[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 702 2013-03-21 15:55:35Z ramonb $ |
---|
[227] | 23 | # |
---|
[23] | 24 | |
---|
[694] | 25 | # vi :set ts=4 |
---|
| 26 | |
---|
[471] | 27 | import sys, getopt, ConfigParser, time, os, socket, string, re |
---|
[699] | 28 | import xdrlib, socket, syslog, xml, xml.sax, shlex |
---|
[318] | 29 | from xml.sax.handler import feature_namespaces |
---|
[623] | 30 | from collections import deque |
---|
[699] | 31 | from glob import glob |
---|
[318] | 32 | |
---|
[691] | 33 | VERSION='0.4+SVN' |
---|
[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 | |
---|
[701] | 60 | global PIDFILE JOBMOND_CONF |
---|
| 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 |
---|
| 72 | usage() |
---|
| 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 | return self.conf_dict[ 'udp_send_channel' ] |
---|
| 356 | |
---|
| 357 | def getSectionLastOption( self, section, option ): |
---|
| 358 | |
---|
| 359 | """ |
---|
| 360 | Get last option set in a config section that could be set multiple times in multiple (include) files. |
---|
| 361 | |
---|
| 362 | i.e.: getSectionLastOption( 'globals', 'send_metadata_interval' ) |
---|
| 363 | """ |
---|
| 364 | |
---|
| 365 | self.checkConfDict() |
---|
| 366 | value = None |
---|
| 367 | |
---|
| 368 | if not self.conf_dict.has_key( section ): |
---|
| 369 | |
---|
| 370 | return None |
---|
| 371 | |
---|
| 372 | # Could be set multiple times in multiple (include) files: get last one set |
---|
| 373 | for c in self.conf_dict[ section ]: |
---|
| 374 | |
---|
| 375 | if c.has_key( option ): |
---|
| 376 | |
---|
| 377 | cluster_name = c[ option ][0] |
---|
| 378 | |
---|
| 379 | return cluster_name |
---|
| 380 | |
---|
| 381 | def getClusterName( self ): |
---|
| 382 | |
---|
| 383 | return self.getSectionLastOption( 'cluster', 'name' ) |
---|
| 384 | |
---|
| 385 | def getVal( self, section, option ): |
---|
| 386 | |
---|
| 387 | return self.getSectionLastOption( section, option ) |
---|
| 388 | |
---|
[691] | 389 | def getInt( self, section, valname ): |
---|
[520] | 390 | |
---|
[691] | 391 | value = self.getVal( section, valname ) |
---|
[520] | 392 | |
---|
[691] | 393 | if not value: |
---|
[699] | 394 | return None |
---|
[520] | 395 | |
---|
[691] | 396 | return int( value ) |
---|
[520] | 397 | |
---|
[691] | 398 | def getStr( self, section, valname ): |
---|
[520] | 399 | |
---|
[691] | 400 | value = self.getVal( section, valname ) |
---|
[520] | 401 | |
---|
[691] | 402 | if not value: |
---|
[699] | 403 | return None |
---|
[520] | 404 | |
---|
[691] | 405 | return str( value ) |
---|
[520] | 406 | |
---|
| 407 | def findGmetric(): |
---|
| 408 | |
---|
[691] | 409 | for dir in os.path.expandvars( '$PATH' ).split( ':' ): |
---|
[520] | 410 | |
---|
[691] | 411 | guess = '%s/%s' %( dir, 'gmetric' ) |
---|
[520] | 412 | |
---|
[691] | 413 | if os.path.exists( guess ): |
---|
[520] | 414 | |
---|
[691] | 415 | return guess |
---|
[520] | 416 | |
---|
[691] | 417 | return False |
---|
[520] | 418 | |
---|
[212] | 419 | def loadConfig( filename ): |
---|
| 420 | |
---|
[691] | 421 | def getlist( cfg_string ): |
---|
[215] | 422 | |
---|
[691] | 423 | my_list = [ ] |
---|
[215] | 424 | |
---|
[691] | 425 | for item_txt in cfg_string.split( ',' ): |
---|
[215] | 426 | |
---|
[691] | 427 | sep_char = None |
---|
[215] | 428 | |
---|
[691] | 429 | item_txt = item_txt.strip() |
---|
[215] | 430 | |
---|
[691] | 431 | for s_char in [ "'", '"' ]: |
---|
[215] | 432 | |
---|
[691] | 433 | if item_txt.find( s_char ) != -1: |
---|
[215] | 434 | |
---|
[691] | 435 | if item_txt.count( s_char ) != 2: |
---|
[215] | 436 | |
---|
[691] | 437 | print 'Missing quote: %s' %item_txt |
---|
| 438 | sys.exit( 1 ) |
---|
[215] | 439 | |
---|
[691] | 440 | else: |
---|
[215] | 441 | |
---|
[691] | 442 | sep_char = s_char |
---|
| 443 | break |
---|
[215] | 444 | |
---|
[691] | 445 | if sep_char: |
---|
[215] | 446 | |
---|
[691] | 447 | item_txt = item_txt.split( sep_char )[1] |
---|
[215] | 448 | |
---|
[691] | 449 | my_list.append( item_txt ) |
---|
[215] | 450 | |
---|
[691] | 451 | return my_list |
---|
[215] | 452 | |
---|
[691] | 453 | cfg = ConfigParser.ConfigParser() |
---|
[212] | 454 | |
---|
[691] | 455 | cfg.read( filename ) |
---|
[212] | 456 | |
---|
[691] | 457 | global DEBUG_LEVEL, DAEMONIZE, BATCH_SERVER, BATCH_POLL_INTERVAL |
---|
| 458 | global GMOND_CONF, DETECT_TIME_DIFFS, BATCH_HOST_TRANSLATE |
---|
| 459 | global BATCH_API, QUEUE, GMETRIC_TARGET, USE_SYSLOG |
---|
| 460 | global SYSLOG_LEVEL, SYSLOG_FACILITY, GMETRIC_BINARY |
---|
[701] | 461 | global METRIC_MAX_VAL_LEN |
---|
[212] | 462 | |
---|
[701] | 463 | DEBUG_LEVEL = cfg.getint( 'DEFAULT', 'DEBUG_LEVEL' ) |
---|
[212] | 464 | |
---|
[701] | 465 | DAEMONIZE = cfg.getboolean( 'DEFAULT', 'DAEMONIZE' ) |
