[23] | 1 | #!/usr/bin/env python |
---|
[225] | 2 | # |
---|
| 3 | # This file is part of Jobmonarch |
---|
| 4 | # |
---|
[691] | 5 | # Copyright (C) 2006-2013 Ramon Bastiaans |
---|
[623] | 6 | # Copyright (C) 2007, 2009 Dave Love (SGE code) |
---|
[225] | 7 | # |
---|
| 8 | # Jobmonarch is free software; you can redistribute it and/or modify |
---|
| 9 | # it under the terms of the GNU General Public License as published by |
---|
| 10 | # the Free Software Foundation; either version 2 of the License, or |
---|
| 11 | # (at your option) any later version. |
---|
| 12 | # |
---|
| 13 | # Jobmonarch is distributed in the hope that it will be useful, |
---|
| 14 | # but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
| 15 | # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
---|
| 16 | # GNU General Public License for more details. |
---|
| 17 | # |
---|
| 18 | # You should have received a copy of the GNU General Public License |
---|
| 19 | # along with this program; if not, write to the Free Software |
---|
| 20 | # Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA |
---|
| 21 | # |
---|
[228] | 22 | # SVN $Id: jobmond.py 852 2013-05-13 19:50:05Z ramonb $ |
---|
[227] | 23 | # |
---|
[23] | 24 | |
---|
[694] | 25 | # vi :set ts=4 |
---|
| 26 | |
---|
[471] | 27 | import sys, getopt, ConfigParser, time, os, socket, string, re |
---|
[837] | 28 | import xdrlib, socket, syslog, xml, xml.sax, shlex, os.path, pwd |
---|
[318] | 29 | from xml.sax.handler import feature_namespaces |
---|
[623] | 30 | from collections import deque |
---|
[699] | 31 | from glob import glob |
---|
[318] | 32 | |
---|
[829] | 33 | VERSION='1.0' |
---|
[307] | 34 | |
---|
[471] | 35 | def usage( ver ): |
---|
| 36 | |
---|
[691] | 37 | print 'jobmond %s' %VERSION |
---|
[471] | 38 | |
---|
[691] | 39 | if ver: |
---|
| 40 | return 0 |
---|
[471] | 41 | |
---|
[691] | 42 | print |
---|
| 43 | print 'Purpose:' |
---|
| 44 | print ' The Job Monitoring Daemon (jobmond) reports batch jobs information and statistics' |
---|
| 45 | print ' to Ganglia, which can be viewed with Job Monarch web frontend' |
---|
| 46 | print |
---|
| 47 | print 'Usage: jobmond [OPTIONS]' |
---|
| 48 | print |
---|
| 49 | print ' -c, --config=FILE The configuration file to use (default: /etc/jobmond.conf)' |
---|
| 50 | print ' -p, --pidfile=FILE Use pid file to store the process id' |
---|
| 51 | print ' -h, --help Print help and exit' |
---|
| 52 | print ' -v, --version Print version and exit' |
---|
| 53 | print |
---|
[307] | 54 | |
---|
[212] | 55 | def processArgs( args ): |
---|
[26] | 56 | |
---|
[701] | 57 | SHORT_L = 'p:hvc:' |
---|
| 58 | LONG_L = [ 'help', 'config=', 'pidfile=', 'version' ] |
---|
[165] | 59 | |
---|
[704] | 60 | global PIDFILE, JOBMOND_CONF |
---|
[701] | 61 | PIDFILE = None |
---|
[61] | 62 | |
---|
[701] | 63 | JOBMOND_CONF = '/etc/jobmond.conf' |
---|
[354] | 64 | |
---|
[691] | 65 | try: |
---|
[68] | 66 | |
---|
[691] | 67 | opts, args = getopt.getopt( args, SHORT_L, LONG_L ) |
---|
[185] | 68 | |
---|
[691] | 69 | except getopt.GetoptError, detail: |
---|
[212] | 70 | |
---|
[691] | 71 | print detail |
---|
[851] | 72 | usage( False ) |
---|
[691] | 73 | sys.exit( 1 ) |
---|
[212] | 74 | |
---|
[691] | 75 | for opt, value in opts: |
---|
[212] | 76 | |
---|
[691] | 77 | if opt in [ '--config', '-c' ]: |
---|
| 78 | |
---|
[701] | 79 | JOBMOND_CONF = value |
---|
[212] | 80 | |
---|
[691] | 81 | if opt in [ '--pidfile', '-p' ]: |
---|
[212] | 82 | |
---|
[701] | 83 | PIDFILE = value |
---|
[691] | 84 | |
---|
| 85 | if opt in [ '--help', '-h' ]: |
---|
[307] | 86 | |
---|
[691] | 87 | usage( False ) |
---|
| 88 | sys.exit( 0 ) |
---|
[212] | 89 | |
---|
[691] | 90 | if opt in [ '--version', '-v' ]: |
---|
[471] | 91 | |
---|
[691] | 92 | usage( True ) |
---|
| 93 | sys.exit( 0 ) |
---|
[471] | 94 | |
---|
[701] | 95 | return loadConfig( JOBMOND_CONF ) |
---|
[212] | 96 | |
---|
[520] | 97 | class GangliaConfigParser: |
---|
| 98 | |
---|
[699] | 99 | def __init__( self, filename ): |
---|
[520] | 100 | |
---|
[699] | 101 | self.conf_lijst = [ ] |
---|
| 102 | self.conf_dict = { } |
---|
| 103 | self.filename = filename |
---|
| 104 | self.file_pointer = file( filename, 'r' ) |
---|
| 105 | self.lexx = shlex.shlex( self.file_pointer ) |
---|
| 106 | self.lexx.whitespace_split = True |
---|
[520] | 107 | |
---|
[699] | 108 | self.parse() |
---|
[520] | 109 | |
---|
[699] | 110 | def __del__( self ): |
---|
[520] | 111 | |
---|
[699] | 112 | """ |
---|
| 113 | Cleanup: close file descriptor |
---|
| 114 | """ |
---|
| 115 | |
---|
| 116 | self.file_pointer.close() |
---|
| 117 | del self.lexx |
---|
| 118 | del self.conf_lijst |
---|
| 119 | |
---|
[691] | 120 | def removeQuotes( self, value ): |
---|
[520] | 121 | |
---|
[699] | 122 | clean_value = value |
---|
| 123 | clean_value = clean_value.replace( "'", "" ) |
---|
| 124 | clean_value = clean_value.replace( '"', '' ) |
---|
| 125 | clean_value = clean_value.strip() |
---|
[520] | 126 | |
---|
[691] | 127 | return clean_value |
---|
[520] | 128 | |
---|
[699] | 129 | def removeBraces( self, value ): |
---|
[520] | 130 | |
---|
[699] | 131 | clean_value = value |
---|
| 132 | clean_value = clean_value.replace( "(", "" ) |
---|
| 133 | clean_value = clean_value.replace( ')', '' ) |
---|
| 134 | clean_value = clean_value.strip() |
---|
[520] | 135 | |
---|
[699] | 136 | return clean_value |
---|
[520] | 137 | |
---|
[699] | 138 | def parse( self ): |
---|
[520] | 139 | |
---|
[699] | 140 | """ |
---|
| 141 | Parse self.filename using shlex scanning. |
---|
| 142 | - Removes /* comments */ |
---|
| 143 | - Traverses (recursively) through all include () statements |
---|
| 144 | - Stores complete valid config tokens in self.conf_list |
---|
[520] | 145 | |
---|
[699] | 146 | i.e.: |
---|
| 147 | ['globals', |
---|
| 148 | '{', |
---|
| 149 | 'daemonize', |
---|
| 150 | '=', |
---|
| 151 | 'yes', |
---|
| 152 | 'setuid', |
---|
| 153 | '=', |
---|
| 154 | 'yes', |
---|
| 155 | 'user', |
---|
| 156 | '=', |
---|
| 157 | 'ganglia', |
---|
| 158 | 'debug_level', |
---|
| 159 | '=', |
---|
| 160 | '0', |
---|
| 161 | <etc> ] |
---|
| 162 | """ |
---|
[520] | 163 | |
---|
[699] | 164 | t = 'bogus' |
---|
| 165 | c = False |
---|
| 166 | i = False |
---|
[520] | 167 | |
---|
[699] | 168 | while t != self.lexx.eof: |
---|
| 169 | #print 'get token' |
---|
| 170 | t = self.lexx.get_token() |
---|
[520] | 171 | |
---|
[699] | 172 | if len( t ) >= 2: |
---|
[520] | 173 | |
---|
[699] | 174 | if len( t ) >= 4: |
---|
[520] | 175 | |
---|
[699] | 176 | if t[:2] == '/*' and t[-2:] == '*/': |
---|
[520] | 177 | |
---|
[699] | 178 | #print 'comment line' |
---|
| 179 | #print 'skipping: %s' %t |
---|
| 180 | continue |
---|
[520] | 181 | |
---|
[699] | 182 | if t == '/*' or t[:2] == '/*': |
---|
| 183 | c = True |
---|
| 184 | #print 'comment start' |
---|
| 185 | #print 'skipping: %s' %t |
---|
| 186 | continue |
---|
[520] | 187 | |
---|
[699] | 188 | if t == '*/' or t[-2:] == '*/': |
---|
| 189 | c = False |
---|
| 190 | #print 'skipping: %s' %t |
---|
| 191 | #print 'comment end' |
---|
| 192 | continue |
---|
| 193 | |
---|
| 194 | if c: |
---|
| 195 | #print 'skipping: %s' %t |
---|
| 196 | continue |
---|
| 197 | |
---|
| 198 | if t == 'include': |
---|
| 199 | i = True |
---|
| 200 | #print 'include start' |
---|
| 201 | #print 'skipping: %s' %t |
---|
| 202 | continue |
---|
| 203 | |
---|
| 204 | if i: |
---|
| 205 | |
---|
| 206 | #print 'include start: %s' %t |
---|
| 207 | |
---|
| 208 | t2 = self.removeQuotes( t ) |
---|
| 209 | t2 = self.removeBraces( t ) |
---|
| 210 | |
---|
| 211 | for in_file in glob( self.removeQuotes(t2) ): |
---|
| 212 | |
---|
| 213 | #print 'including file: %s' %in_file |
---|
| 214 | parse_infile = GangliaConfigParser( in_file ) |
---|
| 215 | |
---|
| 216 | self.conf_lijst = self.conf_lijst + parse_infile.getConfLijst() |
---|
| 217 | |
---|
| 218 | del parse_infile |
---|
| 219 | |
---|
| 220 | i = False |
---|
| 221 | #print 'include end' |
---|
| 222 | #print 'skipping: %s' %t |
---|
| 223 | continue |
---|
| 224 | |
---|
| 225 | #print 'keep: %s' %t |
---|
| 226 | self.conf_lijst.append( self.removeQuotes(t) ) |
---|
| 227 | |
---|
| 228 | def getConfLijst( self ): |
---|
| 229 | |
---|
| 230 | return self.conf_lijst |
---|
| 231 | |
---|
| 232 | def confListToDict( self, parent_list=None ): |
---|
| 233 | |
---|
| 234 | """ |
---|
| 235 | Recursively traverses a conf_list and creates dictionary from it |
---|
| 236 | """ |
---|
| 237 | |
---|
| 238 | new_dict = { } |
---|
| 239 | count = 0 |
---|
| 240 | skip = 0 |
---|
| 241 | |
---|
| 242 | if not parent_list: |
---|
| 243 | parent_list = self.conf_lijst |
---|
| 244 | |
---|
| 245 | #print 'entering confListToDict(): (parent) list size %s' %len(parent_list) |
---|
| 246 | |
---|
| 247 | for n, c in enumerate( parent_list ): |
---|
| 248 | |
---|
| 249 | count = count + 1 |
---|
| 250 | |
---|
| 251 | #print 'CL: n %d c %s' %(n, c) |
---|
| 252 | |
---|
| 253 | if skip > 0: |
---|
| 254 | |
---|
| 255 | #print '- skipped' |
---|
| 256 | skip = skip - 1 |
---|
| 257 | continue |
---|
| 258 | |
---|
| 259 | if (n+1) <= (len( parent_list )-1): |
---|
| 260 | |
---|
| 261 | if parent_list[(n+1)] == '{': |
---|
| 262 | |
---|
| 263 | if not new_dict.has_key( c ): |
---|
| 264 | new_dict[ c ] = [ ] |
---|
| 265 | |
---|
| 266 | (temp_new_dict, skip) = self.confListToDict( parent_list[(n+2):] ) |
---|
| 267 | new_dict[ c ].append( temp_new_dict ) |
---|
| 268 | |
---|
| 269 | if parent_list[(n+1)] == '=' and (n+2) <= (len( parent_list )-1): |
---|
| 270 | |
---|
| 271 | if not new_dict.has_key( c ): |
---|
| 272 | new_dict[ c ] = [ ] |
---|
| 273 | |
---|
| 274 | new_dict[ c ].append( parent_list[ (n+2) ] ) |
---|
| 275 | |
---|
| 276 | skip = 2 |
---|
| 277 | |
---|
| 278 | if parent_list[n] == '}': |
---|
| 279 | |
---|
| 280 | #print 'leaving confListToDict(): new dict = %s' %new_dict |
---|
| 281 | return (new_dict, count) |
---|
| 282 | |
---|
| 283 | def makeConfDict( self ): |
---|
| 284 | |
---|
| 285 | """ |
---|
| 286 | Walks through self.conf_list and creates a dictionary based upon config values |
---|
| 287 | |
---|
| 288 | i.e.: |
---|
| 289 | 'tcp_accept_channel': [{'acl': [{'access': [{'action': ['"allow"'], |
---|
| 290 | 'ip': ['"127.0.0.1"'], |
---|
| 291 | 'mask': ['32']}]}], |
---|
| 292 | 'port': ['8649']}], |
---|
| 293 | 'udp_recv_channel': [{'port': ['8649']}], |
---|
| 294 | 'udp_send_channel': [{'host': ['145.101.32.3'], |
---|
| 295 | 'port': ['8649']}, |
---|
| 296 | {'host': ['145.101.32.207'], |
---|
| 297 | 'port': ['8649']}]} |
---|
| 298 | """ |
---|
| 299 | |
---|
| 300 | new_dict = { } |
---|
| 301 | skip = 0 |
---|
| 302 | |
