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 policy
  19 import traceback
  20 from config import config as cfg
  21 import util.file
  22 from optparse import OptionParser
  23
  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 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 = database.dbLoad("rebootlog")
 160 except:
 161         rebootlog = LogRoll()
 162
 163 parser = parsermodule.getParser(['nodesets'])
 164 parser.set_defaults( timewait=0,
 165                                         skip=0,
 166                                         rins=False,
 167                                         reboot=False,
 168                                         findbad=False,
 169                                         force=False,
 170                                         nosetup=False,
 171                                         verbose=False,
 172                                         stopkey=None,
 173                                         stopvalue=None,
 174                                         quiet=False,
 175                                         )
 176
 177 parser.add_option("", "--stopselect", dest="stopselect", metavar="",
 178                                         help="The select string that must evaluate to true for the node to be considered 'done'")
 179 parser.add_option("", "--findbad", dest="findbad", action="store_true",
 180                                         help="Re-run findbad on the nodes we're going to check before acting.")
 181 parser.add_option("", "--force", dest="force", action="store_true",
 182                                         help="Force action regardless of previous actions/logs.")
 183 parser.add_option("", "--rins", dest="rins", action="store_true",
 184                                         help="Set the boot_state to 'rins' for all nodes.")
 185 parser.add_option("", "--reboot", dest="reboot", action="store_true",
 186                                         help="Actively try to reboot the nodes, keeping a log of actions.")
 187
 188 parser.add_option("", "--verbose", dest="verbose", action="store_true",
 189                                         help="Extra debug output messages.")
 190 parser.add_option("", "--nosetup", dest="nosetup", action="store_true",
 191                                         help="Do not perform the orginary setup phase.")
 192 parser.add_option("", "--skip", dest="skip",
 193                                         help="Number of machines to skip on the input queue.")
 194 parser.add_option("", "--timewait", dest="timewait",
 195                                         help="Minutes to wait between iterations of 10 nodes.")
 196
 197 parser = parsermodule.getParser(['defaults'], parser)
 198 config = parsermodule.parse_args(parser)
 199
 200 # COLLECT nodegroups, nodes and node lists
 201 if config.nodegroup:
 202         ng = api.GetNodeGroups({'name' : config.nodegroup})
 203         nodelist = api.GetNodes(ng[0]['node_ids'])
 204         hostnames = [ n['hostname'] for n in nodelist ]
 205
 206 if config.node or config.nodelist:
 207         if config.node: hostnames = [ config.node ]
 208         else: hostnames = config.getListFromFile(config.nodelist)
 209
 210 if config.nodeselect:
 211         hostnames = node_select(config.nodeselect)
 212
 213 if config.findbad:
 214         # rerun findbad with the nodes in the given nodes.
 215         file = "findbad.txt"
 216         util.file.setFileFromList(file, hostnames)
 217         os.system("./findbad.py --cachenodes --debug=0 --dbname=findbad --increment --nodelist %s" % file)
 218
 219 fb = database.dbLoad("findbad")
 220 # commands:
 221 i = 1
 222 count = 1
 223 for host in hostnames:
 224
 225         #if 'echo' in host or 'hptest-1' in host: continue
 226         try:
 227                 try:
 228                         node = api.GetNodes(host)[0]
 229                 except:
 230                         print traceback.print_exc();
 231                         print "FAILED GETNODES for host: %s" % host
 232                         continue
 233
 234                 print "%-2d" % i, nodegroup_display(node, fb)
 235                 i += 1
 236                 if i < int(config.skip): continue
 237
 238                 if config.stopselect:
 239                         dict_query = query_to_dict(config.stopselect)
 240                         fbnode = fb['nodes'][host]['values']
 241                         observed_state = get_current_state(fbnode)
 242
 243                         if verify(dict_query, fbnode) and observed_state != "dbg ":
 244                                 # evaluates to true, therefore skip.
