findbad.py

   1 #!/usr/bin/python
   2
   3 import os
   4 import sys
   5 import string
   6 import time
   7 import config
   8 import util.file
   9
  10
  11 # QUERY all nodes.
  12 COMON_COTOPURL= "http://summer.cs.princeton.edu/status/tabulator.cgi?" + \
  13                                 "table=table_nodeview&" + \
  14                                 "dumpcols='name,resptime,sshstatus,uptime,lastcotop,cpuspeed,memsize,disksize'&" + \
  15                                 "formatcsv"
  16                                     #"formatcsv&" + \
  17                                         #"select='lastcotop!=0'"
  18
  19 import threading
  20 plc_lock = threading.Lock()
  21 round = 1
  22 externalState = {'round': round, 'nodes': {}}
  23 count = 0
  24
  25
  26 import database
  27 import moncommands
  28 import comon
  29 import threadpool
  30 import syncplcdb
  31 from nodequery import verify,query_to_dict,node_select
  32 import traceback
  33 import plc
  34 api = plc.getAuthAPI()
  35
  36 def collectPingAndSSH(nodename, cohash):
  37         ### RUN PING ######################
  38         ping = moncommands.CMD()
  39         (oval,errval) = ping.run_noexcept("ping -c 1 -q %s | grep rtt" % nodename)
  40
  41         values = {}
  42
  43         if oval == "":
  44                 # An error occurred
  45                 values['ping'] = "NOPING"
  46         else:
  47                 values['ping'] = "PING"
  48
  49         try:
  50                 for port in [22, 806]:
  51                         ssh = moncommands.SSH('root', nodename, port)
  52
  53                         (oval, errval) = ssh.run_noexcept2(""" <<\EOF
  54                                 echo "{"
  55                                 echo '  "kernel":"'`uname -a`'",'
  56                                 echo '  "bmlog":"'`ls /tmp/bm.log`'",'
  57                                 echo '  "bootcd":"'`cat /mnt/cdrom/bootme/ID`'",'
  58                                 echo '  "nm":"'`ps ax | grep nm.py | grep -v grep`'",'
  59                                 echo '  "princeton_comon":"'`ls -d /vservers/princeton_comon`'",'
  60
  61                                 ID=`grep princeton_comon /etc/passwd | awk -F : '{if ( $3 > 500 ) { print $3}}'`
  62
  63                                 echo '  "princeton_comon_running":"'`ls -d /proc/virtual/$ID`'",'
  64                                 echo '  "princeton_comon_procs":"'`vps ax | grep $ID | grep -v grep | wc -l`'",'
  65                                 echo "}"
  66 EOF                     """)
  67
  68                         if len(oval) > 0:
  69                                 values.update(eval(oval))
  70                                 values['sshport'] = port
  71                                 break
  72                         else:
  73                                 values.update({'kernel': "", 'bmlog' : "", 'bootcd' : '', 'nm' :
  74                                 '', 'princeton_comon' : '', 'princeton_comon_running' : '',
  75                                 'princeton_comon_procs' : '', 'sshport' : None})
  76         except:
  77                 print traceback.print_exc()
  78                 sys.exit(1)
  79
  80         ### RUN SSH ######################
  81         b_getbootcd_id = True
  82         #ssh = moncommands.SSH('root', nodename)
  83         #oval = ""
  84         #errval = ""
  85         #(oval, errval) = ssh.run_noexcept('echo `uname -a ; ls /tmp/bm.log`')
  86
  87         oval = values['kernel']
  88         if "2.6.17" in oval or "2.6.2" in oval:
  89                 values['ssh'] = 'SSH'
  90                 values['category'] = 'ALPHA'
  91                 if "bm.log" in values['bmlog']:
  92                         values['state'] = 'DEBUG'
  93                 else:
  94                         values['state'] = 'BOOT'
  95         elif "2.6.12" in oval or "2.6.10" in oval:
  96                 values['ssh'] = 'SSH'
  97                 values['category'] = 'PROD'
  98                 if "bm.log" in values['bmlog']:
  99                         values['state'] = 'DEBUG'
 100                 else:
 101                         values['state'] = 'BOOT'
 102
 103         # NOTE: on 2.6.8 kernels, with 4.2 bootstrapfs, the chroot command fails.  I have no idea why.