---|
| 303 | #print 'entering makeConfDict()' |
---|
| 304 | |
---|
| 305 | for n, c in enumerate( self.conf_lijst ): |
---|
| 306 | |
---|
| 307 | #print 'M: n %d c %s' %(n, c) |
---|
| 308 | |
---|
| 309 | if skip > 0: |
---|
| 310 | |
---|
| 311 | #print '- skipped' |
---|
| 312 | skip = skip - 1 |
---|
| 313 | continue |
---|
| 314 | |
---|
| 315 | if (n+1) <= (len( self.conf_lijst )-1): |
---|
| 316 | |
---|
| 317 | if self.conf_lijst[(n+1)] == '{': |
---|
| 318 | |
---|
| 319 | if not new_dict.has_key( c ): |
---|
| 320 | new_dict[ c ] = [ ] |
---|
| 321 | |
---|
| 322 | ( temp_new_dict, skip ) = self.confListToDict( self.conf_lijst[(n+2):] ) |
---|
| 323 | new_dict[ c ].append( temp_new_dict ) |
---|
| 324 | |
---|
| 325 | if self.conf_lijst[(n+1)] == '=' and (n+2) <= (len( self.conf_lijst )-1): |
---|
| 326 | |
---|
| 327 | if not new_dict.has_key( c ): |
---|
| 328 | new_dict[ c ] = [ ] |
---|
| 329 | |
---|
| 330 | new_dict[ c ].append( self.conf_lijst[ (n+2) ] ) |
---|
| 331 | |
---|
| 332 | skip = 2 |
---|
| 333 | |
---|
| 334 | self.conf_dict = new_dict |
---|
| 335 | #print 'leaving makeConfDict(): conf dict size %d' %len( self.conf_dict ) |
---|
| 336 | |
---|
| 337 | def checkConfDict( self ): |
---|
| 338 | |
---|
| 339 | if len( self.conf_lijst ) == 0: |
---|
| 340 | |
---|
| 341 | raise Exception("Something went wrong generating conf list for %s" %self.file_name ) |
---|
| 342 | |
---|
| 343 | if len( self.conf_dict ) == 0: |
---|
| 344 | |
---|
| 345 | self.makeConfDict() |
---|
| 346 | |
---|
| 347 | def getConfDict( self ): |
---|
| 348 | |
---|
| 349 | self.checkConfDict() |
---|
| 350 | return self.conf_dict |
---|
| 351 | |
---|
| 352 | def getUdpSendChannels( self ): |
---|
| 353 | |
---|
| 354 | self.checkConfDict() |
---|
| 355 | |
---|
[707] | 356 | udp_send_channels = [ ] # IP:PORT |
---|
| 357 | |
---|
| 358 | if not self.conf_dict.has_key( 'udp_send_channel' ): |
---|
| 359 | return None |
---|
| 360 | |
---|
| 361 | for u in self.conf_dict[ 'udp_send_channel' ]: |
---|
| 362 | |
---|
| 363 | if u.has_key( 'mcast_join' ): |
---|
| 364 | |
---|
| 365 | ip = u['mcast_join'][0] |
---|
| 366 | |
---|
| 367 | elif u.has_key( 'host' ): |
---|
| 368 | |
---|
| 369 | ip = u['host'][0] |
---|
| 370 | |
---|
| 371 | port = u['port'][0] |
---|
| 372 | |
---|
| 373 | udp_send_channels.append( ( ip, port ) ) |
---|
| 374 | |
---|
| 375 | if len( udp_send_channels ) == 0: |
---|
| 376 | return None |
---|
| 377 | |
---|
| 378 | return udp_send_channels |
---|
| 379 | |
---|
[699] | 380 | def getSectionLastOption( self, section, option ): |
---|
| 381 | |
---|
| 382 | """ |
---|
| 383 | Get last option set in a config section that could be set multiple times in multiple (include) files. |
---|
| 384 | |
---|
| 385 | i.e.: getSectionLastOption( 'globals', 'send_metadata_interval' ) |
---|
| 386 | """ |
---|
| 387 | |
---|
| 388 | self.checkConfDict() |
---|
| 389 | value = None |
---|
| 390 | |
---|
| 391 | if not self.conf_dict.has_key( section ): |
---|
| 392 | |
---|
| 393 | return None |
---|
| 394 | |
---|
| 395 | # Could be set multiple times in multiple (include) files: get last one set |
---|
| 396 | for c in self.conf_dict[ section ]: |
---|
| 397 | |
---|
| 398 | if c.has_key( option ): |
---|
| 399 | |
---|
[705] | 400 | value = c[ option ][0] |
---|
[699] | 401 | |
---|
[705] | 402 | return value |
---|
[699] | 403 | |
---|
| 404 | def getClusterName( self ): |
---|
| 405 | |
---|
| 406 | return self.getSectionLastOption( 'cluster', 'name' ) |
---|
| 407 | |
---|
| 408 | def getVal( self, section, option ): |
---|
| 409 | |
---|
| 410 | return self.getSectionLastOption( section, option ) |
---|
| 411 | |
---|
[691] | 412 | def getInt( self, section, valname ): |
---|
[520] | 413 | |
---|
[691] | 414 | value = self.getVal( section, valname ) |
---|
[520] | 415 | |
---|
[691] | 416 | if not value: |
---|
[699] | 417 | return None |
---|
[520] | 418 | |
---|
[691] | 419 | return int( value ) |
---|
[520] | 420 | |
---|
[691] | 421 | def getStr( self, section, valname ): |
---|
[520] | 422 | |
---|
[691] | 423 | value = self.getVal( section, valname ) |
---|
[520] | 424 | |
---|
[691] | 425 | if not value: |
---|
[699] | 426 | return None |
---|
[520] | 427 | |
---|
[691] | 428 | return str( value ) |
---|
[520] | 429 | |
---|
| 430 | def findGmetric(): |
---|
| 431 | |
---|
[691] | 432 | for dir in os.path.expandvars( '$PATH' ).split( ':' ): |
---|
[520] | 433 | |
---|
[691] | 434 | guess = '%s/%s' %( dir, 'gmetric' ) |
---|
[520] | 435 | |
---|
[691] | 436 | if os.path.exists( guess ): |
---|
[520] | 437 | |
---|
[691] | 438 | return guess |
---|
[520] | 439 | |
---|
[691] | 440 | return False |
---|
[520] | 441 | |
---|
[212] | 442 | def loadConfig( filename ): |
---|
| 443 | |
---|
[691] | 444 | def getlist( cfg_string ): |
---|
[215] | 445 | |
---|
[691] | 446 | my_list = [ ] |
---|
[215] | 447 | |
---|
[691] | 448 | for item_txt in cfg_string.split( ',' ): |
---|
[215] | 449 | |
---|
[691] | 450 | sep_char = None |
---|
[215] | 451 | |
---|
[691] | 452 | item_txt = item_txt.strip() |
---|
[215] | 453 | |
---|
[691] | 454 | for s_char in [ "'", '"' ]: |
---|
[215] | 455 | |
---|
[691] | 456 | if item_txt.find( s_char ) != -1: |
---|
[215] | 457 | |
---|
[691] | 458 | if item_txt.count( s_char ) != 2: |
---|
[215] | 459 | |
---|
[691] | 460 | print 'Missing quote: %s' %item_txt |
---|
| 461 | sys.exit( 1 ) |
---|
[215] | 462 | |
---|
[691] | 463 | else: |
---|
[215] | 464 | |
---|
[691] | 465 | sep_char = s_char |
---|
| 466 | break |
---|
[215] | 467 | |
---|
[691] | 468 | if sep_char: |
---|
[215] | 469 | |
---|
[691] | 470 | item_txt = item_txt.split( sep_char )[1] |
---|
[215] | 471 | |
---|
[691] | 472 | my_list.append( item_txt ) |
---|
[215] | 473 | |
---|
[691] | 474 | return my_list |
---|
[215] | 475 | |
---|
[784] | 476 | if not os.path.isfile( JOBMOND_CONF ): |
---|
| 477 | |
---|
| 478 | print "Is not a file or does not exist: '%s'" %JOBMOND_CONF |
---|
| 479 | sys.exit( 1 ) |
---|
| 480 | |
---|
| 481 | try: |
---|
| 482 | f = open( JOBMOND_CONF, 'r' ) |
---|
| 483 | except IOError, detail: |
---|
| 484 | print "Cannot read config file: '%s'" %JOBMOND_CONF |
---|
| 485 | sys.exit( 1 ) |
---|
| 486 | else: |
---|
| 487 | f.close() |
---|
| 488 | |
---|
[691] | 489 | cfg = ConfigParser.ConfigParser() |
---|
[212] | 490 | |
---|
[691] | 491 | cfg.read( filename ) |
---|
[212] | 492 | |
---|
[691] | 493 | global DEBUG_LEVEL, DAEMONIZE, BATCH_SERVER, BATCH_POLL_INTERVAL |
---|
| 494 | global GMOND_CONF, DETECT_TIME_DIFFS, BATCH_HOST_TRANSLATE |
---|
| 495 | global BATCH_API, QUEUE, GMETRIC_TARGET, USE_SYSLOG |
---|
| 496 | global SYSLOG_LEVEL, SYSLOG_FACILITY, GMETRIC_BINARY |
---|
[707] | 497 | global METRIC_MAX_VAL_LEN, GMOND_UDP_SEND_CHANNELS |
---|
[212] | 498 | |
---|
[701] | 499 | DEBUG_LEVEL = cfg.getint( 'DEFAULT', 'DEBUG_LEVEL' ) |
---|
[212] | 500 | |
---|
[701] | 501 | DAEMONIZE = cfg.getboolean( 'DEFAULT', 'DAEMONIZE' ) |
---|
[212] | 502 | |
---|
[691] | 503 | SYSLOG_LEVEL = -1 |
---|
[701] | 504 | SYSLOG_FACILITY = None |
---|
[377] | 505 | |
---|
[691] | 506 | try: |
---|
[701] | 507 | USE_SYSLOG = cfg.getboolean( 'DEFAULT', 'USE_SYSLOG' ) |
---|
[212] | 508 | |
---|
[691] | 509 | except ConfigParser.NoOptionError: |
---|
[373] | 510 | |
---|
[701] | 511 | USE_SYSLOG = True |
---|
[373] | 512 | |
---|
[691] | 513 | debug_msg( 0, 'ERROR: no option USE_SYSLOG found: assuming yes' ) |
---|
[373] | 514 | |
---|
[691] | 515 | if USE_SYSLOG: |
---|
[373] | 516 | |
---|
[691] | 517 | try: |
---|
[701] | 518 | SYSLOG_LEVEL = cfg.getint( 'DEFAULT', 'SYSLOG_LEVEL' ) |
---|
[373] | 519 | |
---|
[691] | 520 | except ConfigParser.NoOptionError: |
---|
[373] | 521 | |
---|
[691] | 522 | debug_msg( 0, 'ERROR: no option SYSLOG_LEVEL found: assuming level 0' ) |
---|
[701] | 523 | SYSLOG_LEVEL = 0 |
---|
[373] | 524 | |
---|
[691] | 525 | try: |
---|
[373] | 526 | |
---|
[691] | 527 | SYSLOG_FACILITY = eval( 'syslog.LOG_' + cfg.get( 'DEFAULT', 'SYSLOG_FACILITY' ) ) |
---|
[373] | 528 | |
---|
[691] | 529 | except ConfigParser.NoOptionError: |
---|
[373] | 530 | |
---|
[691] | 531 | SYSLOG_FACILITY = syslog.LOG_DAEMON |
---|
[373] | 532 | |
---|
[691] | 533 | debug_msg( 0, 'ERROR: no option SYSLOG_FACILITY found: assuming facility DAEMON' ) |
---|
[373] | 534 | |
---|
[691] | 535 | try: |
---|
[373] | 536 | |
---|
[701] | 537 | BATCH_SERVER = cfg.get( 'DEFAULT', 'BATCH_SERVER' ) |
---|
[212] | 538 | |
---|
[691] | 539 | except ConfigParser.NoOptionError: |
---|
[265] | 540 | |
---|
[691] | 541 | # Backwards compatibility for old configs |
---|
| 542 | # |
---|
[265] | 543 | |
---|
[701] | 544 | BATCH_SERVER = cfg.get( 'DEFAULT', 'TORQUE_SERVER' ) |
---|
| 545 | api_guess = 'pbs' |
---|
[691] | 546 | |
---|
| 547 | try: |
---|
| 548 | |
---|
[701] | 549 | BATCH_POLL_INTERVAL = cfg.getint( 'DEFAULT', 'BATCH_POLL_INTERVAL' ) |
---|
[265] | 550 | |
---|
[691] | 551 | except ConfigParser.NoOptionError: |
---|
[265] | 552 | |
---|
[691] | 553 | # Backwards compatibility for old configs |
---|
| 554 | # |
---|
[265] | 555 | |
---|
[701] | 556 | BATCH_POLL_INTERVAL = cfg.getint( 'DEFAULT', 'TORQUE_POLL_INTERVAL' ) |
---|
| 557 | api_guess = 'pbs' |
---|
[691] | 558 | |
---|
| 559 | try: |
---|
[212] | 560 | |
---|
[701] | 561 | GMOND_CONF = cfg.get( 'DEFAULT', 'GMOND_CONF' ) |
---|
[353] | 562 | |
---|
[691] | 563 | except ConfigParser.NoOptionError: |
---|
[353] | 564 | |
---|
[703] | 565 | # Not specified: assume /etc/ganglia/gmond.conf |
---|
[691] | 566 | # |
---|
[703] | 567 | GMOND_CONF = '/etc/ganglia/gmond.conf' |
---|
[353] | 568 | |
---|
[707] | 569 | ganglia_cfg = GangliaConfigParser( GMOND_CONF ) |
---|
| 570 | GMETRIC_TARGET = None |
---|
[449] | 571 | |
---|
[707] | 572 | GMOND_UDP_SEND_CHANNELS = ganglia_cfg.getUdpSendChannels() |
---|
[449] | 573 | |
---|
[707] | 574 | if not GMOND_UDP_SEND_CHANNELS: |
---|
[449] | 575 | |
---|
[701] | 576 | debug_msg( 0, "WARNING: Can't parse udp_send_channel from: '%s' - Trying: %s" %( GMOND_CONF, JOBMOND_CONF ) ) |
---|
[520] | 577 | |
---|
[691] | 578 | # Couldn't figure it out: let's see if it's in our jobmond.conf |
---|
| 579 | # |
---|
| 580 | try: |
---|
[520] | 581 | |
---|
[691] | 582 | GMETRIC_TARGET = cfg.get( 'DEFAULT', 'GMETRIC_TARGET' ) |
---|
[520] | 583 | |
---|
[691] | 584 | # Guess not: now just give up |
---|
[701] | 585 | |
---|
[691] | 586 | except ConfigParser.NoOptionError: |
---|
[520] | 587 | |
---|
[691] | 588 | GMETRIC_TARGET = None |
---|
[520] | 589 | |
---|
[691] | 590 | debug_msg( 0, "ERROR: GMETRIC_TARGET not set: internal Gmetric handling aborted. Failing back to DEPRECATED use of gmond.conf/gmetric binary. This will slow down jobmond significantly!" ) |
---|
[520] | 591 | |