 245                                 print "%s evaluates true for %s ; skipping..." % ( config.stopselect, host )
 246                                 continue
 247
 248                 if config.stopkey and config.stopvalue:
 249                         fbnode = fb['nodes'][host]['values']
 250                         observed_state = get_current_state(fbnode)
 251
 252                         if config.stopkey in fbnode:
 253                                 if config.stopvalue in fbnode[config.stopkey] and observed_state != "dbg ":
 254                                         print "%s has stopvalue; skipping..." % host
 255                                         continue
 256                         else:
 257                                 print "stopkey %s not in fbnode record for %s; skipping..." % (config.stopkey, host)
 258                                 print fbnode
 259                                 continue
 260
 261                 if not config.force and rebootlog.find(host, {'action' : ".*reboot"}, 60*60*2):
 262                         print "recently rebooted %s.  skipping... " % host
 263                         continue
 264
 265                 if config.reboot:
 266
 267                         fbnode = fb['nodes'][host]['values']
 268                         observed_state = get_current_state(fbnode)
 269
 270                         if       observed_state == "dbg ":
 271                                 o = RebootDebug(fbnode)
 272
 273                         elif observed_state == "boot" :
 274                                 if config.rins:
 275                                         l = set_node_to_rins(host, fb)
 276                                         if l: rebootlog.add(l)
 277
 278                                 o = RebootBoot(fbnode)
 279
 280                         elif observed_state == "down":
 281                                 if config.rins:
 282                                         l = set_node_to_rins(host, fb)
 283                                         if l: rebootlog.add(l)
 284
 285                                 o = RebootDown(fbnode)
 286
 287
 288                         if o.direct(host):
 289                                 record = {'observation' : "DIRECT_SUCCESS: %s" % observed_state,
 290                                                   'action' : o.action,
 291                                                   'model' : "none",
 292                                                   'time' : time.time()}
 293                         elif o.pcu(host):
 294                                 record = {'observation' : "PCU_SUCCESS: %s" % observed_state,
 295                                                   'action' : o.action,
 296                                                   'model' : "none",
 297                                                   'time' : time.time()}
 298                         elif o.mail(host):
 299                                 record = {'observation' : "MAIL_SUCCESS: %s" % observed_state,
 300                                                   'action' : o.action,
 301                                                   'model' : "none",
 302                                                   'time' : time.time()}
 303                         else:
 304                                 record = {'observation' : "REBOOT_FAILED: %s" %  observed_state,
 305                                                   'action' : "log failure",
 306                                                   'model' : "none",
 307                                                   'time' : time.time()}
 308
 309                                 print "ALL METHODS OF RESTARTING %s FAILED" % host
 310                                 args = {}
 311                                 args['hostname'] = host
 312                                 m = PersistMessage(host, "ALL METHODS FAILED for %(hostname)s" % args,
 313                                                                                          "CANNOT CONTACT", False, db='suspect_persistmessages')
 314                                 m.reset()
 315                                 m.send(['monitor-list@lists.planet-lab.org'])
 316
 317                         l = Log(host, record)
 318                         print l
 319                         rebootlog.add(l)
 320         except KeyboardInterrupt:
 321                 print "Killed by interrupt"
 322                 sys.exit(0)
 323         except:
 324                 print traceback.print_exc();
 325                 print "Continuing..."
 326
 327         time.sleep(1)
 328         if count % 10 == 0:
 329                 print "Saving rebootlog"
 330                 database.dbDump("rebootlog", rebootlog)
 331                 wait_time = int(config.timewait)
 332                 print "Sleeping %d minutes" % wait_time
 333                 ti = 0
 334                 print "Minutes slept: ",
 335                 sys.stdout.flush()
 336                 while ti < wait_time:
 337                         print "%s" % ti,
 338                         sys.stdout.flush()
 339                         time.sleep(60)
 340                         ti = ti+1
 341
 342         count = count + 1
 343
 344 print "Saving rebootlog"
 345 database.dbDump("rebootlog", rebootlog)