init__ (self): self.options=Options() self.options.dry_run=False self.options.verbose=False 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() ] - # for compat with older LocalSubstrate - try: - self.plc_vs_boxes = [ PlcVsBox (h,m) for (h,m) in self.plc_vs_boxes_spec ()] - self.plc_lxc_boxes = [ PlcLxcBox (h,m) for (h,m) in self.plc_lxc_boxes_spec ()] - except: - self.plc_vs_boxes = [ PlcVsBox (h,m) for (h,m) in self.plc_boxes_spec ()] - self.plc_lxc_boxes = [ ] + 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._sensed=False self.vplc_pool = Pool (self.vplc_ips(),"for vplcs",self) self.vnode_pool = Pool (self.vnode_ips(),"for vnodes",self) - self.rescope (plcs_on_vs=plcs_on_vs, plcs_on_lxc=plcs_on_lxc) - - # which plc boxes are we interested in ? - def rescope (self, plcs_on_vs, plcs_on_lxc): - self.plc_boxes=[] - if plcs_on_vs: self.plc_boxes += self.plc_vs_boxes - if plcs_on_lxc: self.plc_boxes += self.plc_lxc_boxes + 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 vp"%len(self.plc_vs_boxes) msg += " %d xp"%len(self.plc_lxc_boxes) msg += " %d tried plc boxes"%len(self.plc_boxes) msg += "]" @@ -1286,7 +1337,7 @@ class Substrate: 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, @@ -1303,8 +1354,6 @@ class Substrate: help='dry run mode') (self.options,args)=parser.parse_args() - self.rescope (plcs_on_vs=True, plcs_on_lxc=True) - boxes=args if self.options.testbox: boxes += [self.test_box] if self.options.builds: boxes += self.build_boxes @@ -1312,6 +1361,8 @@ class Substrate: 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: boxes = self.build_boxes + self.plc_boxes + self.qemu_boxes + [self.test_box]