---|
[701] | 592 | gmetric_bin = findGmetric() |
---|
[520] | 593 | |
---|
[701] | 594 | if gmetric_bin: |
---|
[520] | 595 | |
---|
[701] | 596 | GMETRIC_BINARY = gmetric_bin |
---|
| 597 | else: |
---|
| 598 | debug_msg( 0, "WARNING: Can't find gmetric binary anywhere in $PATH" ) |
---|
[520] | 599 | |
---|
[701] | 600 | try: |
---|
[520] | 601 | |
---|
[701] | 602 | GMETRIC_BINARY = cfg.get( 'DEFAULT', 'GMETRIC_BINARY' ) |
---|
[520] | 603 | |
---|
[701] | 604 | except ConfigParser.NoOptionError: |
---|
[520] | 605 | |
---|
[786] | 606 | print "FATAL ERROR: GMETRIC_BINARY not set and not in $PATH" |
---|
[701] | 607 | sys.exit( 1 ) |
---|
[520] | 608 | |
---|
[701] | 609 | #TODO: is this really still needed or should be automatic |
---|
[691] | 610 | DETECT_TIME_DIFFS = cfg.getboolean( 'DEFAULT', 'DETECT_TIME_DIFFS' ) |
---|
[212] | 611 | |
---|
[701] | 612 | BATCH_HOST_TRANSLATE = getlist( cfg.get( 'DEFAULT', 'BATCH_HOST_TRANSLATE' ) ) |
---|
[215] | 613 | |
---|
[691] | 614 | try: |
---|
[256] | 615 | |
---|
[691] | 616 | BATCH_API = cfg.get( 'DEFAULT', 'BATCH_API' ) |
---|
[266] | 617 | |
---|
[691] | 618 | except ConfigParser.NoOptionError, detail: |
---|
[266] | 619 | |
---|
[691] | 620 | if BATCH_SERVER and api_guess: |
---|
[354] | 621 | |
---|
[691] | 622 | BATCH_API = api_guess |
---|
| 623 | else: |
---|
[786] | 624 | print "FATAL ERROR: BATCH_API not set and can't make guess" |
---|
[691] | 625 | sys.exit( 1 ) |
---|
[317] | 626 | |
---|
[691] | 627 | try: |
---|
[317] | 628 | |
---|
[691] | 629 | QUEUE = getlist( cfg.get( 'DEFAULT', 'QUEUE' ) ) |
---|
[317] | 630 | |
---|
[691] | 631 | except ConfigParser.NoOptionError, detail: |
---|
[317] | 632 | |
---|
[691] | 633 | QUEUE = None |
---|
[353] | 634 | |
---|
[701] | 635 | METRIC_MAX_VAL_LEN = ganglia_cfg.getInt( 'globals', 'max_udp_msg_len' ) |
---|
| 636 | |
---|
[691] | 637 | return True |
---|
[212] | 638 | |
---|
[507] | 639 | def fqdn_parts (fqdn): |
---|
[520] | 640 | |
---|
[691] | 641 | """Return pair of host and domain for fully-qualified domain name arg.""" |
---|
[520] | 642 | |
---|
[691] | 643 | parts = fqdn.split (".") |
---|
[520] | 644 | |
---|
[691] | 645 | return (parts[0], string.join(parts[1:], ".")) |
---|
[507] | 646 | |
---|
[61] | 647 | class DataProcessor: |
---|
[355] | 648 | |
---|
[691] | 649 | """Class for processing of data""" |
---|
[61] | 650 | |
---|
[691] | 651 | binary = None |
---|
[61] | 652 | |
---|
[691] | 653 | def __init__( self, binary=None ): |
---|
[355] | 654 | |
---|
[691] | 655 | """Remember alternate binary location if supplied""" |
---|
[61] | 656 | |
---|
[691] | 657 | global GMETRIC_BINARY, GMOND_CONF |
---|
[449] | 658 | |
---|
[691] | 659 | if binary: |
---|
| 660 | self.binary = binary |
---|
[61] | 661 | |
---|
[707] | 662 | if not self.binary and not GMETRIC_TARGET and not GMOND_UDP_SEND_CHANNELS: |
---|
[691] | 663 | self.binary = GMETRIC_BINARY |
---|
[449] | 664 | |
---|
[691] | 665 | # Timeout for XML |
---|
| 666 | # |
---|
| 667 | # From ganglia's documentation: |
---|
| 668 | # |
---|
| 669 | # 'A metric will be deleted DMAX seconds after it is received, and |
---|
| 670 | # DMAX=0 means eternal life.' |
---|
[61] | 671 | |
---|
[691] | 672 | self.dmax = str( int( int( BATCH_POLL_INTERVAL ) * 2 ) ) |
---|
[80] | 673 | |
---|
[707] | 674 | if GMOND_CONF and not GMETRIC_TARGET and not GMOND_UDP_SEND_CHANNELS: |
---|
[354] | 675 | |
---|
[691] | 676 | incompatible = self.checkGmetricVersion() |
---|
[61] | 677 | |
---|
[691] | 678 | if incompatible: |
---|
[355] | 679 | |
---|
[786] | 680 | print 'Ganglia/Gmetric version not compatible, please upgrade to at least 3.4.0' |
---|
[691] | 681 | sys.exit( 1 ) |
---|
[65] | 682 | |
---|
[691] | 683 | def checkGmetricVersion( self ): |
---|
[355] | 684 | |
---|
[691] | 685 | """ |
---|
[692] | 686 | Check version of gmetric is at least 3.4.0 |
---|
[691] | 687 | for the syntax we use |
---|
| 688 | """ |
---|
[65] | 689 | |
---|
[691] | 690 | global METRIC_MAX_VAL_LEN, GMETRIC_TARGET |
---|
[255] | 691 | |
---|
[691] | 692 | incompatible = 0 |
---|
[341] | 693 | |
---|
[691] | 694 | gfp = os.popen( self.binary + ' --version' ) |
---|
[692] | 695 | lines = gfp.readlines() |
---|
[65] | 696 | |
---|
[691] | 697 | gfp.close() |
---|
[355] | 698 | |
---|
[691] | 699 | for line in lines: |
---|
[355] | 700 | |
---|
[691] | 701 | line = line.split( ' ' ) |
---|
[65] | 702 | |
---|
[691] | 703 | if len( line ) == 2 and str( line ).find( 'gmetric' ) != -1: |
---|
| 704 | |
---|
| 705 | gmetric_version = line[1].split( '\n' )[0] |
---|
[65] | 706 | |
---|
[691] | 707 | version_major = int( gmetric_version.split( '.' )[0] ) |
---|
| 708 | version_minor = int( gmetric_version.split( '.' )[1] ) |
---|
| 709 | version_patch = int( gmetric_version.split( '.' )[2] ) |
---|
[65] | 710 | |
---|
[691] | 711 | incompatible = 0 |
---|
[65] | 712 | |
---|
[691] | 713 | if version_major < 3: |
---|
[65] | 714 | |
---|
[691] | 715 | incompatible = 1 |
---|
| 716 | |
---|
| 717 | elif version_major == 3: |
---|
[65] | 718 | |
---|
[692] | 719 | if version_minor < 4: |
---|
[65] | 720 | |
---|
[692] | 721 | incompatible = 1 |
---|
[65] | 722 | |
---|
[691] | 723 | return incompatible |
---|
[65] | 724 | |
---|
[691] | 725 | def multicastGmetric( self, metricname, metricval, valtype='string', units='' ): |
---|
[355] | 726 | |
---|
[691] | 727 | """Call gmetric binary and multicast""" |
---|
[65] | 728 | |
---|
[691] | 729 | cmd = self.binary |
---|
[65] | 730 | |
---|
[707] | 731 | if GMOND_UDP_SEND_CHANNELS: |
---|
[61] | 732 | |
---|
[707] | 733 | for c_ip, c_port in GMOND_UDP_SEND_CHANNELS: |
---|
| 734 | |
---|
| 735 | metric_debug = "[gmetric %s:%s] name: %s - val: %s - dmax: %s" %( str(c_ip), str(c_port), str( metricname ), str( metricval ), str( self.dmax ) ) |
---|
| 736 | |
---|
| 737 | debug_msg( 10, printTime() + ' ' + metric_debug) |
---|
| 738 | |
---|
| 739 | gm = Gmetric( c_ip, c_port ) |
---|
| 740 | |
---|
| 741 | gm.send( str( metricname ), str( metricval ), str( self.dmax ), valtype, units ) |
---|
| 742 | |
---|
| 743 | elif GMETRIC_TARGET: |
---|
| 744 | |
---|
[691] | 745 | GMETRIC_TARGET_HOST = GMETRIC_TARGET.split( ':' )[0] |
---|
| 746 | GMETRIC_TARGET_PORT = GMETRIC_TARGET.split( ':' )[1] |
---|
[353] | 747 | |
---|
[691] | 748 | metric_debug = "[gmetric] name: %s - val: %s - dmax: %s" %( str( metricname ), str( metricval ), str( self.dmax ) ) |
---|
[353] | 749 | |
---|
[691] | 750 | debug_msg( 10, printTime() + ' ' + metric_debug) |
---|
[353] | 751 | |
---|
[691] | 752 | gm = Gmetric( GMETRIC_TARGET_HOST, GMETRIC_TARGET_PORT ) |
---|
[353] | 753 | |
---|
[691] | 754 | gm.send( str( metricname ), str( metricval ), str( self.dmax ), valtype, units ) |
---|
[353] | 755 | |
---|
[691] | 756 | else: |
---|
| 757 | try: |
---|
| 758 | cmd = cmd + ' -c' + GMOND_CONF |
---|
[353] | 759 | |
---|
[691] | 760 | except NameError: |
---|
[353] | 761 | |
---|
[705] | 762 | debug_msg( 10, 'Assuming /etc/ganglia/gmond.conf for gmetric cmd' ) |
---|
[353] | 763 | |
---|
[691] | 764 | cmd = cmd + ' -n' + str( metricname )+ ' -v"' + str( metricval )+ '" -t' + str( valtype ) + ' -d' + str( self.dmax ) |
---|
[353] | 765 | |
---|
[691] | 766 | if len( units ) > 0: |
---|
[409] | 767 | |
---|
[691] | 768 | cmd = cmd + ' -u"' + units + '"' |
---|
[409] | 769 | |
---|
[691] | 770 | debug_msg( 10, printTime() + ' ' + cmd ) |
---|
[353] | 771 | |
---|
[691] | 772 | os.system( cmd ) |
---|
[353] | 773 | |
---|
[318] | 774 | class DataGatherer: |
---|
[23] | 775 | |
---|
[691] | 776 | """Skeleton class for batch system DataGatherer""" |
---|
[256] | 777 | |
---|
[691] | 778 | def printJobs( self, jobs ): |
---|
[355] | 779 | |
---|
[691] | 780 | """Print a jobinfo overview""" |
---|
[318] | 781 | |
---|
[691] | 782 | for name, attrs in self.jobs.items(): |
---|
[318] | 783 | |
---|
[691] | 784 | print 'job %s' %(name) |
---|
[318] | 785 | |
---|
[691] | 786 | for name, val in attrs.items(): |
---|
[318] | 787 | |
---|
[691] | 788 | print '\t%s = %s' %( name, val ) |
---|
[318] | 789 | |
---|
[691] | 790 | def printJob( self, jobs, job_id ): |
---|
[355] | 791 | |
---|
[691] | 792 | """Print job with job_id from jobs""" |
---|
[318] | 793 | |
---|
[691] | 794 | print 'job %s' %(job_id) |
---|
[318] | 795 | |
---|
[691] | 796 | for name, val in jobs[ job_id ].items(): |
---|
[318] | 797 | |
---|
[691] | 798 | print '\t%s = %s' %( name, val ) |
---|
[318] | 799 | |
---|
[691] | 800 | def getAttr( self, attrs, name ): |
---|
[507] | 801 | |
---|
[691] | 802 | """Return certain attribute from dictionary, if exists""" |
---|
[507] | 803 | |
---|
[691] | 804 | if attrs.has_key( name ): |
---|
[507] | 805 | |
---|
[691] | 806 | return attrs[ name ] |
---|
| 807 | else: |
---|
| 808 | return '' |
---|
[507] | 809 | |
---|
[691] | 810 | def jobDataChanged( self, jobs, job_id, attrs ): |
---|
[507] | 811 | |
---|
[691] | 812 | """Check if job with attrs and job_id in jobs has changed""" |
---|
[507] | 813 | |
---|
[691] | 814 | if jobs.has_key( job_id ): |
---|
[507] | 815 | |
---|
[691] | 816 | oldData = jobs[ job_id ] |
---|
| 817 | else: |
---|
| 818 | return 1 |
---|
[507] | 819 | |
---|
[691] | 820 | for name, val in attrs.items(): |
---|
[507] | 821 | |
---|
[691] | 822 | if oldData.has_key( name ): |
---|
[507] | 823 | |
---|
[691] | 824 | if oldData[ name ] != attrs[ name ]: |
---|
[507] | 825 | |
---|
[691] | 826 | return 1 |
---|
[507] | 827 | |
---|
[691] | 828 | else: |
---|
| 829 | return 1 |
---|
[507] | 830 | |
---|
[691] | 831 | return 0 |
---|
[507] | 832 | |
---|
[691] | 833 | def submitJobData( self ): |
---|
[507] | 834 | |
---|
[691] | 835 | """Submit job info list""" |
---|
[507] | 836 | |
---|
[691] | 837 | global BATCH_API |
---|
[512] | 838 | |
---|
[728] | 839 | self.dp.multicastGmetric( 'zplugin_monarch_heartbeat', str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
[507] | 840 | |
---|
[724] | 841 | running_jobs = 0 |
---|
| 842 | queued_jobs = 0 |
---|
[507] | 843 | |
---|
[691] | 844 | # Count how many running/queued jobs we found |
---|
| 845 | # |
---|
| 846 | for jobid, jobattrs in self.jobs.items(): |
---|
[507] | 847 | |
---|
[691] | 848 | if jobattrs[ 'status' ] == 'Q': |
---|
[507] | 849 | |
---|
[691] | 850 | queued_jobs += 1 |
---|
[507] | 851 | |
---|
[691] | 852 | elif jobattrs[ 'status' ] == 'R': |
---|
[507] | 853 | |
---|
[691] | 854 | running_jobs += 1 |
---|
[507] | 855 | |
---|
[691] | 856 | # Report running/queued jobs as seperate metric for a nice RRD graph |
---|
| 857 | # |
---|
[728] | 858 | self.dp.multicastGmetric( 'zplugin_monarch_rj', str( running_jobs ), 'uint32', 'jobs' ) |
---|
| 859 | self.dp.multicastGmetric( 'zplugin_monarch_qj', str( queued_jobs ), 'uint32', 'jobs' ) |
---|