 104         elif "2.4" in oval or "2.6.8" in oval:
 105                 b_getbootcd_id = False
 106                 values['ssh'] = 'SSH'
 107                 values['category'] = 'OLDBOOTCD'
 108                 values['state'] = 'DEBUG'
 109         elif oval != "":
 110                 values['ssh'] = 'SSH'
 111                 values['category'] = 'UNKNOWN'
 112                 if "bm.log" in values['bmlog']:
 113                         values['state'] = 'DEBUG'
 114                 else:
 115                         values['state'] = 'BOOT'
 116         else:
 117                 # An error occurred.
 118                 b_getbootcd_id = False
 119                 values['ssh'] = 'NOSSH'
 120                 values['category'] = 'ERROR'
 121                 values['state'] = 'DOWN'
 122                 val = errval.strip()
 123                 values['kernel'] = val
 124
 125         #values['kernel'] = val
 126
 127         if b_getbootcd_id:
 128                 # try to get BootCD for all nodes that are not 2.4 nor inaccessible
 129                 #(oval, errval) = ssh.run_noexcept('cat /mnt/cdrom/bootme/ID')
 130                 oval = values['bootcd']
 131                 if "BootCD" in oval:
 132                         values['bootcd'] = oval
 133                         if "v2" in oval and \
 134                                 ( nodename is not "planetlab1.cs.unc.edu" and \
 135                                   nodename is not "planetlab2.cs.unc.edu" ):
 136                                 values['category'] = 'OLDBOOTCD'
 137                 else:
 138                         values['bootcd'] = ""
 139         else:
 140                 values['bootcd'] = ""
 141
 142         # TODO: get bm.log for debug nodes.
 143         # 'zcat /tmp/bm.log'
 144
 145         #(oval, errval) = ssh.run_noexcept('ps ax | grep nm.py | grep -v grep')
 146         oval = values['nm']
 147         if "nm.py" in oval:
 148                 values['nm'] = "Y"
 149         else:
 150                 values['nm'] = "N"
 151
 152         continue_slice_check = True
 153         #(oval, errval) = ssh.run_noexcept('ls -d /vservers/princeton_comon')
 154         oval = values['princeton_comon']
 155         if "princeton_comon" in oval:
 156                 values['princeton_comon'] = "Y"
 157         else:
 158                 values['princeton_comon'] = "N"
 159                 continue_slice_check = False
 160
 161         if continue_slice_check:
 162                 #(oval, errval) = ssh.run_noexcept('ID=`grep princeton_comon /etc/passwd | awk -F : "{if ( \\\$3 > 500 ) { print \\\$3}}"`; ls -d /proc/virtual/$ID')
 163                 oval = values['princeton_comon_running']
 164                 if len(oval) > len('/proc/virtual/'):
 165                         values['princeton_comon_running'] = "Y"
 166                 else:
 167                         values['princeton_comon_running'] = "N"
 168                         continue_slice_check = False
 169         else:
 170                 values['princeton_comon_running'] = "-"
 171
 172         if continue_slice_check:
 173                 #(oval, errval) = ssh.run_noexcept('ID=`grep princeton_comon /etc/passwd | awk -F : "{if ( \\\$3 > 500 ) { print \\\$3}}"`; vps ax | grep $ID | grep -v grep | wc -l')
 174                 oval = values['princeton_comon_procs']
 175                 values['princeton_comon_procs'] = oval
 176         else:
 177                 values['princeton_comon_procs'] = "-"
 178
 179
 180         if nodename in cohash:
 181                 values['comonstats'] = cohash[nodename]
 182         else:
 183                 values['comonstats'] = {'resptime':  '-1',
 184                                                                 'uptime':    '-1',
 185                                                                 'sshstatus': '-1',
 186                                                                 'lastcotop': '-1',
 187                                                                 'cpuspeed' : "null",
 188                                                                 'disksize' : 'null',
 189                                                                 'memsize'  : 'null'}
 190         # include output value
 191         ### GET PLC NODE ######################
 192         b_except = False
 193         plc_lock.acquire()
 194
 195         try:
 196                 d_node = plc.getNodes({'hostname': nodename}, ['pcu_ids', 'site_id', 'date_created', 'last_updated', 'last_contact', 'boot_state', 'nodegroup_ids'])
 197         except:
 198                 b_except = True
 199                 traceback.print_exc()
 200
 201         plc_lock.release()
 202         if b_except: return (None, None)
 203
 204         site_id = -1
 205         if d_node and len(d_node) > 0:
 206                 pcu = d_node[0]['pcu_ids']