---|
[212] | 466 | |
---|
[691] | 467 | SYSLOG_LEVEL = -1 |
---|
[701] | 468 | SYSLOG_FACILITY = None |
---|
[377] | 469 | |
---|
[691] | 470 | try: |
---|
[701] | 471 | USE_SYSLOG = cfg.getboolean( 'DEFAULT', 'USE_SYSLOG' ) |
---|
[212] | 472 | |
---|
[691] | 473 | except ConfigParser.NoOptionError: |
---|
[373] | 474 | |
---|
[701] | 475 | USE_SYSLOG = True |
---|
[373] | 476 | |
---|
[691] | 477 | debug_msg( 0, 'ERROR: no option USE_SYSLOG found: assuming yes' ) |
---|
[373] | 478 | |
---|
[691] | 479 | if USE_SYSLOG: |
---|
[373] | 480 | |
---|
[691] | 481 | try: |
---|
[701] | 482 | SYSLOG_LEVEL = cfg.getint( 'DEFAULT', 'SYSLOG_LEVEL' ) |
---|
[373] | 483 | |
---|
[691] | 484 | except ConfigParser.NoOptionError: |
---|
[373] | 485 | |
---|
[691] | 486 | debug_msg( 0, 'ERROR: no option SYSLOG_LEVEL found: assuming level 0' ) |
---|
[701] | 487 | SYSLOG_LEVEL = 0 |
---|
[373] | 488 | |
---|
[691] | 489 | try: |
---|
[373] | 490 | |
---|
[691] | 491 | SYSLOG_FACILITY = eval( 'syslog.LOG_' + cfg.get( 'DEFAULT', 'SYSLOG_FACILITY' ) ) |
---|
[373] | 492 | |
---|
[691] | 493 | except ConfigParser.NoOptionError: |
---|
[373] | 494 | |
---|
[691] | 495 | SYSLOG_FACILITY = syslog.LOG_DAEMON |
---|
[373] | 496 | |
---|
[691] | 497 | debug_msg( 0, 'ERROR: no option SYSLOG_FACILITY found: assuming facility DAEMON' ) |
---|
[373] | 498 | |
---|
[691] | 499 | try: |
---|
[373] | 500 | |
---|
[701] | 501 | BATCH_SERVER = cfg.get( 'DEFAULT', 'BATCH_SERVER' ) |
---|
[212] | 502 | |
---|
[691] | 503 | except ConfigParser.NoOptionError: |
---|
[265] | 504 | |
---|
[691] | 505 | # Backwards compatibility for old configs |
---|
| 506 | # |
---|
[265] | 507 | |
---|
[701] | 508 | BATCH_SERVER = cfg.get( 'DEFAULT', 'TORQUE_SERVER' ) |
---|
| 509 | api_guess = 'pbs' |
---|
[691] | 510 | |
---|
| 511 | try: |
---|
| 512 | |
---|
[701] | 513 | BATCH_POLL_INTERVAL = cfg.getint( 'DEFAULT', 'BATCH_POLL_INTERVAL' ) |
---|
[265] | 514 | |
---|
[691] | 515 | except ConfigParser.NoOptionError: |
---|
[265] | 516 | |
---|
[691] | 517 | # Backwards compatibility for old configs |
---|
| 518 | # |
---|
[265] | 519 | |
---|
[701] | 520 | BATCH_POLL_INTERVAL = cfg.getint( 'DEFAULT', 'TORQUE_POLL_INTERVAL' ) |
---|
| 521 | api_guess = 'pbs' |
---|
[691] | 522 | |
---|
| 523 | try: |
---|
[212] | 524 | |
---|
[701] | 525 | GMOND_CONF = cfg.get( 'DEFAULT', 'GMOND_CONF' ) |
---|
[353] | 526 | |
---|
[691] | 527 | except ConfigParser.NoOptionError: |
---|
[353] | 528 | |
---|
[691] | 529 | # Not specified: assume /etc/gmond.conf |
---|
| 530 | # |
---|
[701] | 531 | GMOND_CONF = '/etc/gmond.conf' |
---|
[353] | 532 | |
---|
[701] | 533 | ganglia_cfg = GangliaConfigParser( GMOND_CONF ) |
---|
[449] | 534 | |
---|
[691] | 535 | # Let's try to find the GMETRIC_TARGET ourselves first from GMOND_CONF |
---|
| 536 | # |
---|
[701] | 537 | gmetric_dest_ip = ganglia_cfg.getStr( 'udp_send_channel', 'mcast_join' ) |
---|
[449] | 538 | |
---|
[691] | 539 | if not gmetric_dest_ip: |
---|
[449] | 540 | |
---|
[691] | 541 | # Maybe unicast target then |
---|
| 542 | # |
---|
[701] | 543 | gmetric_dest_ip = ganglia_cfg.getStr( 'udp_send_channel', 'host' ) |
---|
[449] | 544 | |
---|
[701] | 545 | gmetric_dest_port = ganglia_cfg.getStr( 'udp_send_channel', 'port' ) |
---|
[520] | 546 | |
---|
[701] | 547 | #TODO: use multiple udp send channels (if found) |
---|
[691] | 548 | if gmetric_dest_ip and gmetric_dest_port: |
---|
[520] | 549 | |
---|
[701] | 550 | GMETRIC_TARGET = '%s:%s' %( gmetric_dest_ip, gmetric_dest_port ) |
---|
[691] | 551 | else: |
---|
[520] | 552 | |
---|
[701] | 553 | debug_msg( 0, "WARNING: Can't parse udp_send_channel from: '%s' - Trying: %s" %( GMOND_CONF, JOBMOND_CONF ) ) |
---|
[520] | 554 | |
---|
[691] | 555 | # Couldn't figure it out: let's see if it's in our jobmond.conf |
---|
| 556 | # |
---|
| 557 | try: |
---|
[520] | 558 | |
---|
[691] | 559 | GMETRIC_TARGET = cfg.get( 'DEFAULT', 'GMETRIC_TARGET' ) |
---|
[520] | 560 | |
---|
[691] | 561 | # Guess not: now just give up |
---|
[701] | 562 | |
---|
[691] | 563 | except ConfigParser.NoOptionError: |
---|
[520] | 564 | |
---|
[691] | 565 | GMETRIC_TARGET = None |
---|
[520] | 566 | |
---|
[691] | 567 | 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] | 568 | |
---|
[701] | 569 | gmetric_bin = findGmetric() |
---|
[520] | 570 | |
---|
[701] | 571 | if gmetric_bin: |
---|
[520] | 572 | |
---|
[701] | 573 | GMETRIC_BINARY = gmetric_bin |
---|
| 574 | else: |
---|
| 575 | debug_msg( 0, "WARNING: Can't find gmetric binary anywhere in $PATH" ) |
---|
[520] | 576 | |
---|
[701] | 577 | try: |
---|
[520] | 578 | |
---|
[701] | 579 | GMETRIC_BINARY = cfg.get( 'DEFAULT', 'GMETRIC_BINARY' ) |
---|
[520] | 580 | |
---|
[701] | 581 | except ConfigParser.NoOptionError: |
---|
[520] | 582 | |
---|