[507] | 860 | |
---|
[691] | 861 | # Report down/offline nodes in batch (PBS only ATM) |
---|
| 862 | # |
---|
| 863 | if BATCH_API == 'pbs': |
---|
[512] | 864 | |
---|
[691] | 865 | domain = fqdn_parts( socket.getfqdn() )[1] |
---|
[514] | 866 | |
---|
[728] | 867 | downed_nodes = list() |
---|
| 868 | offline_nodes = list() |
---|
[691] | 869 | |
---|
| 870 | l = ['state'] |
---|
[512] | 871 | |
---|
[790] | 872 | nodelist = self.getNodeData() |
---|
| 873 | |
---|
| 874 | for name, node in nodelist.items(): |
---|
| 875 | |
---|
[691] | 876 | if ( node[ 'state' ].find( "down" ) != -1 ): |
---|
[512] | 877 | |
---|
[691] | 878 | downed_nodes.append( name ) |
---|
[512] | 879 | |
---|
[691] | 880 | if ( node[ 'state' ].find( "offline" ) != -1 ): |
---|
[512] | 881 | |
---|
[691] | 882 | offline_nodes.append( name ) |
---|
[512] | 883 | |
---|
[728] | 884 | downnodeslist = do_nodelist( downed_nodes ) |
---|
| 885 | offlinenodeslist = do_nodelist( offline_nodes ) |
---|
[512] | 886 | |
---|
[691] | 887 | down_str = 'nodes=%s domain=%s reported=%s' %( string.join( downnodeslist, ';' ), domain, str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
| 888 | offl_str = 'nodes=%s domain=%s reported=%s' %( string.join( offlinenodeslist, ';' ), domain, str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
[728] | 889 | self.dp.multicastGmetric( 'zplugin_monarch_down' , down_str ) |
---|
| 890 | self.dp.multicastGmetric( 'zplugin_monarch_offline', offl_str ) |
---|
[514] | 891 | |
---|
[691] | 892 | # Now let's spread the knowledge |
---|
| 893 | # |
---|
| 894 | for jobid, jobattrs in self.jobs.items(): |
---|
[507] | 895 | |
---|
[691] | 896 | # Make gmetric values for each job: respect max gmetric value length |
---|
| 897 | # |
---|
| 898 | gmetric_val = self.compileGmetricVal( jobid, jobattrs ) |
---|
| 899 | metric_increment = 0 |
---|
[507] | 900 | |
---|
[691] | 901 | # If we have more job info than max gmetric value length allows, split it up |
---|
| 902 | # amongst multiple metrics |
---|
| 903 | # |
---|
| 904 | for val in gmetric_val: |
---|
[507] | 905 | |
---|
[728] | 906 | metric_name = 'zplugin_monarch_job_%s_%s' %( str(metric_increment) , str( jobid ) ) |
---|
| 907 | self.dp.multicastGmetric( metric_name, val ) |
---|
[507] | 908 | |
---|
[691] | 909 | # Increase follow number if this jobinfo is split up amongst more than 1 gmetric |
---|
| 910 | # |
---|
| 911 | metric_increment = metric_increment + 1 |
---|
[507] | 912 | |
---|
[691] | 913 | def compileGmetricVal( self, jobid, jobattrs ): |
---|
[507] | 914 | |
---|
[691] | 915 | """Create a val string for gmetric of jobinfo""" |
---|
[507] | 916 | |
---|
[691] | 917 | gval_lists = [ ] |
---|
| 918 | val_list = { } |
---|
[507] | 919 | |
---|
[691] | 920 | for val_name, val_value in jobattrs.items(): |
---|
[507] | 921 | |
---|
[691] | 922 | # These are our own metric names, i.e.: status, start_timestamp, etc |
---|
| 923 | # |
---|
| 924 | val_list_names_len = len( string.join( val_list.keys() ) ) + len(val_list.keys()) |
---|
[507] | 925 | |
---|
[691] | 926 | # These are their corresponding values |
---|
| 927 | # |
---|
| 928 | val_list_vals_len = len( string.join( val_list.values() ) ) + len(val_list.values()) |
---|
[507] | 929 | |
---|
[691] | 930 | if val_name == 'nodes' and jobattrs['status'] == 'R': |
---|
[507] | 931 | |
---|
[691] | 932 | node_str = None |
---|
[507] | 933 | |
---|
[691] | 934 | for node in val_value: |
---|
[507] | 935 | |
---|
[691] | 936 | if node_str: |
---|
[507] | 937 | |
---|
[691] | 938 | node_str = node_str + ';' + node |
---|
| 939 | else: |
---|
| 940 | node_str = node |
---|
[507] | 941 | |
---|
[691] | 942 | # Make sure if we add this new info, that the total metric's value length does not exceed METRIC_MAX_VAL_LEN |
---|
| 943 | # |
---|
| 944 | if (val_list_names_len + len(val_name) ) + (val_list_vals_len + len(node_str) ) > METRIC_MAX_VAL_LEN: |
---|
[507] | 945 | |
---|
[691] | 946 | # It's too big, we need to make a new gmetric for the additional info |
---|
| 947 | # |
---|
| 948 | val_list[ val_name ] = node_str |
---|
[507] | 949 | |
---|
[691] | 950 | gval_lists.append( val_list ) |
---|
[507] | 951 | |
---|
[691] | 952 | val_list = { } |
---|
| 953 | node_str = None |
---|
[507] | 954 | |
---|
[691] | 955 | val_list[ val_name ] = node_str |
---|
[507] | 956 | |
---|
[691] | 957 | gval_lists.append( val_list ) |
---|
[507] | 958 | |
---|
[691] | 959 | val_list = { } |
---|
[507] | 960 | |
---|
[691] | 961 | elif val_value != '': |
---|
[507] | 962 | |
---|
[691] | 963 | # Make sure if we add this new info, that the total metric's value length does not exceed METRIC_MAX_VAL_LEN |
---|
| 964 | # |
---|
| 965 | if (val_list_names_len + len(val_name) ) + (val_list_vals_len + len(str(val_value)) ) > METRIC_MAX_VAL_LEN: |
---|
[507] | 966 | |
---|
[691] | 967 | # It's too big, we need to make a new gmetric for the additional info |
---|
| 968 | # |
---|
| 969 | gval_lists.append( val_list ) |
---|
[507] | 970 | |
---|
[691] | 971 | val_list = { } |
---|
[507] | 972 | |
---|
[691] | 973 | val_list[ val_name ] = val_value |
---|
[507] | 974 | |
---|
[691] | 975 | if len( val_list ) > 0: |
---|
[507] | 976 | |
---|
[691] | 977 | gval_lists.append( val_list ) |
---|
[507] | 978 | |
---|
[691] | 979 | str_list = [ ] |
---|
[507] | 980 | |
---|
[691] | 981 | # Now append the value names and values together, i.e.: stop_timestamp=value, etc |
---|
| 982 | # |
---|
| 983 | for val_list in gval_lists: |
---|
[507] | 984 | |
---|
[691] | 985 | my_val_str = None |
---|
[507] | 986 | |
---|
[691] | 987 | for val_name, val_value in val_list.items(): |
---|
[507] | 988 | |
---|
[691] | 989 | if type(val_value) == list: |
---|
[579] | 990 | |
---|
[691] | 991 | val_value = val_value.join( ',' ) |
---|
[579] | 992 | |
---|
[691] | 993 | if my_val_str: |
---|
[507] | 994 | |
---|
[691] | 995 | try: |
---|
| 996 | # fixme: It's getting |
---|
| 997 | # ('nodes', None) items |
---|
| 998 | my_val_str = my_val_str + ' ' + val_name + '=' + val_value |
---|
| 999 | except: |
---|
| 1000 | pass |
---|
[623] | 1001 | |
---|
[691] | 1002 | else: |
---|
| 1003 | my_val_str = val_name + '=' + val_value |
---|
[507] | 1004 | |
---|
[691] | 1005 | str_list.append( my_val_str ) |
---|
[507] | 1006 | |
---|
[691] | 1007 | return str_list |
---|
[507] | 1008 | |
---|
[691] | 1009 | def daemon( self ): |
---|
[355] | 1010 | |
---|
[691] | 1011 | """Run as daemon forever""" |
---|
[256] | 1012 | |
---|
[691] | 1013 | # Fork the first child |
---|
| 1014 | # |
---|
| 1015 | pid = os.fork() |
---|
| 1016 | if pid > 0: |
---|
| 1017 | sys.exit(0) # end parent |
---|
[256] | 1018 | |
---|
[691] | 1019 | # creates a session and sets the process group ID |
---|
| 1020 | # |
---|
| 1021 | os.setsid() |
---|
[318] | 1022 | |
---|
[691] | 1023 | # Fork the second child |
---|
| 1024 | # |
---|
| 1025 | pid = os.fork() |
---|
| 1026 | if pid > 0: |
---|
| 1027 | sys.exit(0) # end parent |
---|
[318] | 1028 | |
---|
[691] | 1029 | write_pidfile() |
---|
[318] | 1030 | |
---|
[691] | 1031 | # Go to the root directory and set the umask |
---|
| 1032 | # |
---|
| 1033 | os.chdir('/') |
---|
| 1034 | os.umask(0) |
---|
[318] | 1035 | |
---|
[691] | 1036 | sys.stdin.close() |
---|
| 1037 | sys.stdout.close() |
---|
| 1038 | sys.stderr.close() |
---|
[318] | 1039 | |
---|
[691] | 1040 | os.open('/dev/null', os.O_RDWR) |
---|
| 1041 | os.dup2(0, 1) |
---|
| 1042 | os.dup2(0, 2) |
---|
[318] | 1043 | |
---|
[691] | 1044 | self.run() |
---|
[318] | 1045 | |
---|
[691] | 1046 | def run( self ): |
---|
[355] | 1047 | |
---|
[691] | 1048 | """Main thread""" |
---|
[256] | 1049 | |
---|
[691] | 1050 | while ( 1 ): |
---|
| 1051 | |
---|
| 1052 | self.getJobData() |
---|
| 1053 | self.submitJobData() |
---|
| 1054 | time.sleep( BATCH_POLL_INTERVAL ) |
---|
[256] | 1055 | |
---|
[623] | 1056 | # SGE code by Dave Love <fx@gnu.org>. Tested with SGE 6.0u8 and 6.0u11. May |
---|
| 1057 | # work with SGE 6.1 (else should be easily fixable), but definitely doesn't |
---|
| 1058 | # with 6.2. See also the fixmes. |
---|
[256] | 1059 | |
---|
[507] | 1060 | class NoJobs (Exception): |
---|
[691] | 1061 | """Exception raised by empty job list in qstat output.""" |
---|
| 1062 | pass |
---|
[256] | 1063 | |
---|
[507] | 1064 | class SgeQstatXMLParser(xml.sax.handler.ContentHandler): |
---|
[691] | 1065 | """SAX handler for XML output from Sun Grid Engine's `qstat'.""" |
---|
[318] | 1066 | |
---|
[691] | 1067 | def __init__(self): |
---|
| 1068 | self.value = "" |
---|
| 1069 | self.joblist = [] |
---|
| 1070 | self.job = {} |
---|
| 1071 | self.queue = "" |
---|
| 1072 | self.in_joblist = False |
---|
| 1073 | self.lrequest = False |
---|
| 1074 | self.eltq = deque() |
---|
| 1075 | xml.sax.handler.ContentHandler.__init__(self) |
---|
[318] | 1076 | |
---|
[691] | 1077 | # The structure of the output is as follows (for SGE 6.0). It's |
---|
| 1078 | # similar for 6.1, but radically different for 6.2, and is |
---|
| 1079 | # undocumented generally. Unfortunately it's voluminous, and probably |
---|
| 1080 | # doesn't scale to large clusters/queues. |
---|
[318] | 1081 | |
---|
[691] | 1082 | # <detailed_job_info xmlns:xsd="http://www.w3.org/2001/XMLSchema"> |
---|
| 1083 | # <djob_info> |
---|
| 1084 | # <qmaster_response> <!-- job --> |
---|
| 1085 | # ... |
---|
| 1086 | # <JB_ja_template> |
---|
| 1087 | # <ulong_sublist> |
---|
| 1088 | # ... <!-- start_time, state ... --> |
---|
| 1089 | # </ulong_sublist> |
---|
| 1090 | # </JB_ja_template> |
---|
| 1091 | # <JB_ja_tasks> |
---|
| 1092 | # <ulong_sublist> |
---|
| 1093 | # ... <!-- task info |
---|
| 1094 | # </ulong_sublist> |
---|
| 1095 | # ... |
---|
| 1096 | # </JB_ja_tasks> |
---|
| 1097 | # ... |
---|
| 1098 | # </qmaster_response> |
---|
| 1099 | # </djob_info> |
---|
| 1100 | # <messages> |
---|
| 1101 | # ... |
---|
[318] | 1102 | |
---|
[691] | 1103 | # NB. We might treat each task as a separate job, like |
---|
| 1104 | # straight qstat output, but the web interface expects jobs to |
---|
| 1105 | # be identified by integers, not, say, <job number>.<task>. |
---|
[318] | 1106 | |
---|
[691] | 1107 | # So, I lied. If the job list is empty, we get invalid XML |
---|
| 1108 | # like this, which we need to defend against: |
---|
[318] | 1109 | |
---|
[691] | 1110 | # <unknown_jobs xmlns:xsd="http://www.w3.org/2001/XMLSchema"> |
---|
| 1111 | # <> |
---|
| 1112 | # <ST_name>*</ST_name> |
---|
| 1113 | # </> |
---|
| 1114 | # </unknown_jobs> |
---|
[318] | 1115 | |
---|
[691] | 1116 | def startElement(self, name, attrs): |
---|
| 1117 | self.value = "" |
---|
| 1118 | if name == "djob_info": # job list |
---|
| 1119 | self.in_joblist = True |
---|
