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