[701] | 583 | debug_msg( 0, "FATAL ERROR: GMETRIC_BINARY not set and not in $PATH" ) |
---|
| 584 | sys.exit( 1 ) |
---|
[520] | 585 | |
---|
[701] | 586 | #TODO: is this really still needed or should be automatic |
---|
[691] | 587 | DETECT_TIME_DIFFS = cfg.getboolean( 'DEFAULT', 'DETECT_TIME_DIFFS' ) |
---|
[212] | 588 | |
---|
[701] | 589 | BATCH_HOST_TRANSLATE = getlist( cfg.get( 'DEFAULT', 'BATCH_HOST_TRANSLATE' ) ) |
---|
[215] | 590 | |
---|
[691] | 591 | try: |
---|
[256] | 592 | |
---|
[691] | 593 | BATCH_API = cfg.get( 'DEFAULT', 'BATCH_API' ) |
---|
[266] | 594 | |
---|
[691] | 595 | except ConfigParser.NoOptionError, detail: |
---|
[266] | 596 | |
---|
[691] | 597 | if BATCH_SERVER and api_guess: |
---|
[354] | 598 | |
---|
[691] | 599 | BATCH_API = api_guess |
---|
| 600 | else: |
---|
| 601 | debug_msg( 0, "FATAL ERROR: BATCH_API not set and can't make guess" ) |
---|
| 602 | sys.exit( 1 ) |
---|
[317] | 603 | |
---|
[691] | 604 | try: |
---|
[317] | 605 | |
---|
[691] | 606 | QUEUE = getlist( cfg.get( 'DEFAULT', 'QUEUE' ) ) |
---|
[317] | 607 | |
---|
[691] | 608 | except ConfigParser.NoOptionError, detail: |
---|
[317] | 609 | |
---|
[691] | 610 | QUEUE = None |
---|
[353] | 611 | |
---|
[701] | 612 | METRIC_MAX_VAL_LEN = ganglia_cfg.getInt( 'globals', 'max_udp_msg_len' ) |
---|
| 613 | |
---|
[691] | 614 | return True |
---|
[212] | 615 | |
---|
[507] | 616 | def fqdn_parts (fqdn): |
---|
[520] | 617 | |
---|
[691] | 618 | """Return pair of host and domain for fully-qualified domain name arg.""" |
---|
[520] | 619 | |
---|
[691] | 620 | parts = fqdn.split (".") |
---|
[520] | 621 | |
---|
[691] | 622 | return (parts[0], string.join(parts[1:], ".")) |
---|
[507] | 623 | |
---|
[61] | 624 | class DataProcessor: |
---|
[355] | 625 | |
---|
[691] | 626 | """Class for processing of data""" |
---|
[61] | 627 | |
---|
[691] | 628 | binary = None |
---|
[61] | 629 | |
---|
[691] | 630 | def __init__( self, binary=None ): |
---|
[355] | 631 | |
---|
[691] | 632 | """Remember alternate binary location if supplied""" |
---|
[61] | 633 | |
---|
[691] | 634 | global GMETRIC_BINARY, GMOND_CONF |
---|
[449] | 635 | |
---|
[691] | 636 | if binary: |
---|
| 637 | self.binary = binary |
---|
[61] | 638 | |
---|
[691] | 639 | if not self.binary: |
---|
| 640 | self.binary = GMETRIC_BINARY |
---|
[449] | 641 | |
---|
[691] | 642 | # Timeout for XML |
---|
| 643 | # |
---|
| 644 | # From ganglia's documentation: |
---|
| 645 | # |
---|
| 646 | # 'A metric will be deleted DMAX seconds after it is received, and |
---|
| 647 | # DMAX=0 means eternal life.' |
---|
[61] | 648 | |
---|
[691] | 649 | self.dmax = str( int( int( BATCH_POLL_INTERVAL ) * 2 ) ) |
---|
[80] | 650 | |
---|
[691] | 651 | if GMOND_CONF: |
---|
[354] | 652 | |
---|
[691] | 653 | incompatible = self.checkGmetricVersion() |
---|
[61] | 654 | |
---|
[691] | 655 | if incompatible: |
---|
[355] | 656 | |
---|
[692] | 657 | debug_msg( 0, 'Gmetric version not compatible, please upgrade to at least 3.4.0' ) |
---|
[691] | 658 | sys.exit( 1 ) |
---|
[65] | 659 | |
---|
[691] | 660 | def checkGmetricVersion( self ): |
---|
[355] | 661 | |
---|
[691] | 662 | """ |
---|
[692] | 663 | Check version of gmetric is at least 3.4.0 |
---|
[691] | 664 | for the syntax we use |
---|
| 665 | """ |
---|
[65] | 666 | |
---|
[691] | 667 | global METRIC_MAX_VAL_LEN, GMETRIC_TARGET |
---|
[255] | 668 | |
---|
[691] | 669 | incompatible = 0 |
---|
[341] | 670 | |
---|
[691] | 671 | gfp = os.popen( self.binary + ' --version' ) |
---|
[692] | 672 | lines = gfp.readlines() |
---|
[65] | 673 | |
---|
[691] | 674 | gfp.close() |
---|
[355] | 675 | |
---|
[691] | 676 | for line in lines: |
---|
[355] | 677 | |
---|
[691] | 678 | line = line.split( ' ' ) |
---|
[65] | 679 | |
---|
[691] | 680 | if len( line ) == 2 and str( line ).find( 'gmetric' ) != -1: |
---|
| 681 | |
---|
| 682 | gmetric_version = line[1].split( '\n' )[0] |
---|
[65] | 683 | |
---|
[691] | 684 | version_major = int( gmetric_version.split( '.' )[0] ) |
---|
| 685 | version_minor = int( gmetric_version.split( '.' )[1] ) |
---|
| 686 | version_patch = int( gmetric_version.split( '.' )[2] ) |
---|
[65] | 687 | |
---|
[691] | 688 | incompatible = 0 |
---|
[65] | 689 | |
---|
[691] | 690 | if version_major < 3: |
---|
[65] | 691 | |
---|
[691] | 692 | incompatible = 1 |
---|
| 693 | |
---|
| 694 | elif version_major == 3: |
---|
[65] | 695 | |
---|
[692] | 696 | if version_minor < 4: |
---|
[65] | 697 | |
---|
[692] | 698 | incompatible = 1 |
---|
[65] | 699 | |
---|
[691] | 700 | return incompatible |
---|
[65] | 701 | |
---|
[691] | 702 | def multicastGmetric( self, metricname, metricval, valtype='string', units='' ): |
---|
[355] | 703 | |
---|
[691] | 704 | """Call gmetric binary and multicast""" |
---|
[65] | 705 | |
---|
[691] | 706 | cmd = self.binary |
---|
[65] | 707 | |
---|
