grouprins.py

   1 #!/usr/bin/python
   2
   3 # This script is used to manipulate the operational state of nodes in
   4 # different node groups.  These are basically set operations on nodes via the
   5 # PLC api.
   6 #
   7 # Take the ng name as an argument....
   8 # optionally,
   9 #  * get a list of nodes in the given nodegroup.
  10 #  * set some or all in the set to rins.
  11 #  * restart them all.
  12 #  * do something else to them all.
  13 #
  14
  15 import plc
  16 api = plc.getAuthAPI()
  17
  18 import traceback
  19 import config
  20 import util.file
  21 from optparse import OptionParser
  22
  23 import const
  24 from nodecommon import *
  25 from nodequery import verify,query_to_dict,node_select
  26 import database
  27 from unified_model import *
  28 import os
  29
  30 import time
  31 import parser as parsermodule
  32
  33 from model import *
  34 import bootman          # debug nodes
  35 import reboot           # down nodes without pcu
  36 import mailmonitor      # down nodes with pcu
  37 from emailTxt import mailtxt
  38 #reboot.verbose = 0
  39 import sys
  40
  41 class Reboot(object):
  42         def __init__(self, fbnode):
  43                 self.fbnode = fbnode
  44
  45         def _send_pcunotice(self, host):
  46                 args = {}
  47                 args['hostname'] = host
  48                 try:
  49                         args['pcu_id'] = plc.getpcu(host)['pcu_id']
  50                 except:
  51                         args['pcu_id'] = host
  52
  53                 m = PersistMessage(host, mailtxt.pcudown_one[0] % args,
  54                                                                  mailtxt.pcudown_one[1] % args, True, db='pcu_persistmessages')
  55
  56                 loginbase = plc.siteId(host)
  57                 m.send([const.TECHEMAIL % loginbase])
  58
  59         def pcu(self, host):
  60                 # TODO: It should be possible to diagnose the various conditions of
  61                 #               the PCU here, and send different messages as appropriate.
  62                 print "'%s'" % self.fbnode['pcu']
  63                 if self.fbnode['pcu'] == "PCU" or "PCUOK" in self.fbnode['pcu']:
  64                         self.action = "reboot.reboot('%s')" % host
  65
  66                         pflags = PersistFlags(host, 2*60*60*24, db='pcu_persistflags')
  67                         pflags.resetRecentFlag('pcutried')
  68                         if not pflags.getRecentFlag('pcutried'):
  69                                 pflags.setRecentFlag('pcutried')
  70                                 try:
  71                                         print "CALLING REBOOT!!!"
  72                                         ret = reboot.reboot(host)
  73
  74                                         pflags.save()
  75                                         return ret
  76
  77                                 except Exception,e:
  78                                         print traceback.print_exc(); print e
  79
  80                                         # NOTE: this failure could be an implementation issue on
  81                                         #               our end.  So, extra notices are confusing...
  82                                         # self._send_pcunotice(host)
  83
  84                                         pflags.setRecentFlag('pcufailed')
  85                                         pflags.save()
  86                                         return False
  87                         else:
  88                                 # we've tried the pcu recently, but it didn't work,
  89                                 # so did we send a message about it recently?
  90                                 if not pflags.getRecentFlag('pcumessagesent'):
  91
  92                                         self._send_pcunotice(host)
  93
  94                                         pflags.setRecentFlag('pcumessagesent')
  95                                         pflags.save()
  96                                         # NOTE: this will result in just one message sent at a time.
  97                                         return True
  98
  99                                 else:
 100                                         print "GetRecentFlag()"
 101                                         return False
 102                 else:
 103                         print "NO PCUOK"
 104                         self.action = "None"
 105                         return False
 106
 107         def mail(self, host):
 108
 109                 # Reset every 4 weeks or so
 110                 pflags = PersistFlags(host, 27*60*60*24, db='mail_persistflags')
 111                 if not pflags.getRecentFlag('endrecord'):
 112                         node_end_record(host)
 113                         pflags.setRecentFlag('endrecord')
 114                         pflags.save()
 115
 116                 # Then in either case, run mailmonitor.reboot()
 117                 self.action = "mailmonitor.reboot('%s')" % host
 118                 try:
 119                         return mailmonitor.reboot(host)
 120                 except Exception, e:
 121                         print traceback.print_exc(); print e
 122                         return False
 123
 124 class RebootDebug(Reboot):
 125
 126         def direct(self, host):
 127                 self.action = "bootman.reboot('%s', config, None)" % host
 128                 return bootman.reboot(host, config, None)
 129
 130 class RebootBoot(Reboot):
 131
 132         def direct(self, host):
 133                 self.action = "bootman.reboot('%s', config, 'reboot')" % host
 134                 return bootman.reboot(host, config, 'reboot')
 135
 136 class RebootDown(Reboot):
 137
 138         def direct(self, host):
 139                 self.action = "None"
 140                 return False    # this always fails, since the node will be down.
