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 import auth
  17 api = plc.PLC(auth.auth, auth.plc)
  18
  19 import policy
  20 import traceback
  21 from config import config as cfg
  22 import config as configmodule
  23 from optparse import OptionParser
  24
  25 from nodecommon import *
  26 from nodequery import verify,query_to_dict,node_select
  27 import soltesz
  28 from unified_model import *
  29 import os
  30
  31 import time
  32
  33 from model import *
  34 import bootman          # debug nodes
  35 import monitor          # down nodes with pcu
  36 import reboot           # down nodes without 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([policy.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                 if self.fbnode['pcu'] == "PCU":
  63                         self.action = "reboot.reboot('%s')" % host
  64
  65                         pflags = PersistFlags(host, 2*60*60*24, db='pcu_persistflags')
  66                         if not pflags.getRecentFlag('pcutried'):
  67                                 pflags.setRecentFlag('pcutried')
  68                                 try:
  69                                         ret = reboot.reboot(host)
  70
  71                                         pflags.save()
  72                                         return ret
  73
  74                                 except Exception,e:
  75                                         print traceback.print_exc(); print e
  76
  77                                         # NOTE: this failure could be an implementation issue on
  78                                         #               our end.  So, extra notices are confusing...
  79                                         # self._send_pcunotice(host)
  80
  81                                         pflags.setRecentFlag('pcufailed')
  82                                         pflags.save()
  83                                         return False
  84                         else:
  85                                 # we've tried the pcu recently, but it didn't work,
  86                                 # so did we send a message about it recently?
  87                                 if not pflags.getRecentFlag('pcumessagesent'):
  88
  89                                         self._send_pcunotice(host)
  90
  91                                         pflags.setRecentFlag('pcumessagesent')
  92                                         pflags.save()
  93                                         # NOTE: this will result in just one message sent at a time.
  94                                         return True
  95
  96                                 else:
  97                                         return False
  98                 else:
  99                         self.action = "None"
 100                         return False
 101
 102         def mail(self, host):
 103
 104                 # Reset every 4 weeks or so
 105                 pflags = PersistFlags(host, 27*60*60*24, db='mail_persistflags')
 106                 if not pflags.getRecentFlag('endrecord'):
 107                         node_end_record(host)
 108                         pflags.setRecentFlag('endrecord')
 109                         pflags.save()
 110
 111                 # Then in either case, run monitor.reboot()
 112                 self.action = "monitor.reboot('%s')" % host
 113                 try:
 114                         return monitor.reboot(host)
 115                 except Exception, e:
 116                         print traceback.print_exc(); print e
 117                         return False
 118
 119 class RebootDebug(Reboot):
 120
 121         def direct(self, host):
 122                 self.action = "bootman.reboot('%s', config, None)" % host
 123                 return bootman.reboot(host, config, None)
 124
 125 class RebootBoot(Reboot):
 126
 127         def direct(self, host):
 128                 self.action = "bootman.reboot('%s', config, 'reboot')" % host
 129                 return bootman.reboot(host, config, 'reboot')
 130
 131 class RebootDown(Reboot):
 132
 133         def direct(self, host):
 134                 self.action = "None"
 135                 return False    # this always fails, since the node will be down.
 136
 137 def set_node_to_rins(host, fb):
 138
 139         node = api.GetNodes(host, ['boot_state', 'last_contact', 'last_updated', 'date_created'])
 140         record = {'observation' : node[0],
 141                           'model' : 'USER_REQUEST',
 142                           'action' : 'api.UpdateNode(%s, {"boot_state" : "rins"})' % host,
 143                           'time' : time.time()}
 144         l = Log(host, record)
 145
 146         ret = api.UpdateNode(host, {'boot_state' : 'rins'})
 147         if ret:
 148                 # it's nice to see the current status rather than the previous status on the console
 149                 node = api.GetNodes(host)[0]
 150                 print l
 151                 print "%-2d" % (i-1), nodegroup_display(node, fb)
 152                 return l
 153         else:
 154                 print "FAILED TO UPDATE NODE BOOT STATE : %s" % host
 155                 return None
 156
 157
 158 try:
 159         rebootlog = soltesz.dbLoad("rebootlog")
 160 except:
 161         rebootlog = LogRoll()
 162
 163 parser = OptionParser()
 164 parser.set_defaults(nodegroup=None,
 165                                         node=None,
 166                                         nodelist=None,
 167                                         nodeselect=None,
 168                                         timewait=0,
 169                                         skip=0,
 170                                         rins=False,
 171                                         reboot=False,
 172                                         findbad=False,
 173                                         force=False,
 174                                         nosetup=False,
 175                                         verbose=False,
 176                                         stopkey=None,
 177                                         stopvalue=None,
 178                                         quiet=False,
 179                                         )
 180 parser.add_option("", "--node", dest="node", metavar="nodename.edu",
 181                                         help="A single node name to add to the nodegroup")
 182 parser.add_option("", "--nodelist", dest="nodelist", metavar="list.txt",
 183                                         help="Use all nodes in the given file for operation.")