[691] | 708 | if GMETRIC_TARGET: |
---|
[61] | 709 | |
---|
[691] | 710 | GMETRIC_TARGET_HOST = GMETRIC_TARGET.split( ':' )[0] |
---|
| 711 | GMETRIC_TARGET_PORT = GMETRIC_TARGET.split( ':' )[1] |
---|
[353] | 712 | |
---|
[691] | 713 | metric_debug = "[gmetric] name: %s - val: %s - dmax: %s" %( str( metricname ), str( metricval ), str( self.dmax ) ) |
---|
[353] | 714 | |
---|
[691] | 715 | debug_msg( 10, printTime() + ' ' + metric_debug) |
---|
[353] | 716 | |
---|
[691] | 717 | gm = Gmetric( GMETRIC_TARGET_HOST, GMETRIC_TARGET_PORT ) |
---|
[353] | 718 | |
---|
[691] | 719 | gm.send( str( metricname ), str( metricval ), str( self.dmax ), valtype, units ) |
---|
[353] | 720 | |
---|
[691] | 721 | else: |
---|
| 722 | try: |
---|
| 723 | cmd = cmd + ' -c' + GMOND_CONF |
---|
[353] | 724 | |
---|
[691] | 725 | except NameError: |
---|
[353] | 726 | |
---|
[691] | 727 | debug_msg( 10, 'Assuming /etc/gmond.conf for gmetric cmd (omitting)' ) |
---|
[353] | 728 | |
---|
[691] | 729 | cmd = cmd + ' -n' + str( metricname )+ ' -v"' + str( metricval )+ '" -t' + str( valtype ) + ' -d' + str( self.dmax ) |
---|
[353] | 730 | |
---|
[691] | 731 | if len( units ) > 0: |
---|
[409] | 732 | |
---|
[691] | 733 | cmd = cmd + ' -u"' + units + '"' |
---|
[409] | 734 | |
---|
[691] | 735 | debug_msg( 10, printTime() + ' ' + cmd ) |
---|
[353] | 736 | |
---|
[691] | 737 | os.system( cmd ) |
---|
[353] | 738 | |
---|
[318] | 739 | class DataGatherer: |
---|
[23] | 740 | |
---|
[691] | 741 | """Skeleton class for batch system DataGatherer""" |
---|
[256] | 742 | |
---|
[691] | 743 | def printJobs( self, jobs ): |
---|
[355] | 744 | |
---|
[691] | 745 | """Print a jobinfo overview""" |
---|
[318] | 746 | |
---|
[691] | 747 | for name, attrs in self.jobs.items(): |
---|
[318] | 748 | |
---|
[691] | 749 | print 'job %s' %(name) |
---|
[318] | 750 | |
---|
[691] | 751 | for name, val in attrs.items(): |
---|
[318] | 752 | |
---|
[691] | 753 | print '\t%s = %s' %( name, val ) |
---|
[318] | 754 | |
---|
[691] | 755 | def printJob( self, jobs, job_id ): |
---|
[355] | 756 | |
---|
[691] | 757 | """Print job with job_id from jobs""" |
---|
[318] | 758 | |
---|
[691] | 759 | print 'job %s' %(job_id) |
---|
[318] | 760 | |
---|
[691] | 761 | for name, val in jobs[ job_id ].items(): |
---|
[318] | 762 | |
---|
[691] | 763 | print '\t%s = %s' %( name, val ) |
---|
[318] | 764 | |
---|
[691] | 765 | def getAttr( self, attrs, name ): |
---|
[507] | 766 | |
---|
[691] | 767 | """Return certain attribute from dictionary, if exists""" |
---|
[507] | 768 | |
---|
[691] | 769 | if attrs.has_key( name ): |
---|
[507] | 770 | |
---|
[691] | 771 | return attrs[ name ] |
---|
| 772 | else: |
---|
| 773 | return '' |
---|
[507] | 774 | |
---|
[691] | 775 | def jobDataChanged( self, jobs, job_id, attrs ): |
---|
[507] | 776 | |
---|
[691] | 777 | """Check if job with attrs and job_id in jobs has changed""" |
---|
[507] | 778 | |
---|
[691] | 779 | if jobs.has_key( job_id ): |
---|
[507] | 780 | |
---|
[691] | 781 | oldData = jobs[ job_id ] |
---|
| 782 | else: |
---|
| 783 | return 1 |
---|
[507] | 784 | |
---|
[691] | 785 | for name, val in attrs.items(): |
---|
[507] | 786 | |
---|
[691] | 787 | if oldData.has_key( name ): |
---|
[507] | 788 | |
---|
[691] | 789 | if oldData[ name ] != attrs[ name ]: |
---|
[507] | 790 | |
---|
[691] | 791 | return 1 |
---|
[507] | 792 | |
---|
[691] | 793 | else: |
---|
| 794 | return 1 |
---|
[507] | 795 | |
---|
[691] | 796 | return 0 |
---|
[507] | 797 | |
---|
[691] | 798 | def submitJobData( self ): |
---|
[507] | 799 | |
---|
[691] | 800 | """Submit job info list""" |
---|
[507] | 801 | |
---|
[691] | 802 | global BATCH_API |
---|
[512] | 803 | |
---|
[691] | 804 | self.dp.multicastGmetric( 'MONARCH-HEARTBEAT', str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
[507] | 805 | |
---|
[691] | 806 | running_jobs = 0 |
---|
| 807 | queued_jobs = 0 |
---|
[507] | 808 | |
---|
[691] | 809 | # Count how many running/queued jobs we found |
---|
| 810 | # |
---|
| 811 | for jobid, jobattrs in self.jobs.items(): |
---|
[507] | 812 | |
---|
[691] | 813 | if jobattrs[ 'status' ] == 'Q': |
---|
[507] | 814 | |
---|
[691] | 815 | queued_jobs += 1 |
---|
[507] | 816 | |
---|
[691] | 817 | elif jobattrs[ 'status' ] == 'R': |
---|
[507] | 818 | |
---|
[691] | 819 | running_jobs += 1 |
---|
[507] | 820 | |
---|
[691] | 821 | # Report running/queued jobs as seperate metric for a nice RRD graph |
---|
| 822 | # |
---|
| 823 | self.dp.multicastGmetric( 'MONARCH-RJ', str( running_jobs ), 'uint32', 'jobs' ) |
---|
| 824 | self.dp.multicastGmetric( 'MONARCH-QJ', str( queued_jobs ), 'uint32', 'jobs' ) |
---|
[507] | 825 | |
---|
[691] | 826 | # Report down/offline nodes in batch (PBS only ATM) |
---|
| 827 | # |
---|
| 828 | if BATCH_API == 'pbs': |
---|
[512] | 829 | |
---|