 141
 142 def set_node_to_rins(host, fb):
 143
 144         node = api.GetNodes(host, ['boot_state', 'last_contact', 'last_updated', 'date_created'])
 145         record = {'observation' : node[0],
 146                           'model' : 'USER_REQUEST',
 147                           'action' : 'api.UpdateNode(%s, {"boot_state" : "rins"})' % host,
 148                           'time' : time.time()}
 149         l = Log(host, record)
 150
 151         ret = api.UpdateNode(host, {'boot_state' : 'rins'})
 152         if ret:
 153                 # it's nice to see the current status rather than the previous status on the console
 154                 node = api.GetNodes(host)[0]
 155                 print l
 156                 print "%-2d" % (i-1), nodegroup_display(node, fb)
 157                 return l
 158         else:
 159                 print "FAILED TO UPDATE NODE BOOT STATE : %s" % host
 160                 return None
 161
 162
 163 try:
 164         rebootlog = database.dbLoad("rebootlog")
 165 except:
 166         rebootlog = LogRoll()
 167
 168 parser = parsermodule.getParser(['nodesets'])
 169 parser.set_defaults( timewait=0,
 170                                         skip=0,
 171                                         rins=False,
 172                                         reboot=False,
 173                                         findbad=False,
 174                                         force=False,
 175                                         nosetup=False,
 176                                         verbose=False,
 177                                         stopkey=None,
 178                                         stopvalue=None,
 179                                         quiet=False,
 180                                         )
 181
 182 parser.add_option("", "--stopselect", dest="stopselect", metavar="",
 183                                         help="The select string that must evaluate to true for the node to be considered 'done'")
 184 parser.add_option("", "--findbad", dest="findbad", action="store_true",
 185                                         help="Re-run findbad on the nodes we're going to check before acting.")
 186 parser.add_option("", "--force", dest="force", action="store_true",
 187                                         help="Force action regardless of previous actions/logs.")
 188 parser.add_option("", "--rins", dest="rins", action="store_true",
 189                                         help="Set the boot_state to 'rins' for all nodes.")
 190 parser.add_option("", "--reboot", dest="reboot", action="store_true",
 191                                         help="Actively try to reboot the nodes, keeping a log of actions.")
 192
 193 parser.add_option("", "--verbose", dest="verbose", action="store_true",
 194                                         help="Extra debug output messages.")
 195 parser.add_option("", "--nosetup", dest="nosetup", action="store_true",
 196                                         help="Do not perform the orginary setup phase.")
 197 parser.add_option("", "--skip", dest="skip",
 198                                         help="Number of machines to skip on the input queue.")
 199 parser.add_option("", "--timewait", dest="timewait",
 200                                         help="Minutes to wait between iterations of 10 nodes.")
 201
 202 parser = parsermodule.getParser(['defaults'], parser)
 203 config = parsermodule.parse_args(parser)
 204
 205 # COLLECT nodegroups, nodes and node lists
 206 if config.nodegroup:
 207         ng = api.GetNodeGroups({'name' : config.nodegroup})
 208         nodelist = api.GetNodes(ng[0]['node_ids'])
 209         hostnames = [ n['hostname'] for n in nodelist ]
 210
 211 if config.node or config.nodelist:
 212         if config.node: hostnames = [ config.node ]
 213         else: hostnames = config.getListFromFile(config.nodelist)
 214
 215 fb = database.dbLoad("findbad")
 216
 217 if config.nodeselect:
 218         hostnames = node_select(config.nodeselect, fb['nodes'].keys(), fb)
 219
 220 if config.findbad:
 221         # rerun findbad with the nodes in the given nodes.