| 1120 | # The job container is "qmaster_response" in SGE 6.0 |
---|
| 1121 | # and 6.1, but "element" in 6.2. This is only the very |
---|
| 1122 | # start of what's necessary for 6.2, though (sigh). |
---|
| 1123 | elif (name == "qmaster_response" or name == "element") \ |
---|
| 1124 | and self.eltq[-1] == "djob_info": # job |
---|
| 1125 | self.job = {"job_state": "U", "slots": 0, |
---|
| 1126 | "nodes": [], "queued_timestamp": "", |
---|
| 1127 | "queued_timestamp": "", "queue": "", |
---|
| 1128 | "ppn": "0", "RN_max": 0, |
---|
| 1129 | # fixme in endElement |
---|
| 1130 | "requested_memory": 0, "requested_time": 0 |
---|
| 1131 | } |
---|
| 1132 | self.joblist.append(self.job) |
---|
| 1133 | elif name == "qstat_l_requests": # resource request |
---|
| 1134 | self.lrequest = True |
---|
| 1135 | elif name == "unknown_jobs": |
---|
| 1136 | raise NoJobs |
---|
| 1137 | self.eltq.append (name) |
---|
[318] | 1138 | |
---|
[691] | 1139 | def characters(self, ch): |
---|
| 1140 | self.value += ch |
---|
[318] | 1141 | |
---|
[691] | 1142 | def endElement(self, name): |
---|
| 1143 | """Snarf job elements contents into job dictionary. |
---|
| 1144 | Translate keys if appropriate.""" |
---|
[318] | 1145 | |
---|
[691] | 1146 | name_trans = { |
---|
| 1147 | "JB_job_number": "number", |
---|
| 1148 | "JB_job_name": "name", "JB_owner": "owner", |
---|
| 1149 | "queue_name": "queue", "JAT_start_time": "start_timestamp", |
---|
| 1150 | "JB_submission_time": "queued_timestamp" |
---|
| 1151 | } |
---|
| 1152 | value = self.value |
---|
| 1153 | self.eltq.pop () |
---|
[318] | 1154 | |
---|
[691] | 1155 | if name == "djob_info": |
---|
| 1156 | self.in_joblist = False |
---|
| 1157 | self.job = {} |
---|
| 1158 | elif name == "JAT_master_queue": |
---|
| 1159 | self.job["queue"] = value.split("@")[0] |
---|
| 1160 | elif name == "JG_qhostname": |
---|
| 1161 | if not (value in self.job["nodes"]): |
---|
| 1162 | self.job["nodes"].append(value) |
---|
| 1163 | elif name == "JG_slots": # slots in use |
---|
| 1164 | self.job["slots"] += int(value) |
---|
| 1165 | elif name == "RN_max": # requested slots (tasks or parallel) |
---|
| 1166 | self.job["RN_max"] = max (self.job["RN_max"], |
---|
| 1167 | int(value)) |
---|
| 1168 | elif name == "JAT_state": # job state (bitwise or) |
---|
| 1169 | value = int (value) |
---|
| 1170 | # Status values from sge_jobL.h |
---|
| 1171 | #define JIDLE 0x00000000 |
---|
| 1172 | #define JHELD 0x00000010 |
---|
| 1173 | #define JMIGRATING 0x00000020 |
---|
| 1174 | #define JQUEUED 0x00000040 |
---|
| 1175 | #define JRUNNING 0x00000080 |
---|
| 1176 | #define JSUSPENDED 0x00000100 |
---|
| 1177 | #define JTRANSFERING 0x00000200 |
---|
| 1178 | #define JDELETED 0x00000400 |
---|
| 1179 | #define JWAITING 0x00000800 |
---|
| 1180 | #define JEXITING 0x00001000 |
---|
| 1181 | #define JWRITTEN 0x00002000 |
---|
| 1182 | #define JSUSPENDED_ON_THRESHOLD 0x00010000 |
---|
| 1183 | #define JFINISHED 0x00010000 |
---|
| 1184 | if value & 0x80: |
---|
| 1185 | self.job["status"] = "R" |
---|
| 1186 | elif value & 0x40: |
---|
| 1187 | self.job["status"] = "Q" |
---|
| 1188 | else: |
---|
| 1189 | self.job["status"] = "O" # `other' |
---|
| 1190 | elif name == "CE_name" and self.lrequest and self.value in \ |
---|
| 1191 | ("h_cpu", "s_cpu", "cpu", "h_core", "s_core"): |
---|
| 1192 | # We're in a container for an interesting resource |
---|
| 1193 | # request; record which type. |
---|
| 1194 | self.lrequest = self.value |
---|
| 1195 | elif name == "CE_doubleval" and self.lrequest: |
---|
| 1196 | # if we're in a container for an interesting |
---|
| 1197 | # resource request, use the maxmimum of the hard |
---|
| 1198 | # and soft requests to record the requested CPU |
---|
| 1199 | # or core. Fixme: I'm not sure if this logic is |
---|
| 1200 | # right. |
---|
| 1201 | if self.lrequest in ("h_core", "s_core"): |
---|
| 1202 | self.job["requested_memory"] = \ |
---|
| 1203 | max (float (value), |
---|
| 1204 | self.job["requested_memory"]) |
---|
| 1205 | # Fixme: Check what cpu means, c.f [hs]_cpu. |
---|
| 1206 | elif self.lrequest in ("h_cpu", "s_cpu", "cpu"): |
---|
| 1207 | self.job["requested_time"] = \ |
---|
| 1208 | max (float (value), |
---|
| 1209 | self.job["requested_time"]) |
---|
| 1210 | elif name == "qstat_l_requests": |
---|
| 1211 | self.lrequest = False |
---|
| 1212 | elif self.job and self.in_joblist: |
---|
| 1213 | if name in name_trans: |
---|
| 1214 | name = name_trans[name] |
---|
| 1215 | self.job[name] = value |
---|
[318] | 1216 | |
---|
[507] | 1217 | # Abstracted from PBS original. |
---|
| 1218 | # Fixme: Is it worth (or appropriate for PBS) sorting the result? |
---|
[520] | 1219 | # |
---|
| 1220 | def do_nodelist( nodes ): |
---|
| 1221 | |
---|
[691] | 1222 | """Translate node list as appropriate.""" |
---|
[520] | 1223 | |
---|
[691] | 1224 | nodeslist = [ ] |
---|
| 1225 | my_domain = fqdn_parts( socket.getfqdn() )[1] |
---|
[520] | 1226 | |
---|
[691] | 1227 | for node in nodes: |
---|
[520] | 1228 | |
---|
[691] | 1229 | host = node.split( '/' )[0] # not relevant for SGE |
---|
| 1230 | h, host_domain = fqdn_parts(host) |
---|
[520] | 1231 | |
---|
[691] | 1232 | if host_domain == my_domain: |
---|
[520] | 1233 | |
---|
[691] | 1234 | host = h |
---|
[520] | 1235 | |
---|
[691] | 1236 | if nodeslist.count( host ) == 0: |
---|
[520] | 1237 | |
---|
[691] | 1238 | for translate_pattern in BATCH_HOST_TRANSLATE: |
---|
[520] | 1239 | |
---|
[691] | 1240 | if translate_pattern.find( '/' ) != -1: |
---|
[520] | 1241 | |
---|
[691] | 1242 | translate_orig = \ |
---|
| 1243 | translate_pattern.split( '/' )[1] |
---|
| 1244 | translate_new = \ |
---|
| 1245 | translate_pattern.split( '/' )[2] |
---|
| 1246 | host = re.sub( translate_orig, |
---|
| 1247 | translate_new, host ) |
---|
| 1248 | if not host in nodeslist: |
---|
| 1249 | nodeslist.append( host ) |
---|
| 1250 | return nodeslist |
---|
[318] | 1251 | |
---|
[837] | 1252 | class SLURMDataGatherer( DataGatherer ): |
---|
| 1253 | |
---|
| 1254 | global pyslurm |
---|
| 1255 | |
---|
| 1256 | """This is the DataGatherer for SLURM""" |
---|
| 1257 | |
---|
| 1258 | def __init__( self ): |
---|
| 1259 | |
---|
| 1260 | """Setup appropriate variables""" |
---|
| 1261 | |
---|
| 1262 | self.jobs = { } |
---|
| 1263 | self.timeoffset = 0 |
---|
| 1264 | self.dp = DataProcessor() |
---|
| 1265 | |
---|
| 1266 | def getNodeData( self ): |
---|
| 1267 | |
---|
| 1268 | slurm_type = pyslurm.node() |
---|
| 1269 | |
---|
| 1270 | nodedict = slurm_type.get() |
---|
| 1271 | |
---|
| 1272 | return nodedict |
---|
| 1273 | |
---|
| 1274 | def getJobData( self ): |
---|
| 1275 | |
---|
| 1276 | """Gather all data on current jobs""" |
---|
| 1277 | |
---|
| 1278 | joblist = {} |
---|
| 1279 | |
---|
| 1280 | self.cur_time = time.time() |
---|
| 1281 | |
---|
| 1282 | slurm_type = pyslurm.job() |
---|
| 1283 | joblist = slurm_type.get() |
---|
| 1284 | |
---|
| 1285 | jobs_processed = [ ] |
---|
| 1286 | |
---|
| 1287 | for name, attrs in joblist.items(): |
---|
| 1288 | display_queue = 1 |
---|
| 1289 | job_id = name |
---|
| 1290 | |
---|
| 1291 | name = self.getAttr( attrs, 'name' ) |
---|
| 1292 | queue = self.getAttr( attrs, 'partition' ) |
---|
| 1293 | |
---|
| 1294 | if QUEUE: |
---|
| 1295 | for q in QUEUE: |
---|
| 1296 | if q == queue: |
---|
| 1297 | display_queue = 1 |
---|
| 1298 | break |
---|
| 1299 | else: |
---|
| 1300 | display_queue = 0 |
---|
| 1301 | continue |
---|
| 1302 | if display_queue == 0: |
---|
| 1303 | continue |
---|
| 1304 | |
---|
| 1305 | owner_uid = attrs[ 'user_id' ] |
---|
| 1306 | ( owner, owner_pw, owner_uid, owner_gid, owner_gecos, owner_dir, owner_shell ) = pwd.getpwuid( owner_uid ) |
---|
| 1307 | |
---|
| 1308 | requested_time = self.getAttr( attrs, 'time_limit' ) |
---|
[852] | 1309 | min_memory = self.getAttr( attrs, 'pn_min_memory' ) |
---|
[837] | 1310 | |
---|
[852] | 1311 | if min_memory == 0: |
---|
| 1312 | |
---|
| 1313 | requested_memory = '' |
---|
| 1314 | |
---|
| 1315 | else: |
---|
| 1316 | requested_memory = min_memory |
---|
| 1317 | |
---|
[837] | 1318 | ppn = self.getAttr( attrs, 'ntasks_per_node' ) |
---|
| 1319 | |
---|
| 1320 | ( something, status_long ) = self.getAttr( attrs, 'job_state' ) |
---|
| 1321 | |
---|
| 1322 | status = 'Q' |
---|
| 1323 | |
---|
| 1324 | if status_long == 'RUNNING': |
---|
| 1325 | |
---|
| 1326 | status = 'R' |
---|
| 1327 | |
---|
| 1328 | elif status_long == 'COMPLETED': |
---|
| 1329 | |
---|
| 1330 | continue |
---|
| 1331 | |
---|
| 1332 | jobs_processed.append( job_id ) |
---|
| 1333 | |
---|
| 1334 | queued_timestamp = self.getAttr( attrs, 'submit_time' ) |
---|
| 1335 | |
---|
| 1336 | start_timestamp = '' |
---|
| 1337 | nodeslist = '' |
---|
| 1338 | |
---|
| 1339 | if status == 'R': |
---|
| 1340 | |
---|
| 1341 | start_timestamp = self.getAttr( attrs, 'start_time' ) |
---|
[851] | 1342 | nodes = attrs[ 'nodes' ] |
---|
[837] | 1343 | |
---|
[851] | 1344 | if not nodes: |
---|
[837] | 1345 | |
---|
[851] | 1346 | # This should not happen |
---|
| 1347 | |
---|
| 1348 | # Something wrong: running but 'nodes' returned empty by pyslurm |
---|
| 1349 | # Possible pyslurm bug: abort/quit/warning |
---|
| 1350 | |
---|
| 1351 | err_msg = 'FATAL ERROR: job %s running but nodes returned empty: pyslurm bugged?' %job_id |
---|
| 1352 | |
---|
| 1353 | print err_msg |
---|
| 1354 | debug_msg( 0, err_msg ) |
---|
| 1355 | sys.exit(1) |
---|
| 1356 | |
---|
| 1357 | my_nodelist = [ ] |
---|
| 1358 | |
---|
| 1359 | slurm_hostlist = pyslurm.hostlist() |
---|
| 1360 | slurm_hostlist.create( nodes ) |
---|
| 1361 | slurm_hostlist.uniq() |
---|
| 1362 | |
---|
| 1363 | while slurm_hostlist.count() > 0: |
---|
| 1364 | |
---|
| 1365 | my_nodelist.append( slurm_hostlist.pop() ) |
---|
| 1366 | |
---|
| 1367 | slurm_hostlist.destroy() |
---|
| 1368 | |
---|
| 1369 | del slurm_hostlist |
---|
| 1370 | |
---|
| 1371 | nodeslist = do_nodelist( my_nodelist ) |
---|
| 1372 | |
---|
[837] | 1373 | if DETECT_TIME_DIFFS: |
---|
| 1374 | |
---|
| 1375 | # If a job start if later than our current date, |
---|
| 1376 | # that must mean the Torque server's time is later |
---|
| 1377 | # than our local time. |
---|
| 1378 | |
---|
| 1379 | if int( start_timestamp ) > int( int( self.cur_time ) + int( self.timeoffset ) ): |
---|
| 1380 | |
---|
| 1381 | self.timeoffset = int( int(start_timestamp) - int(self.cur_time) ) |
---|
| 1382 | |
---|
| 1383 | elif status == 'Q': |
---|
| 1384 | |
---|
| 1385 | nodeslist = str( attrs[ 'num_nodes' ] ) |
---|
| 1386 | |
---|
| 1387 | else: |
---|
| 1388 | start_timestamp = '' |
---|
| 1389 | nodeslist = '' |
---|
| 1390 | |
---|
| 1391 | myAttrs = { } |
---|
| 1392 | |
---|
| 1393 | myAttrs[ 'name' ] = str( name ) |