[691] | 830 | domain = fqdn_parts( socket.getfqdn() )[1] |
---|
[514] | 831 | |
---|
[691] | 832 | downed_nodes = list() |
---|
| 833 | offline_nodes = list() |
---|
| 834 | |
---|
| 835 | l = ['state'] |
---|
| 836 | |
---|
| 837 | for name, node in self.pq.getnodes().items(): |
---|
[512] | 838 | |
---|
[691] | 839 | if ( node[ 'state' ].find( "down" ) != -1 ): |
---|
[512] | 840 | |
---|
[691] | 841 | downed_nodes.append( name ) |
---|
[512] | 842 | |
---|
[691] | 843 | if ( node[ 'state' ].find( "offline" ) != -1 ): |
---|
[512] | 844 | |
---|
[691] | 845 | offline_nodes.append( name ) |
---|
[512] | 846 | |
---|
[691] | 847 | downnodeslist = do_nodelist( downed_nodes ) |
---|
| 848 | offlinenodeslist = do_nodelist( offline_nodes ) |
---|
[512] | 849 | |
---|
[691] | 850 | down_str = 'nodes=%s domain=%s reported=%s' %( string.join( downnodeslist, ';' ), domain, str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
| 851 | offl_str = 'nodes=%s domain=%s reported=%s' %( string.join( offlinenodeslist, ';' ), domain, str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
| 852 | self.dp.multicastGmetric( 'MONARCH-DOWN' , down_str ) |
---|
| 853 | self.dp.multicastGmetric( 'MONARCH-OFFLINE', offl_str ) |
---|
[514] | 854 | |
---|
[691] | 855 | # Now let's spread the knowledge |
---|
| 856 | # |
---|
| 857 | for jobid, jobattrs in self.jobs.items(): |
---|
[507] | 858 | |
---|
[691] | 859 | # Make gmetric values for each job: respect max gmetric value length |
---|
| 860 | # |
---|
| 861 | gmetric_val = self.compileGmetricVal( jobid, jobattrs ) |
---|
| 862 | metric_increment = 0 |
---|
[507] | 863 | |
---|
[691] | 864 | # If we have more job info than max gmetric value length allows, split it up |
---|
| 865 | # amongst multiple metrics |
---|
| 866 | # |
---|
| 867 | for val in gmetric_val: |
---|
[507] | 868 | |
---|
[691] | 869 | self.dp.multicastGmetric( 'MONARCH-JOB-' + jobid + '-' + str(metric_increment), val ) |
---|
[507] | 870 | |
---|
[691] | 871 | # Increase follow number if this jobinfo is split up amongst more than 1 gmetric |
---|
| 872 | # |
---|
| 873 | metric_increment = metric_increment + 1 |
---|
[507] | 874 | |
---|
[691] | 875 | def compileGmetricVal( self, jobid, jobattrs ): |
---|
[507] | 876 | |
---|
[691] | 877 | """Create a val string for gmetric of jobinfo""" |
---|
[507] | 878 | |
---|
[691] | 879 | gval_lists = [ ] |
---|
| 880 | val_list = { } |
---|
[507] | 881 | |
---|
[691] | 882 | for val_name, val_value in jobattrs.items(): |
---|
[507] | 883 | |
---|
[691] | 884 | # These are our own metric names, i.e.: status, start_timestamp, etc |
---|
| 885 | # |
---|
| 886 | val_list_names_len = len( string.join( val_list.keys() ) ) + len(val_list.keys()) |
---|
[507] | 887 | |
---|
[691] | 888 | # These are their corresponding values |
---|
| 889 | # |
---|
| 890 | val_list_vals_len = len( string.join( val_list.values() ) ) + len(val_list.values()) |
---|
[507] | 891 | |
---|
[691] | 892 | if val_name == 'nodes' and jobattrs['status'] == 'R': |
---|
[507] | 893 | |
---|
[691] | 894 | node_str = None |
---|
[507] | 895 | |
---|
[691] | 896 | for node in val_value: |
---|
[507] | 897 | |
---|
[691] | 898 | if node_str: |
---|
[507] | 899 | |
---|
[691] | 900 | node_str = node_str + ';' + node |
---|
| 901 | else: |
---|
| 902 | node_str = node |
---|
[507] | 903 | |
---|
[691] | 904 | # Make sure if we add this new info, that the total metric's value length does not exceed METRIC_MAX_VAL_LEN |
---|
| 905 | # |
---|
| 906 | if (val_list_names_len + len(val_name) ) + (val_list_vals_len + len(node_str) ) > METRIC_MAX_VAL_LEN: |
---|
[507] | 907 | |
---|
[691] | 908 | # It's too big, we need to make a new gmetric for the additional info |
---|
| 909 | # |
---|
| 910 | val_list[ val_name ] = node_str |
---|
[507] | 911 | |
---|
[691] | 912 | gval_lists.append( val_list ) |
---|
[507] | 913 | |
---|
[691] | 914 | val_list = { } |
---|
| 915 | node_str = None |
---|
[507] | 916 | |
---|
[691] | 917 | val_list[ val_name ] = node_str |
---|
[507] | 918 | |
---|
[691] | 919 | gval_lists.append( val_list ) |
---|
[507] | 920 | |
---|
[691] | 921 | val_list = { } |
---|
[507] | 922 | |
---|
[691] | 923 | elif val_value != '': |
---|
[507] | 924 | |
---|
[691] | 925 | # Make sure if we add this new info, that the total metric's value length does not exceed METRIC_MAX_VAL_LEN |
---|
| 926 | # |
---|
| 927 | if (val_list_names_len + len(val_name) ) + (val_list_vals_len + len(str(val_value)) ) > METRIC_MAX_VAL_LEN: |
---|
[507] | 928 | |
---|
[691] | 929 | # It's too big, we need to make a new gmetric for the additional info |
---|
| 930 | # |
---|
| 931 | gval_lists.append( val_list ) |
---|
[507] | 932 | |
---|
[691] | 933 | val_list = { } |
---|
[507] | 934 | |
---|
[691] | 935 | val_list[ val_name ] = val_value |
---|
[507] | 936 | |
---|
[691] | 937 | if len( val_list ) > 0: |