 207                 if len(pcu) > 0:
 208                         values['pcu'] = "PCU"
 209                 else:
 210                         values['pcu'] = "NOPCU"
 211                 site_id = d_node[0]['site_id']
 212                 last_contact = d_node[0]['last_contact']
 213                 nodegroups = [ i['name'] for i in api.GetNodeGroups(d_node[0]['nodegroup_ids']) ]
 214                 values['plcnode'] = {'status' : 'SUCCESS',
 215                                                         'pcu_ids': pcu,
 216                                                         'boot_state' : d_node[0]['boot_state'],
 217                                                         'site_id': site_id,
 218                                                         'nodegroups' : nodegroups,
 219                                                         'last_contact': last_contact,
 220                                                         'date_created': d_node[0]['date_created'],
 221                                                         'last_updated': d_node[0]['last_updated']}
 222         else:
 223                 values['pcu']     = "UNKNOWN"
 224                 values['plcnode'] = {'status' : "GN_FAILED"}
 225
 226
 227         ### GET PLC SITE ######################
 228         b_except = False
 229         plc_lock.acquire()
 230
 231         try:
 232                 d_site = plc.getSites({'site_id': site_id},
 233                                                         ['max_slices', 'slice_ids', 'node_ids', 'login_base'])
 234         except:
 235                 b_except = True
 236                 traceback.print_exc()
 237
 238         plc_lock.release()
 239         if b_except: return (None, None)
 240
 241         if d_site and len(d_site) > 0:
 242                 max_slices = d_site[0]['max_slices']
 243                 num_slices = len(d_site[0]['slice_ids'])
 244                 num_nodes = len(d_site[0]['node_ids'])
 245                 loginbase = d_site[0]['login_base']
 246                 values['plcsite'] = {'num_nodes' : num_nodes,
 247                                                         'max_slices' : max_slices,
 248                                                         'num_slices' : num_slices,
 249                                                         'login_base' : loginbase,
 250                                                         'status'     : 'SUCCESS'}
 251         else:
 252                 values['plcsite'] = {'status' : "GS_FAILED"}
 253
 254         values['checked'] = time.time()
 255
 256         return (nodename, values)
 257
 258 def recordPingAndSSH(request, result):
 259         global externalState
 260         global count
 261         (nodename, values) = result
 262
 263         if values is not None:
 264                 global_round = externalState['round']
 265                 externalState['nodes'][nodename]['values'] = values
 266                 externalState['nodes'][nodename]['round'] = global_round
 267
 268                 count += 1
 269                 print "%d %s %s" % (count, nodename, externalState['nodes'][nodename]['values'])
 270                 if count % 20 == 0:
 271                         database.dbDump(config.dbname, externalState)
 272
 273 # this will be called when an exception occurs within a thread
 274 def handle_exception(request, result):
 275         print "Exception occured in request %s" % request.requestID
 276         for i in result:
 277                 print "Result: %s" % i
 278
 279
 280 def checkAndRecordState(l_nodes, cohash):
 281         global externalState
 282         global count
 283         global_round = externalState['round']
 284
 285         tp = threadpool.ThreadPool(20)
 286
 287         # CREATE all the work requests
 288         for nodename in l_nodes:
 289                 if nodename not in externalState['nodes']:
 290                         externalState['nodes'][nodename] = {'round': 0, 'values': []}
 291
 292                 node_round   = externalState['nodes'][nodename]['round']
 293                 if node_round < global_round:
 294                         # recreate node stats when refreshed
 295                         #print "%s" % nodename
 296                         req = threadpool.WorkRequest(collectPingAndSSH, [nodename, cohash], {},
 297                                                                                  None, recordPingAndSSH, handle_exception)
 298                         tp.putRequest(req)
 299                 else:
 300                         # We just skip it, since it's "up to date"
 301                         count += 1
 302                         print "%d %s %s" % (count, nodename, externalState['nodes'][nodename]['values'])
 303                         pass
 304
 305         # WAIT while all the work requests are processed.