 184 parser.add_option("", "--nodegroup", dest="nodegroup", metavar="NodegroupName",
 185                                         help="Specify a nodegroup to perform actions on")
 186 parser.add_option("", "--nodeselect", dest="nodeselect", metavar="querystring",
 187                                         help="Specify a query to perform on findbad db")
 188
 189 parser.add_option("", "--verbose", dest="verbose", action="store_true",
 190                                         help="Extra debug output messages.")
 191 parser.add_option("", "--nosetup", dest="nosetup", action="store_true",
 192                                         help="Do not perform the orginary setup phase.")
 193
 194 parser.add_option("", "--skip", dest="skip",
 195                                         help="Number of machines to skip on the input queue.")
 196 parser.add_option("", "--timewait", dest="timewait",
 197                                         help="Minutes to wait between iterations of 10 nodes.")
 198
 199 parser.add_option("", "--stopselect", dest="stopselect", metavar="",
 200                                         help="The select string that must evaluate to true for the node to be considered 'done'")
 201
 202 parser.add_option("", "--stopkey", dest="stopkey", metavar="",
 203                                         help="")
 204 parser.add_option("", "--stopvalue", dest="stopvalue", metavar="",
 205                                         help="")
 206
 207 parser.add_option("", "--findbad", dest="findbad", action="store_true",
 208                                         help="Re-run findbad on the nodes we're going to check before acting.")
 209 parser.add_option("", "--force", dest="force", action="store_true",
 210                                         help="Force action regardless of previous actions/logs.")
 211 parser.add_option("", "--rins", dest="rins", action="store_true",
 212                                         help="Set the boot_state to 'rins' for all nodes.")
 213 parser.add_option("", "--reboot", dest="reboot", action="store_true",
 214                                         help="Actively try to reboot the nodes, keeping a log of actions.")
 215 #config = config(parser)
 216 config = cfg(parser)
 217 config.parse_args()
 218
 219 # COLLECT nodegroups, nodes and node lists
 220 if config.nodegroup:
 221         ng = api.GetNodeGroups({'name' : config.nodegroup})
 222         nodelist = api.GetNodes(ng[0]['node_ids'])
 223         hostnames = [ n['hostname'] for n in nodelist ]
 224
 225 if config.node or config.nodelist:
 226         if config.node: hostnames = [ config.node ]
 227         else: hostnames = config.getListFromFile(config.nodelist)
 228
 229 if config.nodeselect:
 230         hostnames = node_select(config.nodeselect)
 231
 232 if config.findbad:
 233         # rerun findbad with the nodes in the given nodes.
 234         file = "findbad.txt"
 235         configmodule.setFileFromList(file, hostnames)
 236         os.system("./findbad.py --cachenodes --debug=0 --dbname=findbad --increment --nodelist %s" % file)
 237
 238 fb = soltesz.dbLoad("findbad")
 239 # commands:
 240 i = 1
 241 count = 1
 242 for host in hostnames:
 243
 244         #if 'echo' in host or 'hptest-1' in host: continue
 245         try:
 246                 try:
 247                         node = api.GetNodes(host)[0]
 248                 except:
 249                         print traceback.print_exc();
 250                         print "FAILED GETNODES for host: %s" % host
 251                         continue
 252
 253                 print "%-2d" % i, nodegroup_display(node, fb)
 254                 i += 1
 255                 if i < int(config.skip): continue
 256
 257                 if config.stopselect:
 258                         dict_query = query_to_dict(config.stopselect)
 259                         fbnode = fb['nodes'][host]['values']
 260                         observed_state = get_current_state(fbnode)
 261
 262                         if verify(dict_query, fbnode) and observed_state != "dbg ":
 263                                 # evaluates to true, therefore skip.