---|
| 1394 | myAttrs[ 'queue' ] = str( queue ) |
---|
| 1395 | myAttrs[ 'owner' ] = str( owner ) |
---|
| 1396 | myAttrs[ 'requested_time' ] = str( requested_time ) |
---|
| 1397 | myAttrs[ 'requested_memory' ] = str( requested_memory ) |
---|
| 1398 | myAttrs[ 'ppn' ] = str( ppn ) |
---|
| 1399 | myAttrs[ 'status' ] = str( status ) |
---|
| 1400 | myAttrs[ 'start_timestamp' ] = str( start_timestamp ) |
---|
| 1401 | myAttrs[ 'queued_timestamp' ] = str( queued_timestamp ) |
---|
| 1402 | myAttrs[ 'reported' ] = str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) |
---|
| 1403 | myAttrs[ 'nodes' ] = nodeslist |
---|
| 1404 | myAttrs[ 'domain' ] = fqdn_parts( socket.getfqdn() )[1] |
---|
| 1405 | myAttrs[ 'poll_interval' ] = str( BATCH_POLL_INTERVAL ) |
---|
| 1406 | |
---|
| 1407 | if self.jobDataChanged( self.jobs, job_id, myAttrs ) and myAttrs['status'] in [ 'R', 'Q' ]: |
---|
| 1408 | |
---|
| 1409 | self.jobs[ job_id ] = myAttrs |
---|
| 1410 | |
---|
| 1411 | for id, attrs in self.jobs.items(): |
---|
| 1412 | |
---|
| 1413 | if id not in jobs_processed: |
---|
| 1414 | |
---|
| 1415 | # This one isn't there anymore; toedeledoki! |
---|
| 1416 | # |
---|
| 1417 | del self.jobs[ id ] |
---|
| 1418 | |
---|
[318] | 1419 | class SgeDataGatherer(DataGatherer): |
---|
| 1420 | |
---|
[691] | 1421 | jobs = {} |
---|
[61] | 1422 | |
---|
[691] | 1423 | def __init__( self ): |
---|
| 1424 | self.jobs = {} |
---|
| 1425 | self.timeoffset = 0 |
---|
| 1426 | self.dp = DataProcessor() |
---|
[318] | 1427 | |
---|
[691] | 1428 | def getJobData( self ): |
---|
| 1429 | """Gather all data on current jobs in SGE""" |
---|
[318] | 1430 | |
---|
[691] | 1431 | import popen2 |
---|
[318] | 1432 | |
---|
[691] | 1433 | self.cur_time = 0 |
---|
| 1434 | queues = "" |
---|
| 1435 | if QUEUE: # only for specific queues |
---|
| 1436 | # Fixme: assumes queue names don't contain single |
---|
| 1437 | # quote or comma. Don't know what the SGE rules are. |
---|
| 1438 | queues = " -q '" + string.join (QUEUE, ",") + "'" |
---|
| 1439 | # Note the comment in SgeQstatXMLParser about scaling with |
---|
| 1440 | # this method of getting data. I haven't found better one. |
---|
| 1441 | # Output with args `-xml -ext -f -r' is easier to parse |
---|
| 1442 | # in some ways, harder in others, but it doesn't provide |
---|
| 1443 | # the submission time (at least SGE 6.0). The pipeline |
---|
| 1444 | # into sed corrects bogus XML observed with a configuration |
---|
| 1445 | # of SGE 6.0u8, which otherwise causes the parsing to hang. |
---|
| 1446 | piping = popen2.Popen3("qstat -u '*' -j '*' -xml | \ |
---|
[623] | 1447 | sed -e 's/reported usage>/reported_usage>/g' -e 's;<\/*JATASK:.*>;;'" \ |
---|
[691] | 1448 | + queues, True) |
---|
| 1449 | qstatparser = SgeQstatXMLParser() |
---|
| 1450 | parse_err = 0 |
---|
| 1451 | try: |
---|
| 1452 | xml.sax.parse(piping.fromchild, qstatparser) |
---|
| 1453 | except NoJobs: |
---|
| 1454 | pass |
---|
| 1455 | except: |
---|
| 1456 | parse_err = 1 |
---|
[704] | 1457 | if piping.wait(): |
---|
| 1458 | debug_msg(10, "qstat error, skipping until next polling interval: " + piping.childerr.readline()) |
---|
[691] | 1459 | return None |
---|
| 1460 | elif parse_err: |
---|
| 1461 | debug_msg(10, "Bad XML output from qstat"()) |
---|
| 1462 | exit (1) |
---|
| 1463 | for f in piping.fromchild, piping.tochild, piping.childerr: |
---|
| 1464 | f.close() |
---|
| 1465 | self.cur_time = time.time() |
---|
| 1466 | jobs_processed = [] |
---|
| 1467 | for job in qstatparser.joblist: |
---|
| 1468 | job_id = job["number"] |
---|
| 1469 | if job["status"] in [ 'Q', 'R' ]: |
---|
| 1470 | jobs_processed.append(job_id) |
---|
| 1471 | if job["status"] == "R": |
---|
| 1472 | job["nodes"] = do_nodelist (job["nodes"]) |
---|
| 1473 | # Fixme: why is job["nodes"] sometimes null? |
---|
| 1474 | try: |
---|
| 1475 | # Fixme: Is this sensible? The |
---|
| 1476 | # PBS-type PPN isn't something you use |
---|
| 1477 | # with SGE. |
---|
[704] | 1478 | job["ppn"] = float(job["slots"]) / len(job["nodes"]) |
---|
[691] | 1479 | except: |
---|
| 1480 | job["ppn"] = 0 |
---|
| 1481 | if DETECT_TIME_DIFFS: |
---|
| 1482 | # If a job start is later than our |
---|
| 1483 | # current date, that must mean |
---|
| 1484 | # the SGE server's time is later |
---|
| 1485 | # than our local time. |
---|
[704] | 1486 | start_timestamp = int (job["start_timestamp"]) |
---|
| 1487 | if start_timestamp > int(self.cur_time) + int(self.timeoffset): |
---|
[318] | 1488 | |
---|
[704] | 1489 | self.timeoffset = start_timestamp - int(self.cur_time) |
---|
[691] | 1490 | else: |
---|
| 1491 | # fixme: Note sure what this should be: |
---|
| 1492 | job["ppn"] = job["RN_max"] |
---|
| 1493 | job["nodes"] = "1" |
---|
[318] | 1494 | |
---|
[691] | 1495 | myAttrs = {} |
---|
| 1496 | for attr in ["name", "queue", "owner", |
---|
| 1497 | "requested_time", "status", |
---|
| 1498 | "requested_memory", "ppn", |
---|
| 1499 | "start_timestamp", "queued_timestamp"]: |
---|
| 1500 | myAttrs[attr] = str(job[attr]) |
---|
| 1501 | myAttrs["nodes"] = job["nodes"] |
---|
[704] | 1502 | myAttrs["reported"] = str(int(self.cur_time) + int(self.timeoffset)) |
---|
[691] | 1503 | myAttrs["domain"] = fqdn_parts(socket.getfqdn())[1] |
---|
| 1504 | myAttrs["poll_interval"] = str(BATCH_POLL_INTERVAL) |
---|
[318] | 1505 | |
---|
[704] | 1506 | if self.jobDataChanged(self.jobs, job_id, myAttrs) and myAttrs["status"] in ["R", "Q"]: |
---|
[691] | 1507 | self.jobs[job_id] = myAttrs |
---|
| 1508 | for id, attrs in self.jobs.items(): |
---|
| 1509 | if id not in jobs_processed: |
---|
| 1510 | del self.jobs[id] |
---|
[318] | 1511 | |
---|
[524] | 1512 | # LSF code by Mahmoud Hanafi <hanafim@users.sourceforge.nt> |
---|
| 1513 | # Requres LSFObject http://sourceforge.net/projects/lsfobject |
---|
| 1514 | # |
---|
| 1515 | class LsfDataGatherer(DataGatherer): |
---|
[525] | 1516 | |
---|
[691] | 1517 | """This is the DataGatherer for LSf""" |
---|
[524] | 1518 | |
---|
[691] | 1519 | global lsfObject |
---|
[524] | 1520 | |
---|
[691] | 1521 | def __init__( self ): |
---|
[525] | 1522 | |
---|
[691] | 1523 | self.jobs = { } |
---|
| 1524 | self.timeoffset = 0 |
---|
| 1525 | self.dp = DataProcessor() |
---|
| 1526 | self.initLsfQuery() |
---|
[524] | 1527 | |
---|
[691] | 1528 | def _countDuplicatesInList( self, dupedList ): |
---|
[525] | 1529 | |
---|
[691] | 1530 | countDupes = { } |
---|
[525] | 1531 | |
---|
[691] | 1532 | for item in dupedList: |
---|
[525] | 1533 | |
---|
[691] | 1534 | if not countDupes.has_key( item ): |
---|
[525] | 1535 | |
---|
[691] | 1536 | countDupes[ item ] = 1 |
---|
| 1537 | else: |
---|
| 1538 | countDupes[ item ] = countDupes[ item ] + 1 |
---|
[525] | 1539 | |
---|
[691] | 1540 | dupeCountList = [ ] |
---|
[525] | 1541 | |
---|
[691] | 1542 | for item, count in countDupes.items(): |
---|
[525] | 1543 | |
---|
[691] | 1544 | dupeCountList.append( ( item, count ) ) |
---|
[525] | 1545 | |
---|
[691] | 1546 | return dupeCountList |
---|
[524] | 1547 | # |
---|
| 1548 | #lst = ['I1','I2','I1','I3','I4','I4','I7','I7','I7','I7','I7'] |
---|
| 1549 | #print _countDuplicatesInList(lst) |
---|
| 1550 | #[('I1', 2), ('I3', 1), ('I2', 1), ('I4', 2), ('I7', 5)] |
---|
| 1551 | ######################## |
---|
| 1552 | |
---|
[691] | 1553 | def initLsfQuery( self ): |
---|
| 1554 | self.pq = None |
---|
| 1555 | self.pq = lsfObject.jobInfoEntObject() |
---|
[524] | 1556 | |
---|
[691] | 1557 | def getJobData( self, known_jobs="" ): |
---|
| 1558 | """Gather all data on current jobs in LSF""" |
---|
| 1559 | if len( known_jobs ) > 0: |
---|
| 1560 | jobs = known_jobs |
---|
| 1561 | else: |
---|
| 1562 | jobs = { } |
---|
| 1563 | joblist = {} |
---|
| 1564 | joblist = self.pq.getJobInfo() |
---|
| 1565 | nodelist = '' |
---|
[524] | 1566 | |
---|
[691] | 1567 | self.cur_time = time.time() |
---|
[524] | 1568 | |
---|
[691] | 1569 | jobs_processed = [ ] |
---|
[524] | 1570 | |
---|
[691] | 1571 | for name, attrs in joblist.items(): |
---|
| 1572 | job_id = str(name) |
---|
| 1573 | jobs_processed.append( job_id ) |
---|
| 1574 | name = self.getAttr( attrs, 'jobName' ) |
---|
| 1575 | queue = self.getAttr( self.getAttr( attrs, 'submit') , 'queue' ) |
---|
| 1576 | owner = self.getAttr( attrs, 'user' ) |
---|
[524] | 1577 | |
---|
| 1578 | ### THIS IS THE rLimit List index values |
---|
[691] | 1579 | #define LSF_RLIMIT_CPU 0 /* cpu time in milliseconds */ |
---|
| 1580 | #define LSF_RLIMIT_FSIZE 1 /* maximum file size */ |
---|
| 1581 | #define LSF_RLIMIT_DATA 2 /* data size */ |
---|
| 1582 | #define LSF_RLIMIT_STACK 3 /* stack size */ |
---|
| 1583 | #define LSF_RLIMIT_CORE 4 /* core file size */ |
---|
| 1584 | #define LSF_RLIMIT_RSS 5 /* resident set size */ |
---|
| 1585 | #define LSF_RLIMIT_NOFILE 6 /* open files */ |
---|
| 1586 | #define LSF_RLIMIT_OPEN_MAX 7 /* (from HP-UX) */ |
---|
| 1587 | #define LSF_RLIMIT_VMEM 8 /* maximum swap mem */ |
---|
[524] | 1588 | #define LSF_RLIMIT_SWAP 8 |
---|
[691] | 1589 | #define LSF_RLIMIT_RUN 9 /* max wall-clock time limit */ |
---|
| 1590 | #define LSF_RLIMIT_PROCESS 10 /* process number limit */ |
---|
| 1591 | #define LSF_RLIMIT_THREAD 11 /* thread number limit (introduced in LSF6.0) */ |
---|
| 1592 | #define LSF_RLIM_NLIMITS 12 /* number of resource limits */ |
---|
[524] | 1593 | |
---|
[691] | 1594 | requested_time = self.getAttr( self.getAttr( attrs, 'submit') , 'rLimits' )[9] |
---|
| 1595 | if requested_time == -1: |
---|
| 1596 | requested_time = "" |
---|
| 1597 | requested_memory = self.getAttr( self.getAttr( attrs, 'submit') , 'rLimits' )[8] |
---|
| 1598 | if requested_memory == -1: |
---|
| 1599 | requested_memory = "" |
---|
[524] | 1600 | # This tries to get proc per node. We don't support this right now |
---|
[691] | 1601 | ppn = 0 #self.getAttr( self.getAttr( attrs, 'SubmitList') , 'numProessors' ) |
---|
| 1602 | requested_cpus = self.getAttr( self.getAttr( attrs, 'submit') , 'numProcessors' ) |
---|
| 1603 | if requested_cpus == None or requested_cpus == "": |
---|
| 1604 | requested_cpus = 1 |
---|
[524] | 1605 | |
---|
[691] | 1606 | if QUEUE: |
---|
| 1607 | for q in QUEUE: |
---|
| 1608 | if q == queue: |
---|
| 1609 | display_queue = 1 |
---|
| 1610 | break |
---|
| 1611 | else: |
---|
| 1612 | display_queue = 0 |
---|
| 1613 | continue |
---|
| 1614 | if display_queue == 0: |
---|
| 1615 | continue |
---|
[524] | 1616 | |
---|
[691] | 1617 | runState = self.getAttr( attrs, 'status' ) |
---|
| 1618 | if runState == 4: |
---|
| 1619 | status = 'R' |
---|
| 1620 | else: |
---|
| 1621 | status = 'Q' |
---|
| 1622 | queued_timestamp = self.getAttr( attrs, 'submitTime' ) |
---|
[524] | 1623 | |
---|