 222         file = "findbad.txt"
 223         util.file.setFileFromList(file, hostnames)
 224         os.system("./findbad.py --cachenodes --debug=0 --dbname=findbad --increment --nodelist %s" % file)
 225
 226 # commands:
 227 i = 1
 228 count = 1
 229 for host in hostnames:
 230
 231         #if 'echo' in host or 'hptest-1' in host: continue
 232         try:
 233                 try:
 234                         node = api.GetNodes(host)[0]
 235                 except:
 236                         print traceback.print_exc();
 237                         print "FAILED GETNODES for host: %s" % host
 238                         continue
 239
 240                 print "%-2d" % i, nodegroup_display(node, fb)
 241                 i += 1
 242                 if i-1 <= int(config.skip): continue
 243
 244                 if config.stopselect:
 245                         dict_query = query_to_dict(config.stopselect)
 246                         fbnode = fb['nodes'][host]['values']
 247                         observed_state = get_current_state(fbnode)
 248
 249                         if verify(dict_query, fbnode) and observed_state != "dbg ":
 250                                 # evaluates to true, therefore skip.
 251                                 print "%s evaluates true for %s ; skipping..." % ( config.stopselect, host )
 252                                 continue
 253
 254                 if config.stopkey and config.stopvalue:
 255                         fbnode = fb['nodes'][host]['values']
 256                         observed_state = get_current_state(fbnode)
 257
 258                         if config.stopkey in fbnode:
 259                                 if config.stopvalue in fbnode[config.stopkey] and observed_state != "dbg ":
 260                                         print "%s has stopvalue; skipping..." % host
 261                                         continue
 262                         else:
 263                                 print "stopkey %s not in fbnode record for %s; skipping..." % (config.stopkey, host)
 264                                 print fbnode
 265                                 continue
 266
 267                 if not config.force and rebootlog.find(host, {'action' : ".*reboot"}, 60*60*2):
 268                         print "recently rebooted %s.  skipping... " % host
 269                         continue
 270
 271                 if config.reboot:
 272
 273                         fbnode = fb['nodes'][host]['values']
 274                         observed_state = get_current_state(fbnode)
 275
 276                         if       observed_state == "dbg ":
 277                                 o = RebootDebug(fbnode)
 278
 279                         elif observed_state == "boot" :
 280                                 if config.rins:
 281                                         l = set_node_to_rins(host, fb)
 282                                         if l: rebootlog.add(l)
 283
 284                                 o = RebootBoot(fbnode)
 285
 286                         elif observed_state == "down":
 287                                 if config.rins:
 288                                         l = set_node_to_rins(host, fb)
 289                                         if l: rebootlog.add(l)
 290
 291                                 o = RebootDown(fbnode)
 292
 293
 294                         if o.direct(host):
 295                                 record = {'observation' : "DIRECT_SUCCESS: %s" % observed_state,
 296                                                   'action' : o.action,
 297                                                   'model' : "none",
 298                                                   'time' : time.time()}
 299                         elif o.pcu(host):
 300                                 record = {'observation' : "PCU_SUCCESS: %s" % observed_state,
 301                                                   'action' : o.action,
 302                                                   'model' : "none",
 303                                                   'time' : time.time()}
 304                         elif o.mail(host):
 305                                 record = {'observation' : "MAIL_SUCCESS: %s" % observed_state,
 306                                                   'action' : o.action,
 307                                                   'model' : "none",
 308                                                   'time' : time.time()}
 309                         else:
 310                                 record = {'observation' : "REBOOT_FAILED: %s" %  observed_state,
 311                                                   'action' : "log failure",
 312                                                   'model' : "none",
 313                                                   'time' : time.time()}
 314
 315                                 print "ALL METHODS OF RESTARTING %s FAILED" % host
 316                                 args = {}
 317                                 args['hostname'] = host
 318                                 m = PersistMessage(host, "ALL METHODS FAILED for %(hostname)s" % args,
 319                                                                                          "CANNOT CONTACT", False, db='suspect_persistmessages')
 320                                 m.reset()
 321                                 m.send(['monitor-list@lists.planet-lab.org'])
 322
 323                         l = Log(host, record)
 324                         print l
 325                         rebootlog.add(l)
 326         except KeyboardInterrupt:
 327                 print "Killed by interrupt"
 328                 sys.exit(0)
 329         except:
 330                 print traceback.print_exc();
 331                 print "Continuing..."
 332
 333         time.sleep(1)
 334         if count % 10 == 0:
 335                 print "Saving rebootlog"
 336                 database.dbDump("rebootlog", rebootlog)
 337                 wait_time = int(config.timewait)
 338                 print "Sleeping %d minutes" % wait_time
 339                 ti = 0
 340                 print "Minutes slept: ",
 341                 sys.stdout.flush()
 342                 while ti < wait_time:
 343                         print "%s" % ti,
 344                         sys.stdout.flush()
 345                         time.sleep(60)
 346                         ti = ti+1
 347
 348         count = count + 1
 349
 350 print "Saving rebootlog"
 351 database.dbDump("rebootlog", rebootlog)