---|
[507] | 938 | |
---|
[691] | 939 | gval_lists.append( val_list ) |
---|
[507] | 940 | |
---|
[691] | 941 | str_list = [ ] |
---|
[507] | 942 | |
---|
[691] | 943 | # Now append the value names and values together, i.e.: stop_timestamp=value, etc |
---|
| 944 | # |
---|
| 945 | for val_list in gval_lists: |
---|
[507] | 946 | |
---|
[691] | 947 | my_val_str = None |
---|
[507] | 948 | |
---|
[691] | 949 | for val_name, val_value in val_list.items(): |
---|
[507] | 950 | |
---|
[691] | 951 | if type(val_value) == list: |
---|
[579] | 952 | |
---|
[691] | 953 | val_value = val_value.join( ',' ) |
---|
[579] | 954 | |
---|
[691] | 955 | if my_val_str: |
---|
[507] | 956 | |
---|
[691] | 957 | try: |
---|
| 958 | # fixme: It's getting |
---|
| 959 | # ('nodes', None) items |
---|
| 960 | my_val_str = my_val_str + ' ' + val_name + '=' + val_value |
---|
| 961 | except: |
---|
| 962 | pass |
---|
[623] | 963 | |
---|
[691] | 964 | else: |
---|
| 965 | my_val_str = val_name + '=' + val_value |
---|
[507] | 966 | |
---|
[691] | 967 | str_list.append( my_val_str ) |
---|
[507] | 968 | |
---|
[691] | 969 | return str_list |
---|
[507] | 970 | |
---|
[691] | 971 | def daemon( self ): |
---|
[355] | 972 | |
---|
[691] | 973 | """Run as daemon forever""" |
---|
[256] | 974 | |
---|
[691] | 975 | # Fork the first child |
---|
| 976 | # |
---|
| 977 | pid = os.fork() |
---|
| 978 | if pid > 0: |
---|
| 979 | sys.exit(0) # end parent |
---|
[256] | 980 | |
---|
[691] | 981 | # creates a session and sets the process group ID |
---|
| 982 | # |
---|
| 983 | os.setsid() |
---|
[318] | 984 | |
---|
[691] | 985 | # Fork the second child |
---|
| 986 | # |
---|
| 987 | pid = os.fork() |
---|
| 988 | if pid > 0: |
---|
| 989 | sys.exit(0) # end parent |
---|
[318] | 990 | |
---|
[691] | 991 | write_pidfile() |
---|
[318] | 992 | |
---|
[691] | 993 | # Go to the root directory and set the umask |
---|
| 994 | # |
---|
| 995 | os.chdir('/') |
---|
| 996 | os.umask(0) |
---|
[318] | 997 | |
---|
[691] | 998 | sys.stdin.close() |
---|
| 999 | sys.stdout.close() |
---|
| 1000 | sys.stderr.close() |
---|
[318] | 1001 | |
---|
[691] | 1002 | os.open('/dev/null', os.O_RDWR) |
---|
| 1003 | os.dup2(0, 1) |
---|
| 1004 | os.dup2(0, 2) |
---|
[318] | 1005 | |
---|
[691] | 1006 | self.run() |
---|
[318] | 1007 | |
---|
[691] | 1008 | def run( self ): |
---|
[355] | 1009 | |
---|
[691] | 1010 | """Main thread""" |
---|
[256] | 1011 | |
---|
[691] | 1012 | while ( 1 ): |
---|
| 1013 | |
---|
| 1014 | self.getJobData() |
---|
| 1015 | self.submitJobData() |
---|
| 1016 | time.sleep( BATCH_POLL_INTERVAL ) |
---|
[256] | 1017 | |
---|
[623] | 1018 | # SGE code by Dave Love <fx@gnu.org>. Tested with SGE 6.0u8 and 6.0u11. May |
---|
| 1019 | # work with SGE 6.1 (else should be easily fixable), but definitely doesn't |
---|
| 1020 | # with 6.2. See also the fixmes. |
---|
[256] | 1021 | |
---|
[507] | 1022 | class NoJobs (Exception): |
---|
[691] | 1023 | """Exception raised by empty job list in qstat output.""" |
---|
| 1024 | pass |
---|
[256] | 1025 | |
---|
[507] | 1026 | class SgeQstatXMLParser(xml.sax.handler.ContentHandler): |
---|
[691] | 1027 | """SAX handler for XML output from Sun Grid Engine's `qstat'.""" |
---|
[318] | 1028 | |
---|
[691] | 1029 | def __init__(self): |
---|
| 1030 | self.value = "" |
---|
| 1031 | self.joblist = [] |
---|
| 1032 | self.job = {} |
---|
| 1033 | self.queue = "" |
---|
| 1034 | self.in_joblist = False |
---|
| 1035 | self.lrequest = False |
---|
| 1036 | self.eltq = deque() |
---|
| 1037 | xml.sax.handler.ContentHandler.__init__(self) |
---|
[318] | 1038 | |
---|
[691] | 1039 | # The structure of the output is as follows (for SGE 6.0). It's |
---|
| 1040 | # similar for 6.1, but radically different for 6.2, and is |
---|
| 1041 | # undocumented generally. Unfortunately it's voluminous, and probably |
---|
| 1042 | # doesn't scale to large clusters/queues. |
---|
[318] | 1043 | |
---|
[691] | 1044 | # <detailed_job_info xmlns:xsd="http://www.w3.org/2001/XMLSchema"> |
---|
| 1045 | # <djob_info> |
---|
| 1046 | # <qmaster_response> <!-- job --> |
---|
| 1047 | # ... |
---|
| 1048 | # <JB_ja_template> |
---|
| 1049 | # <ulong_sublist> |
---|
| 1050 | # ... <!-- start_time, state ... --> |
---|
| 1051 | # </ulong_sublist> |
---|
| 1052 | # </JB_ja_template> |
---|
| 1053 | # <JB_ja_tasks> |
---|
| 1054 | # <ulong_sublist> |
---|
| 1055 | # ... <!-- task info |
---|
| 1056 | # </ulong_sublist> |
---|
| 1057 | # ... |
---|
| 1058 | # </JB_ja_tasks> |
---|
| 1059 | # ... |
---|
| 1060 | # </qmaster_response> |
---|
| 1061 | # </djob_info> |
---|
| 1062 | # <messages> |
---|
| 1063 | # ... |
---|
[318] | 1064 | |
---|
[691] | 1065 | # NB. We might treat each task as a separate job, like |
---|
| 1066 | # straight qstat output, but the web interface expects jobs to |