[691] | 1624 | if status == 'R': |
---|
| 1625 | start_timestamp = self.getAttr( attrs, 'startTime' ) |
---|
| 1626 | nodesCpu = dict(self._countDuplicatesInList(self.getAttr( attrs, 'exHosts' ))) |
---|
| 1627 | nodelist = nodesCpu.keys() |
---|
[524] | 1628 | |
---|
[691] | 1629 | if DETECT_TIME_DIFFS: |
---|
[524] | 1630 | |
---|
[691] | 1631 | # If a job start if later than our current date, |
---|
| 1632 | # that must mean the Torque server's time is later |
---|
| 1633 | # than our local time. |
---|
[524] | 1634 | |
---|
[691] | 1635 | if int(start_timestamp) > int( int(self.cur_time) + int(self.timeoffset) ): |
---|
[524] | 1636 | |
---|
[691] | 1637 | self.timeoffset = int( int(start_timestamp) - int(self.cur_time) ) |
---|
[524] | 1638 | |
---|
[691] | 1639 | elif status == 'Q': |
---|
| 1640 | start_timestamp = '' |
---|
| 1641 | count_mynodes = 0 |
---|
| 1642 | numeric_node = 1 |
---|
| 1643 | nodelist = '' |
---|
[524] | 1644 | |
---|
[691] | 1645 | myAttrs = { } |
---|
| 1646 | if name == "": |
---|
| 1647 | myAttrs['name'] = "none" |
---|
| 1648 | else: |
---|
| 1649 | myAttrs['name'] = name |
---|
[524] | 1650 | |
---|
[691] | 1651 | myAttrs[ 'owner' ] = owner |
---|
| 1652 | myAttrs[ 'requested_time' ] = str(requested_time) |
---|
| 1653 | myAttrs[ 'requested_memory' ] = str(requested_memory) |
---|
| 1654 | myAttrs[ 'requested_cpus' ] = str(requested_cpus) |
---|
| 1655 | myAttrs[ 'ppn' ] = str( ppn ) |
---|
| 1656 | myAttrs[ 'status' ] = status |
---|
| 1657 | myAttrs[ 'start_timestamp' ] = str(start_timestamp) |
---|
| 1658 | myAttrs[ 'queue' ] = str(queue) |
---|
| 1659 | myAttrs[ 'queued_timestamp' ] = str(queued_timestamp) |
---|
| 1660 | myAttrs[ 'reported' ] = str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) |
---|
| 1661 | myAttrs[ 'nodes' ] = do_nodelist( nodelist ) |
---|
| 1662 | myAttrs[ 'domain' ] = fqdn_parts( socket.getfqdn() )[1] |
---|
| 1663 | myAttrs[ 'poll_interval' ] = str(BATCH_POLL_INTERVAL) |
---|
[524] | 1664 | |
---|
[691] | 1665 | if self.jobDataChanged( jobs, job_id, myAttrs ) and myAttrs['status'] in [ 'R', 'Q' ]: |
---|
| 1666 | jobs[ job_id ] = myAttrs |
---|
[524] | 1667 | |
---|
[691] | 1668 | debug_msg( 10, printTime() + ' job %s state changed' %(job_id) ) |
---|
[524] | 1669 | |
---|
[691] | 1670 | for id, attrs in jobs.items(): |
---|
| 1671 | if id not in jobs_processed: |
---|
| 1672 | # This one isn't there anymore |
---|
| 1673 | # |
---|
| 1674 | del jobs[ id ] |
---|
| 1675 | self.jobs=jobs |
---|
[524] | 1676 | |
---|
| 1677 | |
---|
[355] | 1678 | class PbsDataGatherer( DataGatherer ): |
---|
[318] | 1679 | |
---|
[691] | 1680 | """This is the DataGatherer for PBS and Torque""" |
---|
[318] | 1681 | |
---|
[691] | 1682 | global PBSQuery, PBSError |
---|
[256] | 1683 | |
---|
[691] | 1684 | def __init__( self ): |
---|
[354] | 1685 | |
---|
[691] | 1686 | """Setup appropriate variables""" |
---|
[23] | 1687 | |
---|
[785] | 1688 | self.jobs = { } |
---|
| 1689 | self.timeoffset = 0 |
---|
| 1690 | self.dp = DataProcessor() |
---|
[354] | 1691 | |
---|
[691] | 1692 | self.initPbsQuery() |
---|
[23] | 1693 | |
---|
[691] | 1694 | def initPbsQuery( self ): |
---|
[91] | 1695 | |
---|
[785] | 1696 | self.pq = None |
---|
[354] | 1697 | |
---|
[788] | 1698 | try: |
---|
[354] | 1699 | |
---|
[788] | 1700 | if( BATCH_SERVER ): |
---|
[91] | 1701 | |
---|
[788] | 1702 | self.pq = PBSQuery( BATCH_SERVER ) |
---|
| 1703 | else: |
---|
| 1704 | self.pq = PBSQuery() |
---|
| 1705 | |
---|
| 1706 | except PBSError, details: |
---|
| 1707 | print 'Cannot connect to pbs server' |
---|
| 1708 | print details |
---|
| 1709 | sys.exit( 1 ) |
---|
| 1710 | |
---|
[691] | 1711 | try: |
---|
[791] | 1712 | # TODO: actually use new data structure |
---|
[691] | 1713 | self.pq.old_data_structure() |
---|
[656] | 1714 | |
---|
[691] | 1715 | except AttributeError: |
---|
[656] | 1716 | |
---|
[691] | 1717 | # pbs_query is older |
---|
| 1718 | # |
---|
| 1719 | pass |
---|
[656] | 1720 | |
---|
[790] | 1721 | def getNodeData( self ): |
---|
| 1722 | |
---|
| 1723 | nodedict = { } |
---|
| 1724 | |
---|
| 1725 | try: |
---|
| 1726 | nodedict = self.pq.getnodes() |
---|
| 1727 | |
---|
| 1728 | except PBSError, detail: |
---|
| 1729 | |
---|
| 1730 | debug_msg( 10, "PBS server unavailable, skipping until next polling interval: " + str( detail ) ) |
---|
| 1731 | |
---|
| 1732 | return nodedict |
---|
| 1733 | |
---|
[691] | 1734 | def getJobData( self ): |
---|
[354] | 1735 | |
---|
[691] | 1736 | """Gather all data on current jobs in Torque""" |
---|
[26] | 1737 | |
---|
[785] | 1738 | joblist = {} |
---|
| 1739 | self.cur_time = 0 |
---|
[349] | 1740 | |
---|
[691] | 1741 | try: |
---|
| 1742 | joblist = self.pq.getjobs() |
---|
[785] | 1743 | self.cur_time = time.time() |
---|
[354] | 1744 | |
---|
[691] | 1745 | except PBSError, detail: |
---|
[354] | 1746 | |
---|
[790] | 1747 | debug_msg( 10, "PBS server unavailable, skipping until next polling interval: " + str( detail ) ) |
---|
[691] | 1748 | return None |
---|
[354] | 1749 | |
---|
[691] | 1750 | jobs_processed = [ ] |
---|
[26] | 1751 | |
---|
[691] | 1752 | for name, attrs in joblist.items(): |
---|
[785] | 1753 | display_queue = 1 |
---|
| 1754 | job_id = name.split( '.' )[0] |
---|
[26] | 1755 | |
---|
[785] | 1756 | name = self.getAttr( attrs, 'Job_Name' ) |
---|
| 1757 | queue = self.getAttr( attrs, 'queue' ) |
---|
[317] | 1758 | |
---|
[691] | 1759 | if QUEUE: |
---|
| 1760 | for q in QUEUE: |
---|
| 1761 | if q == queue: |
---|
| 1762 | display_queue = 1 |
---|
| 1763 | break |
---|
| 1764 | else: |
---|
| 1765 | display_queue = 0 |
---|
| 1766 | continue |
---|
| 1767 | if display_queue == 0: |
---|
| 1768 | continue |
---|
[317] | 1769 | |
---|
| 1770 | |
---|
[691] | 1771 | owner = self.getAttr( attrs, 'Job_Owner' ).split( '@' )[0] |
---|
[785] | 1772 | requested_time = self.getAttr( attrs, 'Resource_List.walltime' ) |
---|
| 1773 | requested_memory = self.getAttr( attrs, 'Resource_List.mem' ) |
---|
[95] | 1774 | |
---|
[785] | 1775 | mynoderequest = self.getAttr( attrs, 'Resource_List.nodes' ) |
---|
[95] | 1776 | |
---|
[785] | 1777 | ppn = '' |
---|
[281] | 1778 | |
---|
[691] | 1779 | if mynoderequest.find( ':' ) != -1 and mynoderequest.find( 'ppn' ) != -1: |
---|
[95] | 1780 | |
---|
[785] | 1781 | mynoderequest_fields = mynoderequest.split( ':' ) |
---|
[281] | 1782 | |
---|
[691] | 1783 | for mynoderequest_field in mynoderequest_fields: |
---|
[281] | 1784 | |
---|
[691] | 1785 | if mynoderequest_field.find( 'ppn' ) != -1: |
---|
[281] | 1786 | |
---|
[785] | 1787 | ppn = mynoderequest_field.split( 'ppn=' )[1] |
---|
[281] | 1788 | |
---|
[785] | 1789 | status = self.getAttr( attrs, 'job_state' ) |
---|
[25] | 1790 | |
---|
[691] | 1791 | if status in [ 'Q', 'R' ]: |
---|
[450] | 1792 | |
---|
[691] | 1793 | jobs_processed.append( job_id ) |
---|
[450] | 1794 | |
---|
[785] | 1795 | queued_timestamp = self.getAttr( attrs, 'ctime' ) |
---|
[243] | 1796 | |
---|
[691] | 1797 | if status == 'R': |
---|
[133] | 1798 | |
---|
[785] | 1799 | start_timestamp = self.getAttr( attrs, 'mtime' ) |
---|
| 1800 | nodes = self.getAttr( attrs, 'exec_host' ).split( '+' ) |
---|
[133] | 1801 | |
---|
[785] | 1802 | nodeslist = do_nodelist( nodes ) |
---|
[354] | 1803 | |
---|
[691] | 1804 | if DETECT_TIME_DIFFS: |
---|
[185] | 1805 | |
---|
[691] | 1806 | # If a job start if later than our current date, |
---|
| 1807 | # that must mean the Torque server's time is later |
---|
| 1808 | # than our local time. |
---|
| 1809 | |
---|
| 1810 | if int( start_timestamp ) > int( int( self.cur_time ) + int( self.timeoffset ) ): |
---|
[185] | 1811 | |
---|
[785] | 1812 | self.timeoffset = int( int(start_timestamp) - int(self.cur_time) ) |
---|
[185] | 1813 | |
---|
[691] | 1814 | elif status == 'Q': |
---|
[95] | 1815 | |
---|
[691] | 1816 | # 'mynodequest' can be a string in the following syntax according to the |
---|
| 1817 | # Torque Administator's manual: |
---|
| 1818 | # |
---|
| 1819 | # {<node_count> | <hostname>}[:ppn=<ppn>][:<property>[:<property>]...][+ ...] |
---|
| 1820 | # {<node_count> | <hostname>}[:ppn=<ppn>][:<property>[:<property>]...][+ ...] |
---|
| 1821 | # etc |
---|
| 1822 | # |
---|
[451] | 1823 | |
---|
[691] | 1824 | # |
---|
| 1825 | # For now we only count the amount of nodes request and ignore properties |
---|
| 1826 | # |
---|
[451] | 1827 | |
---|
[785] | 1828 | start_timestamp = '' |
---|
| 1829 | count_mynodes = 0 |
---|
[354] | 1830 | |
---|
[691] | 1831 | for node in mynoderequest.split( '+' ): |
---|
[67] | 1832 | |
---|
[691] | 1833 | # Just grab the {node_count|hostname} part and ignore properties |
---|
| 1834 | # |
---|
[785] | 1835 | nodepart = node.split( ':' )[0] |
---|
[67] | 1836 | |
---|
[691] | 1837 | # Let's assume a node_count value |
---|
| 1838 | # |
---|
[785] | 1839 | numeric_node = 1 |
---|
[451] | 1840 | |
---|
[691] | 1841 | # Chop the value up into characters |
---|
| 1842 | # |
---|
| 1843 | for letter in nodepart: |
---|
[67] | 1844 | |
---|
[691] | 1845 | # If this char is not a digit (0-9), this must be a hostname |
---|
| 1846 | # |
---|
| 1847 | if letter not in string.digits: |
---|
[133] | 1848 | |
---|
[785] | 1849 | numeric_node = 0 |
---|
[133] | 1850 | |
---|
[691] | 1851 | # If this is a hostname, just count this as one (1) node |
---|
| 1852 | # |
---|
| 1853 | if not numeric_node: |
---|
[354] | 1854 | |
---|
[785] | 1855 | count_mynodes = count_mynodes + 1 |
---|
[691] | 1856 | else: |
---|
[451] | 1857 | |
---|
[691] | 1858 | # If this a number, it must be the node_count |
---|
| 1859 | # and increase our count with it's value |
---|
| 1860 | # |
---|
| 1861 | try: |
---|
[785] | 1862 | count_mynodes = count_mynodes + int( nodepart ) |
---|
[354] | 1863 | |
---|
[691] | 1864 | except ValueError, detail: |
---|
[354] | 1865 | |
---|
[691] | 1866 | # When we arrive here I must be bugged or very confused |
---|
| 1867 | # THIS SHOULD NOT HAPPEN! |
---|
| 1868 | # |
---|
| 1869 | debug_msg( 10, str( detail ) ) |
---|
| 1870 | debug_msg( 10, "Encountered weird node in Resources_List?!" ) |
---|
| 1871 | debug_msg( 10, 'nodepart = ' + str( nodepart ) ) |
---|
| 1872 | debug_msg( 10, 'job = ' + str( name ) ) |
---|
| 1873 | debug_msg( 10, 'attrs = ' + str( attrs ) ) |
---|
| 1874 | |
---|
[785] | 1875 | nodeslist = str( count_mynodes ) |
---|
[691] | 1876 | else: |
---|
[785] | 1877 | start_timestamp = '' |
---|
| 1878 | nodeslist = '' |
---|
[133] | 1879 | |
---|
[691] | 1880 | myAttrs = { } |
---|
[26] | 1881 | |
---|
[785] | 1882 | myAttrs[ 'name' ] = str( name ) |
---|
| 1883 | myAttrs[ 'queue' ] = str( queue ) |
---|
| 1884 | myAttrs[ 'owner' ] = str( owner ) |
---|
| 1885 | myAttrs[ 'requested_time' ] = str( requested_time ) |