 264                                 print "%s evaluates true for %s ; skipping..." % ( config.stopselect, host )
 265                                 continue
 266
 267                 if config.stopkey and config.stopvalue:
 268                         fbnode = fb['nodes'][host]['values']
 269                         observed_state = get_current_state(fbnode)
 270
 271                         if config.stopkey in fbnode:
 272                                 if config.stopvalue in fbnode[config.stopkey] and observed_state != "dbg ":
 273                                         print "%s has stopvalue; skipping..." % host
 274                                         continue
 275                         else:
 276                                 print "stopkey %s not in fbnode record for %s; skipping..." % (config.stopkey, host)
 277                                 print fbnode
 278                                 continue
 279
 280                 if not config.force and rebootlog.find(host, {'action' : ".*reboot"}, 60*60*2):
 281                         print "recently rebooted %s.  skipping... " % host
 282                         continue
 283
 284                 if config.reboot:
 285
 286                         fbnode = fb['nodes'][host]['values']
 287                         observed_state = get_current_state(fbnode)
 288
 289                         if       observed_state == "dbg ":
 290                                 o = RebootDebug(fbnode)
 291
 292                         elif observed_state == "boot" :
 293                                 if config.rins:
 294                                         l = set_node_to_rins(host, fb)
 295                                         if l: rebootlog.add(l)
 296
 297                                 o = RebootBoot(fbnode)
 298
 299                         elif observed_state == "down":
 300                                 if config.rins:
 301                                         l = set_node_to_rins(host, fb)
 302                                         if l: rebootlog.add(l)
 303
 304                                 o = RebootDown(fbnode)
 305
 306
 307                         if o.direct(host):
 308                                 record = {'observation' : "DIRECT_SUCCESS: %s" % observed_state,
 309                                                   'action' : o.action,
 310                                                   'model' : "none",
 311                                                   'time' : time.time()}
 312                         elif o.pcu(host):
 313                                 record = {'observation' : "PCU_SUCCESS: %s" % observed_state,
 314                                                   'action' : o.action,
 315                                                   'model' : "none",
 316                                                   'time' : time.time()}
 317                         elif o.mail(host):
 318                                 record = {'observation' : "MAIL_SUCCESS: %s" % observed_state,
 319                                                   'action' : o.action,
 320                                                   'model' : "none",
 321                                                   'time' : time.time()}
 322                         else:
 323                                 record = {'observation' : "REBOOT_FAILED: %s" %  observed_state,
 324                                                   'action' : "log failure",
 325                                                   'model' : "none",
 326                                                   'time' : time.time()}
 327
 328                                 print "ALL METHODS OF RESTARTING %s FAILED" % host
 329                                 args = {}
 330                                 args['hostname'] = host
 331                                 m = PersistMessage(host, "ALL METHODS FAILED for %(hostname)s" % args,
 332                                                                                          "CANNOT CONTACT", False, db='suspect_persistmessages')
 333                                 m.reset()
 334                                 m.send(['monitor-list@lists.planet-lab.org'])
 335
 336                         l = Log(host, record)
 337                         print l
 338                         rebootlog.add(l)
 339         except KeyboardInterrupt:
 340                 print "Killed by interrupt"
 341                 sys.exit(0)
 342         except:
 343                 print traceback.print_exc();
 344                 print "Continuing..."
 345
 346         time.sleep(1)
 347         if count % 10 == 0:
 348                 print "Saving rebootlog"
 349                 soltesz.dbDump("rebootlog", rebootlog)
 350                 wait_time = int(config.timewait)
 351                 print "Sleeping %d minutes" % wait_time
 352                 ti = 0
 353                 print "Minutes slept: ",
 354                 sys.stdout.flush()
 355                 while ti < wait_time:
 356                         print "%s" % ti,
 357                         sys.stdout.flush()
 358                         time.sleep(60)
 359                         ti = ti+1
 360
 361         count = count + 1
 362
 363 print "Saving rebootlog"
 364 soltesz.dbDump("rebootlog", rebootlog)