policy.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 os
  16 import time
  17 import traceback
  18 import sys
  19 from optparse import OptionParser
  20
  21 from monitor import config
  22 from monitor import parser as parsermodule
  23 from monitor.common import *
  24 from monitor.model import *
  25 from monitor.wrapper import plc
  26 from monitor.wrapper import plccache
  27 from monitor.database.info.model import *
  28 from monitor.database.info.interface import *
  29
  30 from nodequery import verify,query_to_dict,node_select
  31
  32 api = plc.getAuthAPI()
  33
  34 def logic():
  35
  36         plc.nodeBootState(host, 'rins')
  37         node_end_record(host)
  38
  39 def main(hostnames, sitenames):
  40         # commands:
  41         i = 1
  42         node_count = 1
  43         site_count = 1
  44         #print "hosts: %s" % hostnames
  45         for i,host in enumerate(hostnames):
  46                 try:
  47                         lb = plccache.plcdb_hn2lb[host]
  48                 except:
  49                         print "unknown host in plcdb_hn2lb %s" % host
  50                         email_exception(host)
  51                         continue
  52
  53                 nodeblack = BlacklistRecord.get_by(hostname=host)
  54
  55                 if nodeblack and not nodeblack.expired():
  56                         print "skipping %s due to blacklist.  will expire %s" % (host, nodeblack.willExpire() )
  57                         continue
  58
  59                 sitehist = SiteInterface.get_or_make(loginbase=lb)
  60
  61                 recent_actions = sitehist.getRecentActions(hostname=host)
  62
  63                 nodehist = HistoryNodeRecord.findby_or_create(hostname=host)
  64
  65                 print "%s %s %s" % (i, nodehist.hostname, nodehist.status)
  66                 if nodehist.status == 'good' and \
  67                         changed_lessthan(nodehist.last_changed, 1.0) and \
  68                         found_within(recent_actions, 'down_notice', 7.0) and \
  69                         not found_within(recent_actions, 'online_notice', 0.5):
  70                                 # NOTE: chronicly flapping nodes will not get 'online' notices
  71                                 #               since, they are never up long enough to be 'good'.
  72                             # NOTE: searching for down_notice proves that the node has
  73                                 #               gone through a 'down' state first, rather than just
  74                                 #               flapping through: good, offline, online, ...
  75                                 #
  76                                 # NOTE: there is a narrow window in which this command must be
  77                                 #               evaluated, otherwise the notice will not go out.
  78                                 #               this is not ideal.
  79                                 sitehist.sendMessage('online_notice', hostname=host, viart=False)
  80                                 print "send message for host %s online" % host
  81
  82
  83                 # if a node is offline and doesn't have a PCU, remind the user that they should have one.
  84                 if not nodehist.haspcu and nodehist.status in ['offline', 'down'] and \
  85                         changed_greaterthan(nodehist.last_changed,1.0) and \
  86                         not found_within(recent_actions, 'pcumissing_notice', 7.0):
  87
  88                                 sitehist.sendMessage('pcumissing_notice', hostname=host)
  89                                 print "send message for host %s pcumissing_notice" % host
  90
  91                 # if it is offline and HAS a PCU, then try to use it.
  92                 if nodehist.haspcu and nodehist.status in ['offline', 'down'] and \
  93                         changed_greaterthan(nodehist.last_changed,1.0) and \
  94                         not found_between(recent_actions, 'try_reboot', 3.5, 1):
  95
  96                                 sitehist.attemptReboot(host)
  97                                 print "send message for host %s try_reboot" % host
  98
  99                 # NOTE: non-intuitive is that found_between(try_reboot, 3.5, 1)
 100                 #               will be false for a day after the above condition is satisfied
 101                 if nodehist.haspcu and nodehist.status in ['offline', 'down'] and \
 102                         changed_greaterthan(nodehist.last_changed,1.5) and \
 103                         found_between(recent_actions, 'try_reboot', 3.5, 1) and \
 104                         not found_within(recent_actions, 'pcufailed_notice', 3.5):
 105
 106                                 # send pcu failure message
 107                                 #act = ActionRecord(**kwargs)
 108                                 sitehist.sendMessage('pcufailed_notice', hostname=host)
 109                                 print "send message for host %s PCU Failure" % host
 110
 111                 if nodehist.status == 'monitordebug' and \
 112                         changed_greaterthan(nodehist.last_changed, 1) and \
 113                         not found_between(recent_actions, 'bootmanager_restore', 0.5, 0):
 114                                 # send down node notice
 115                                 # delay 0.5 days before retrying...
 116
 117                                 print "send message for host %s bootmanager_restore" % host
 118                                 sitehist.runBootManager(host)
 119                         #       sitehist.sendMessage('retry_bootman', hostname=host)
 120
 121                 if nodehist.status == 'down' and \
 122                         changed_greaterthan(nodehist.last_changed, 2) and \
 123                         not found_within(recent_actions, 'down_notice', 3.5):
 124                                 # send down node notice
 125
 126                                 sitehist.sendMessage('down_notice', hostname=host)
 127                                 print "send message for host %s down" % host
 128
 129                 node_count = node_count + 1
 130                 print "time: ", time.strftime('%Y-%m-%d %H:%M:%S')
 131                 sys.stdout.flush()
 132                 session.flush()
 133
 134         for i,site in enumerate(sitenames):
 135                 sitehist = SiteInterface.get_or_make(loginbase=site)
 136                 siteblack = BlacklistRecord.get_by(loginbase=site)
 137
 138                 if siteblack and not siteblack.expired():
 139                         print "skipping %s due to blacklist.  will expire %s" % (site, siteblack.willExpire() )
 140                         continue
 141
 142                 # TODO: make query only return records within a certin time range,
 143                 #               i.e. greater than 0.5 days ago. or 5 days, etc.