---|
| 1886 | myAttrs[ 'requested_memory' ] = str( requested_memory ) |
---|
| 1887 | myAttrs[ 'ppn' ] = str( ppn ) |
---|
| 1888 | myAttrs[ 'status' ] = str( status ) |
---|
| 1889 | myAttrs[ 'start_timestamp' ] = str( start_timestamp ) |
---|
| 1890 | myAttrs[ 'queued_timestamp' ] = str( queued_timestamp ) |
---|
| 1891 | myAttrs[ 'reported' ] = str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) |
---|
| 1892 | myAttrs[ 'nodes' ] = nodeslist |
---|
| 1893 | myAttrs[ 'domain' ] = fqdn_parts( socket.getfqdn() )[1] |
---|
[691] | 1894 | myAttrs[ 'poll_interval' ] = str( BATCH_POLL_INTERVAL ) |
---|
[354] | 1895 | |
---|
[691] | 1896 | if self.jobDataChanged( self.jobs, job_id, myAttrs ) and myAttrs['status'] in [ 'R', 'Q' ]: |
---|
[61] | 1897 | |
---|
[785] | 1898 | self.jobs[ job_id ] = myAttrs |
---|
[26] | 1899 | |
---|
[691] | 1900 | for id, attrs in self.jobs.items(): |
---|
[76] | 1901 | |
---|
[691] | 1902 | if id not in jobs_processed: |
---|
[76] | 1903 | |
---|
[691] | 1904 | # This one isn't there anymore; toedeledoki! |
---|
| 1905 | # |
---|
| 1906 | del self.jobs[ id ] |
---|
[76] | 1907 | |
---|
[362] | 1908 | GMETRIC_DEFAULT_TYPE = 'string' |
---|
| 1909 | GMETRIC_DEFAULT_HOST = '127.0.0.1' |
---|
| 1910 | GMETRIC_DEFAULT_PORT = '8649' |
---|
[700] | 1911 | GMETRIC_DEFAULT_UNITS = '' |
---|
[362] | 1912 | |
---|
| 1913 | class Gmetric: |
---|
| 1914 | |
---|
[691] | 1915 | global GMETRIC_DEFAULT_HOST, GMETRIC_DEFAULT_PORT |
---|
[362] | 1916 | |
---|
[700] | 1917 | slope = { 'zero' : 0, 'positive' : 1, 'negative' : 2, 'both' : 3, 'unspecified' : 4 } |
---|
| 1918 | type = ( '', 'string', 'uint16', 'int16', 'uint32', 'int32', 'float', 'double', 'timestamp' ) |
---|
| 1919 | protocol = ( 'udp', 'multicast' ) |
---|
[362] | 1920 | |
---|
[691] | 1921 | def __init__( self, host=GMETRIC_DEFAULT_HOST, port=GMETRIC_DEFAULT_PORT ): |
---|
[700] | 1922 | |
---|
[691] | 1923 | global GMETRIC_DEFAULT_TYPE |
---|
[362] | 1924 | |
---|
[691] | 1925 | self.prot = self.checkHostProtocol( host ) |
---|
[700] | 1926 | self.data_msg = xdrlib.Packer() |
---|
| 1927 | self.meta_msg = xdrlib.Packer() |
---|
[691] | 1928 | self.socket = socket.socket( socket.AF_INET, socket.SOCK_DGRAM ) |
---|
[362] | 1929 | |
---|
[691] | 1930 | if self.prot not in self.protocol: |
---|
[362] | 1931 | |
---|
[691] | 1932 | raise ValueError( "Protocol must be one of: " + str( self.protocol ) ) |
---|
[362] | 1933 | |
---|
[691] | 1934 | if self.prot == 'multicast': |
---|
[362] | 1935 | |
---|
[691] | 1936 | # Set multicast options |
---|
| 1937 | # |
---|
| 1938 | self.socket.setsockopt( socket.IPPROTO_IP, socket.IP_MULTICAST_TTL, 20 ) |
---|
[362] | 1939 | |
---|
[691] | 1940 | self.hostport = ( host, int( port ) ) |
---|
| 1941 | self.slopestr = 'both' |
---|
| 1942 | self.tmax = 60 |
---|
[362] | 1943 | |
---|
[691] | 1944 | def checkHostProtocol( self, ip ): |
---|
[362] | 1945 | |
---|
[691] | 1946 | """Detect if a ip adress is a multicast address""" |
---|
[471] | 1947 | |
---|
[691] | 1948 | MULTICAST_ADDRESS_MIN = ( "224", "0", "0", "0" ) |
---|
| 1949 | MULTICAST_ADDRESS_MAX = ( "239", "255", "255", "255" ) |
---|
[362] | 1950 | |
---|
[700] | 1951 | ip_fields = ip.split( '.' ) |
---|
[362] | 1952 | |
---|
[691] | 1953 | if ip_fields >= MULTICAST_ADDRESS_MIN and ip_fields <= MULTICAST_ADDRESS_MAX: |
---|
[362] | 1954 | |
---|
[691] | 1955 | return 'multicast' |
---|
| 1956 | else: |
---|
| 1957 | return 'udp' |
---|
[362] | 1958 | |
---|
[691] | 1959 | def send( self, name, value, dmax, typestr = '', units = '' ): |
---|
[362] | 1960 | |
---|
[691] | 1961 | if len( units ) == 0: |
---|
[700] | 1962 | units = GMETRIC_DEFAULT_UNITS |
---|
[471] | 1963 | |
---|
[691] | 1964 | if len( typestr ) == 0: |
---|
[700] | 1965 | typestr = GMETRIC_DEFAULT_TYPE |
---|
[362] | 1966 | |
---|
[700] | 1967 | (meta_msg, data_msg) = self.makexdr( name, value, typestr, units, self.slopestr, self.tmax, dmax ) |
---|
[409] | 1968 | |
---|
[700] | 1969 | meta_rt = self.socket.sendto( meta_msg, self.hostport ) |
---|
| 1970 | data_rt = self.socket.sendto( data_msg, self.hostport ) |
---|
[362] | 1971 | |
---|
[700] | 1972 | return ( meta_rt, data_rt ) |
---|
[362] | 1973 | |
---|
[700] | 1974 | def makexdr( self, name, value, typestr, unitstr, slopestr, tmax, dmax, group=None, spoof=None ): |
---|
| 1975 | |
---|
| 1976 | hostname = "unset" |
---|
| 1977 | |
---|
[691] | 1978 | if slopestr not in self.slope: |
---|
[362] | 1979 | |
---|
[691] | 1980 | raise ValueError( "Slope must be one of: " + str( self.slope.keys() ) ) |
---|
[362] | 1981 | |
---|
[691] | 1982 | if typestr not in self.type: |
---|
[362] | 1983 | |
---|
[691] | 1984 | raise ValueError( "Type must be one of: " + str( self.type ) ) |
---|
[362] | 1985 | |
---|
[691] | 1986 | if len( name ) == 0: |
---|
[362] | 1987 | |
---|
[691] | 1988 | raise ValueError( "Name must be non-empty" ) |
---|
[362] | 1989 | |
---|
[700] | 1990 | self.meta_msg.reset() |
---|
| 1991 | self.meta_msg.pack_int( 128 ) |
---|
[362] | 1992 | |
---|
[700] | 1993 | if not spoof: |
---|
| 1994 | self.meta_msg.pack_string( hostname ) |
---|
| 1995 | else: |
---|
| 1996 | self.meta_msg.pack_string( spoof ) |
---|
[362] | 1997 | |
---|
[700] | 1998 | self.meta_msg.pack_string( name ) |
---|
| 1999 | |
---|
| 2000 | if not spoof: |
---|
| 2001 | self.meta_msg.pack_int( 0 ) |
---|
| 2002 | else: |
---|
| 2003 | self.meta_msg.pack_int( 1 ) |
---|
| 2004 | |
---|
| 2005 | self.meta_msg.pack_string( typestr ) |
---|
| 2006 | self.meta_msg.pack_string( name ) |
---|
| 2007 | self.meta_msg.pack_string( unitstr ) |
---|
| 2008 | self.meta_msg.pack_int( self.slope[ slopestr ] ) |
---|
| 2009 | self.meta_msg.pack_uint( int( tmax ) ) |
---|
| 2010 | self.meta_msg.pack_uint( int( dmax ) ) |
---|
| 2011 | |
---|
| 2012 | if not group: |
---|
| 2013 | self.meta_msg.pack_int( 0 ) |
---|
| 2014 | else: |
---|
| 2015 | self.meta_msg.pack_int( 1 ) |
---|
| 2016 | self.meta_msg.pack_string( "GROUP" ) |
---|
| 2017 | self.meta_msg.pack_string( group ) |
---|
| 2018 | |
---|
| 2019 | self.data_msg.reset() |
---|
| 2020 | self.data_msg.pack_int( 128+5 ) |
---|
| 2021 | |
---|
| 2022 | if not spoof: |
---|
| 2023 | self.data_msg.pack_string( hostname ) |
---|
| 2024 | else: |
---|
| 2025 | self.data_msg.pack_string( spoof ) |
---|
| 2026 | |
---|
| 2027 | self.data_msg.pack_string( name ) |
---|
| 2028 | |
---|
| 2029 | if not spoof: |
---|
| 2030 | self.data_msg.pack_int( 0 ) |
---|
| 2031 | else: |
---|
| 2032 | self.data_msg.pack_int( 1 ) |
---|
| 2033 | |
---|
| 2034 | self.data_msg.pack_string( "%s" ) |
---|
| 2035 | self.data_msg.pack_string( str( value ) ) |
---|
| 2036 | |
---|
| 2037 | return ( self.meta_msg.get_buffer(), self.data_msg.get_buffer() ) |
---|
| 2038 | |
---|
[26] | 2039 | def printTime( ): |
---|
[354] | 2040 | |
---|
[691] | 2041 | """Print current time/date in human readable format for log/debug""" |
---|
[26] | 2042 | |
---|
[691] | 2043 | return time.strftime("%a, %d %b %Y %H:%M:%S") |
---|
[26] | 2044 | |
---|
| 2045 | def debug_msg( level, msg ): |
---|
[354] | 2046 | |
---|
[691] | 2047 | """Print msg if at or above current debug level""" |
---|
[26] | 2048 | |
---|
[691] | 2049 | global DAEMONIZE, DEBUG_LEVEL, SYSLOG_LEVEL |
---|
[377] | 2050 | |
---|
[691] | 2051 | if (not DAEMONIZE and DEBUG_LEVEL >= level): |
---|
| 2052 | sys.stderr.write( msg + '\n' ) |
---|
[26] | 2053 | |
---|
[691] | 2054 | if (DAEMONIZE and USE_SYSLOG and SYSLOG_LEVEL >= level): |
---|
| 2055 | syslog.syslog( msg ) |
---|
[373] | 2056 | |
---|
[307] | 2057 | def write_pidfile(): |
---|
| 2058 | |
---|
[691] | 2059 | # Write pidfile if PIDFILE is set |
---|
| 2060 | # |
---|
| 2061 | if PIDFILE: |
---|
[307] | 2062 | |
---|
[691] | 2063 | pid = os.getpid() |
---|
[354] | 2064 | |
---|
[691] | 2065 | pidfile = open( PIDFILE, 'w' ) |
---|
[354] | 2066 | |
---|
[691] | 2067 | pidfile.write( str( pid ) ) |
---|
| 2068 | pidfile.close() |
---|
[307] | 2069 | |
---|
[23] | 2070 | def main(): |
---|
[354] | 2071 | |
---|
[691] | 2072 | """Application start""" |
---|
[23] | 2073 | |
---|
[837] | 2074 | global PBSQuery, PBSError, lsfObject, pyslurm |
---|
[691] | 2075 | global SYSLOG_FACILITY, USE_SYSLOG, BATCH_API, DAEMONIZE |
---|
[256] | 2076 | |
---|
[691] | 2077 | if not processArgs( sys.argv[1:] ): |
---|
[354] | 2078 | |
---|
[691] | 2079 | sys.exit( 1 ) |
---|
[212] | 2080 | |
---|
[691] | 2081 | # Load appropriate DataGatherer depending on which BATCH_API is set |
---|
| 2082 | # and any required modules for the Gatherer |
---|
| 2083 | # |
---|
| 2084 | if BATCH_API == 'pbs': |
---|
[256] | 2085 | |
---|
[691] | 2086 | try: |
---|
| 2087 | from PBSQuery import PBSQuery, PBSError |
---|
[256] | 2088 | |
---|
[787] | 2089 | except ImportError, details: |
---|
[256] | 2090 | |
---|
[787] | 2091 | print "FATAL ERROR: BATCH_API set to 'pbs' but python module 'pbs_python' cannot be loaded" |
---|
| 2092 | print details |
---|
[691] | 2093 | sys.exit( 1 ) |
---|
[256] | 2094 | |
---|
[691] | 2095 | gather = PbsDataGatherer() |
---|
[256] | 2096 | |
---|
[691] | 2097 | elif BATCH_API == 'sge': |
---|
[256] | 2098 | |
---|
[691] | 2099 | # Tested with SGE 6.0u11. |
---|
| 2100 | # |
---|
| 2101 | gather = SgeDataGatherer() |
---|
[368] | 2102 | |
---|
[691] | 2103 | elif BATCH_API == 'lsf': |
---|
[368] | 2104 | |
---|
[691] | 2105 | try: |
---|
| 2106 | from lsfObject import lsfObject |
---|
| 2107 | except: |
---|
[786] | 2108 | print "FATAL ERROR: BATCH_API set to 'lsf' but python module is not found or installed" |
---|
[691] | 2109 | sys.exit( 1) |
---|
[256] | 2110 | |
---|
[691] | 2111 | gather = LsfDataGatherer() |
---|
[524] | 2112 | |
---|
[837] | 2113 | elif BATCH_API == 'slurm': |
---|
| 2114 | |
---|
| 2115 | try: |
---|
| 2116 | import pyslurm |
---|
| 2117 | except: |
---|
| 2118 | print "FATAL ERROR: BATCH_API set to 'slurm' but python module is not found or installed" |
---|
| 2119 | sys.exit( 1) |
---|
| 2120 | |
---|
| 2121 | gather = SLURMDataGatherer() |
---|
| 2122 | |
---|
[691] | 2123 | else: |
---|
[786] | 2124 | print "FATAL ERROR: unknown BATCH_API '" + BATCH_API + "' is not supported" |
---|
[354] | 2125 | |
---|
[691] | 2126 | sys.exit( 1 ) |
---|
[256] | 2127 | |
---|
[691] | 2128 | if( DAEMONIZE and USE_SYSLOG ): |
---|
[373] | 2129 | |
---|
[691] | 2130 | syslog.openlog( 'jobmond', syslog.LOG_NOWAIT, SYSLOG_FACILITY ) |
---|
[373] | 2131 | |
---|
[691] | 2132 | if DAEMONIZE: |
---|
[354] | 2133 | |
---|
[691] | 2134 | gather.daemon() |
---|
| 2135 | else: |
---|
| 2136 | gather.run() |
---|
[23] | 2137 | |
---|
[256] | 2138 | # wh00t? someone started me! :) |
---|
[65] | 2139 | # |
---|
[23] | 2140 | if __name__ == '__main__': |
---|
[691] | 2141 | main() |
---|