 306         begin = time.time()
 307         while 1:
 308                 try:
 309                         time.sleep(1)
 310                         tp.poll()
 311                         # if more than two hours
 312                         if time.time() - begin > (60*60*1.5):
 313                                 print "findbad.py has run out of time!!!!!!"
 314                                 database.dbDump(config.dbname, externalState)
 315                                 os._exit(1)
 316                 except KeyboardInterrupt:
 317                         print "Interrupted!"
 318                         break
 319                 except threadpool.NoResultsPending:
 320                         print "All results collected."
 321                         break
 322
 323         database.dbDump(config.dbname, externalState)
 324
 325
 326
 327 def main():
 328         global externalState
 329
 330         externalState = database.if_cached_else(1, config.dbname, lambda : externalState)
 331
 332         if config.increment:
 333                 # update global round number to force refreshes across all nodes
 334                 externalState['round'] += 1
 335
 336         cotop = comon.Comon()
 337         # lastcotop measures whether cotop is actually running.  this is a better
 338         # metric than sshstatus, or other values from CoMon
 339         cotop_url = COMON_COTOPURL
 340
 341         # history information for all nodes
 342         #cohash = {}
 343         cohash = cotop.coget(cotop_url)
 344         l_nodes = syncplcdb.create_plcdb()
 345         if config.nodelist:
 346                 f_nodes = util.file.getListFromFile(config.nodelist)
 347                 l_nodes = filter(lambda x: x['hostname'] in f_nodes, l_nodes)
 348         elif config.node:
 349                 f_nodes = [config.node]
 350                 l_nodes = filter(lambda x: x['hostname'] in f_nodes, l_nodes)
 351         elif config.nodegroup:
 352                 ng = api.GetNodeGroups({'name' : config.nodegroup})
 353                 l_nodes = api.GetNodes(ng[0]['node_ids'])
 354         elif config.site:
 355                 site = api.GetSites(config.site)
 356                 l_nodes = api.GetNodes(site[0]['node_ids'], ['hostname'])
 357
 358         l_nodes = [node['hostname'] for node in l_nodes]
 359
 360         # perform this query after the above options, so that the filter above
 361         # does not break.
 362         if config.nodeselect:
 363                 fb = database.dbLoad("findbad")
 364                 l_nodes = node_select(config.nodeselect, fb['nodes'].keys(), fb)
 365
 366         print "fetching %s hosts" % len(l_nodes)
 367
 368         checkAndRecordState(l_nodes, cohash)
 369
 370         return 0
 371
 372
 373 if __name__ == '__main__':
 374         import parser as parsermodule
 375
 376         parser = parsermodule.getParser(['nodesets'])
 377
 378         parser.set_defaults( increment=False, dbname="findbad", cachenodes=False)
 379         parser.add_option("", "--cachenodes", action="store_true",
 380                                                 help="Cache node lookup from PLC")
 381         parser.add_option("", "--dbname", dest="dbname", metavar="FILE",
 382                                                 help="Specify the name of the database to which the information is saved")
 383         parser.add_option("-i", "--increment", action="store_true", dest="increment",
 384                                                 help="Increment round number to force refresh or retry")
 385
 386         parser = parsermodule.getParser(['defaults'], parser)
 387
 388         cfg = parsermodule.parse_args(parser)
 389
 390         try:
 391                 main()
 392         except Exception, err:
 393                 print traceback.print_exc()
 394                 print "Exception: %s" % err
 395                 print "Saving data... exitting."
 396                 database.dbDump(config.dbname, externalState)
 397                 sys.exit(0)