 144                 recent_actions = sitehist.getRecentActions(loginbase=site)
 145
 146                 print "%s %s %s" % (i, sitehist.db.loginbase, sitehist.db.status)
 147                 if sitehist.db.status == 'down':
 148                         if  not found_within(recent_actions, 'pause_penalty', 30) and \
 149                                 not found_within(recent_actions, 'increase_penalty', 7) and \
 150                                 changed_greaterthan(sitehist.db.last_changed, 7):
 151
 152                                 # TODO: catch errors
 153                                 sitehist.increasePenalty()
 154                                 #sitehist.applyPenalty()
 155                                 sitehist.sendMessage('increase_penalty')
 156
 157                                 print "send message for site %s penalty increase" % site
 158
 159                 if sitehist.db.status == 'good':
 160                         # clear penalty
 161                         # NOTE: because 'all clear' should have an indefinite status, we
 162                         #               have a boolean value rather than a 'recent action'
 163                         if sitehist.db.penalty_applied:
 164                                 # send message that penalties are cleared.
 165
 166                                 sitehist.clearPenalty()
 167                                 #sitehist.applyPenalty()
 168                                 sitehist.sendMessage('clear_penalty')
 169                                 sitehist.closeTicket()
 170
 171                                 print "send message for site %s penalty cleared" % site
 172
 173                 # find all ticket ids for site ( could be on the site record? )
 174                 # determine if there are penalties within the last 30 days?
 175                 # if so, add a 'pause_penalty' action.
 176                 if sitehist.db.message_id != 0 and sitehist.db.message_status == 'open' and \
 177                         sitehist.db.penalty_level > 0 and not found_within(recent_actions, 'pause_penalty', 30):
 178                         #       pause escalation
 179                         print "Pausing penalties for %s" % site
 180                         sitehist.pausePenalty()
 181
 182                 site_count = site_count + 1
 183
 184                 print "time: ", time.strftime('%Y-%m-%d %H:%M:%S')
 185                 sys.stdout.flush()
 186                 session.flush()
 187
 188         session.flush()
 189         return
 190
 191
 192 if __name__ == "__main__":
 193         parser = parsermodule.getParser(['nodesets'])
 194         parser.set_defaults( timewait=0,
 195                                                 skip=0,
 196                                                 rins=False,
 197                                                 reboot=False,
 198                                                 findbad=False,
 199                                                 force=False,
 200                                                 nosetup=False,
 201                                                 verbose=False,
 202                                                 quiet=False,)
 203
 204         parser.add_option("", "--stopselect", dest="stopselect", metavar="",
 205                                                 help="The select string that must evaluate to true for the node to be considered 'done'")
 206         parser.add_option("", "--findbad", dest="findbad", action="store_true",
 207                                                 help="Re-run findbad on the nodes we're going to check before acting.")
 208         parser.add_option("", "--force", dest="force", action="store_true",
 209                                                 help="Force action regardless of previous actions/logs.")
 210         parser.add_option("", "--rins", dest="rins", action="store_true",
 211                                                 help="Set the boot_state to 'rins' for all nodes.")
 212         parser.add_option("", "--reboot", dest="reboot", action="store_true",
 213                                                 help="Actively try to reboot the nodes, keeping a log of actions.")
 214
 215         parser.add_option("", "--verbose", dest="verbose", action="store_true",
 216                                                 help="Extra debug output messages.")
 217         parser.add_option("", "--nosetup", dest="nosetup", action="store_true",
 218                                                 help="Do not perform the orginary setup phase.")
 219         parser.add_option("", "--skip", dest="skip",
 220                                                 help="Number of machines to skip on the input queue.")
 221         parser.add_option("", "--timewait", dest="timewait",
 222                                                 help="Minutes to wait between iterations of 10 nodes.")
 223
 224         parser = parsermodule.getParser(['defaults'], parser)
 225         config = parsermodule.parse_args(parser)
 226
 227         fbquery = HistoryNodeRecord.query.all()
 228         hostnames = [ n.hostname for n in fbquery ]
 229
 230         fbquery = HistorySiteRecord.query.all()
 231         sitenames = [ s.loginbase for s in fbquery ]
 232
 233         if config.site:
 234                 # TODO: replace with calls to local db.  the api fails so often that
 235                 #               these calls should be regarded as unreliable.
 236                 l_nodes = plccache.GetNodesBySite(config.site)
 237                 filter_hostnames = [ n['hostname'] for n in l_nodes ]
 238
 239                 hostnames = filter(lambda x: x in filter_hostnames, hostnames)
 240                 sitenames = [config.site]
 241
 242         if config.node:
 243                 hostnames = [ config.node ]
 244                 sitenames = [ plccache.plcdb_hn2lb[config.node] ]
 245
 246         try:
 247                 main(hostnames, sitenames)
 248                 session.flush()
 249         except KeyboardInterrupt:
 250                 print "Killed by interrupt"
 251                 session.flush()
 252                 sys.exit(0)
 253         except:
 254                 #email_exception()
 255                 print traceback.print_exc();
 256                 print "fail all..."