---|
| 1067 | # be identified by integers, not, say, <job number>.<task>. |
---|
[318] | 1068 | |
---|
[691] | 1069 | # So, I lied. If the job list is empty, we get invalid XML |
---|
| 1070 | # like this, which we need to defend against: |
---|
[318] | 1071 | |
---|
[691] | 1072 | # <unknown_jobs xmlns:xsd="http://www.w3.org/2001/XMLSchema"> |
---|
| 1073 | # <> |
---|
| 1074 | # <ST_name>*</ST_name> |
---|
| 1075 | # </> |
---|
| 1076 | # </unknown_jobs> |
---|
[318] | 1077 | |
---|
[691] | 1078 | def startElement(self, name, attrs): |
---|
| 1079 | self.value = "" |
---|
| 1080 | if name == "djob_info": # job list |
---|
| 1081 | self.in_joblist = True |
---|
| 1082 | # The job container is "qmaster_response" in SGE 6.0 |
---|
| 1083 | # and 6.1, but "element" in 6.2. This is only the very |
---|
| 1084 | # start of what's necessary for 6.2, though (sigh). |
---|
| 1085 | elif (name == "qmaster_response" or name == "element") \ |
---|
| 1086 | and self.eltq[-1] == "djob_info": # job |
---|
| 1087 | self.job = {"job_state": "U", "slots": 0, |
---|
| 1088 | "nodes": [], "queued_timestamp": "", |
---|
| 1089 | "queued_timestamp": "", "queue": "", |
---|
| 1090 | "ppn": "0", "RN_max": 0, |
---|
| 1091 | # fixme in endElement |
---|
| 1092 | "requested_memory": 0, "requested_time": 0 |
---|
| 1093 | } |
---|
| 1094 | self.joblist.append(self.job) |
---|
| 1095 | elif name == "qstat_l_requests": # resource request |
---|
| 1096 | self.lrequest = True |
---|
| 1097 | elif name == "unknown_jobs": |
---|
| 1098 | raise NoJobs |
---|
| 1099 | self.eltq.append (name) |
---|
[318] | 1100 | |
---|
[691] | 1101 | def characters(self, ch): |
---|
| 1102 | self.value += ch |
---|
[318] | 1103 | |
---|
[691] | 1104 | def endElement(self, name): |
---|
| 1105 | """Snarf job elements contents into job dictionary. |
---|
| 1106 | Translate keys if appropriate.""" |
---|
[318] | 1107 | |
---|
[691] | 1108 | name_trans = { |
---|
| 1109 | "JB_job_number": "number", |
---|
| 1110 | "JB_job_name": "name", "JB_owner": "owner", |
---|
| 1111 | "queue_name": "queue", "JAT_start_time": "start_timestamp", |
---|
| 1112 | "JB_submission_time": "queued_timestamp" |
---|
| 1113 | } |
---|
| 1114 | value = self.value |
---|
| 1115 | self.eltq.pop () |
---|
[318] | 1116 | |
---|
[691] | 1117 | if name == "djob_info": |
---|
| 1118 | self.in_joblist = False |
---|
| 1119 | self.job = {} |
---|
| 1120 | elif name == "JAT_master_queue": |
---|
| 1121 | self.job["queue"] = value.split("@")[0] |
---|
| 1122 | elif name == "JG_qhostname": |
---|
| 1123 | if not (value in self.job["nodes"]): |
---|
| 1124 | self.job["nodes"].append(value) |
---|
| 1125 | elif name == "JG_slots": # slots in use |
---|
| 1126 | self.job["slots"] += int(value) |
---|
| 1127 | elif name == "RN_max": # requested slots (tasks or parallel) |
---|
| 1128 | self.job["RN_max"] = max (self.job["RN_max"], |
---|
| 1129 | int(value)) |
---|
| 1130 | elif name == "JAT_state": # job state (bitwise or) |
---|
| 1131 | value = int (value) |
---|
| 1132 | # Status values from sge_jobL.h |
---|
| 1133 | #define JIDLE 0x00000000 |
---|
| 1134 | #define JHELD 0x00000010 |
---|
| 1135 | #define JMIGRATING 0x00000020 |
---|
| 1136 | #define JQUEUED 0x00000040 |
---|
| 1137 | #define JRUNNING 0x00000080 |
---|
| 1138 | #define JSUSPENDED 0x00000100 |
---|
| 1139 | #define JTRANSFERING 0x00000200 |
---|
| 1140 | #define JDELETED 0x00000400 |
---|
| 1141 | #define JWAITING 0x00000800 |
---|
| 1142 | #define JEXITING 0x00001000 |
---|
| 1143 | #define JWRITTEN 0x00002000 |
---|
| 1144 | #define JSUSPENDED_ON_THRESHOLD 0x00010000 |
---|
| 1145 | #define JFINISHED 0x00010000 |
---|
| 1146 | if value & 0x80: |
---|
| 1147 | self.job["status"] = "R" |
---|
| 1148 | elif value & 0x40: |
---|
| 1149 | self.job["status"] = "Q" |
---|
| 1150 | else: |
---|
| 1151 | self.job["status"] = "O" # `other' |
---|
| 1152 | elif name == "CE_name" and self.lrequest and self.value in \ |
---|
| 1153 | ("h_cpu", "s_cpu", "cpu", "h_core", "s_core"): |
---|
| 1154 | # We're in a container for an interesting resource |
---|
| 1155 | # request; record which type. |
---|
| 1156 | self.lrequest = self.value |
---|
| 1157 | elif name == "CE_doubleval" and self.lrequest: |
---|
| 1158 | # if we're in a container for an interesting |
---|
| 1159 | # resource request, use the maxmimum of the hard |
---|
| 1160 | # and soft requests to record the requested CPU |
---|
| 1161 | # or core. Fixme: I'm not sure if this logic is |
---|
| 1162 | # right. |
---|
| 1163 | if self.lrequest in ("h_core", "s_core"): |
---|
| 1164 | self.job["requested_memory"] = \ |
---|
| 1165 | max (float (value), |
---|
| 1166 | self.job["requested_memory"]) |
---|
| 1167 | # Fixme: Check what cpu means, c.f [hs]_cpu. |
---|
| 1168 | elif self.lrequest in ("h_cpu", "s_cpu", "cpu"): |
---|
| 1169 | self.job["requested_time"] = \ |
---|
| 1170 | max (float (value), |
---|
| 1171 | self.job["requested_time"]) |
---|
| 1172 | elif name == "qstat_l_requests": |
---|
| 1173 | self.lrequest = False |
---|
| 1174 | elif self.job and self.in_joblist: |
---|
| 1175 | if name in name_trans: |
---|
| 1176 | name = name_trans[name] |
---|
| 1177 | self.job[name] = value |
---|
[318] | 1178 | |
---|
[507] | 1179 | # Abstracted from PBS original. |
---|
| 1180 | # Fixme: Is it worth (or appropriate for PBS) sorting the result? |
---|
[520] | 1181 | # |
---|
| 1182 | def do_nodelist( nodes ): |
---|
| 1183 | |
---|
[691] | 1184 | """Translate node list as appropriate.""" |
---|
[520] | 1185 | |
---|
[691] | 1186 | nodeslist = [ ] |
---|
| 1187 | my_domain = fqdn_parts( socket.getfqdn() )[1] |
---|
[520] | 1188 | |
---|
[691] | 1189 | for node in nodes: |
---|
[520] | 1190 | |
---|
[691] | 1191 | host = node.split( '/' )[0] # not relevant for SGE |
---|
| 1192 | h, host_domain = fqdn_parts(host) |
---|
[520] | 1193 | |
---|
[691] | 1194 | if host_domain == my_domain: |
---|
[520] | 1195 | |
---|
[691] | 1196 | host = h |
---|
[520] | 1197 | |
---|
[691] | 1198 | if nodeslist.count( host ) == 0: |
---|
[520] | 1199 | |
---|
[691] | 1200 | for translate_pattern in BATCH_HOST_TRANSLATE: |
---|
[520] | 1201 | |
---|
[691] | 1202 | if translate_pattern.find( '/' ) != -1: |
---|
[520] | 1203 | |
---|
[691] | 1204 | translate_orig = \ |
---|
| 1205 | translate_pattern.split( '/' )[1] |
---|
| 1206 | translate_new = \ |
---|
| 1207 | translate_pattern.split( '/' )[2] |
---|
| 1208 | host = re.sub( translate_orig, |
---|
| 1209 | translate_new, host ) |
---|
| 1210 | if not host in nodeslist: |
---|
| 1211 | nodeslist.append( host ) |
---|
| 1212 | return nodeslist |
---|
[318] | 1213 | |
---|
| 1214 | class SgeDataGatherer(DataGatherer): |
---|
| 1215 | |
---|
[691] | 1216 | jobs = {} |
---|
[61] | 1217 | |
---|
[691] | 1218 | def __init__( self ): |
---|
| 1219 | self.jobs = {} |
---|
| 1220 | self.timeoffset = 0 |
---|
| 1221 | self.dp = DataProcessor() |
---|
[318] | 1222 | |
---|
[691] | 1223 | def getJobData( self ): |
---|
| 1224 | """Gather all data on current jobs in SGE""" |
---|
[318] | 1225 | |
---|
[691] | 1226 | import popen2 |
---|
[318] | 1227 | |
---|
[691] | 1228 | self.cur_time = 0 |
---|
| 1229 | queues = "" |
---|
| 1230 | if QUEUE: # only for specific queues |
---|
| 1231 | # Fixme: assumes queue names don't contain single |
---|
| 1232 | # quote or comma. Don't know what the SGE rules are. |
---|
| 1233 | queues = " -q '" + string.join (QUEUE, ",") + "'" |
---|
| 1234 | # Note the comment in SgeQstatXMLParser about scaling with |
---|
| 1235 | # this method of getting data. I haven't found better one. |
---|
| 1236 | # Output with args `-xml -ext -f -r' is easier to parse |
---|
| 1237 | # in some ways, harder in others, but it doesn't provide |
---|
| 1238 | # the submission time (at least SGE 6.0). The pipeline |
---|
| 1239 | # into sed corrects bogus XML observed with a configuration |
---|
| 1240 | # of SGE 6.0u8, which otherwise causes the parsing to hang. |
---|
| 1241 | piping = popen2.Popen3("qstat -u '*' -j '*' -xml | \ |
---|
[623] | 1242 | sed -e 's/reported usage>/reported_usage>/g' -e 's;<\/*JATASK:.*>;;'" \ |
---|
[691] | 1243 | + queues, True) |
---|
| 1244 | qstatparser = SgeQstatXMLParser() |
---|
| 1245 | parse_err = 0 |
---|
| 1246 | try: |
---|
| 1247 | xml.sax.parse(piping.fromchild, qstatparser) |
---|
| 1248 | except NoJobs: |
---|
| 1249 | pass |
---|
| 1250 | except: |
---|
| 1251 | parse_err = 1 |
---|
| 1252 | if piping.wait(): |
---|
| 1253 | debug_msg(10, |
---|
| 1254 | "qstat error, skipping until next polling interval: " |
---|
| 1255 | + piping.childerr.readline()) |
---|
| 1256 | return None |
---|
| 1257 | elif parse_err: |
---|
| 1258 | debug_msg(10, "Bad XML output from qstat"()) |
---|
| 1259 | exit (1) |
---|
| 1260 | for f in piping.fromchild, piping.tochild, piping.childerr: |
---|
| 1261 | f.close() |
---|
| 1262 | self.cur_time = time.time() |
---|
| 1263 | jobs_processed = [] |
---|
| 1264 | for job in qstatparser.joblist: |
---|
| 1265 | job_id = job["number"] |
---|
| 1266 | if job["status"] in [ 'Q', 'R' ]: |
---|
| 1267 | jobs_processed.append(job_id) |
---|
| 1268 | if job["status"] == "R": |
---|
| 1269 | job["nodes"] = do_nodelist (job["nodes"]) |
---|
| 1270 | # Fixme: why is job["nodes"] sometimes null? |
---|
| 1271 | try: |
---|
| 1272 | # Fixme: Is this sensible? The |
---|
| 1273 | # PBS-type PPN isn't something you use |
---|
| 1274 | # with SGE. |
---|
| 1275 | job["ppn"] = float(job["slots"]) / \ |
---|
| 1276 | len(job["nodes"]) |
---|
| 1277 | except: |
---|
| 1278 | job["ppn"] = 0 |
---|
| 1279 | if DETECT_TIME_DIFFS: |
---|
| 1280 | # If a job start is later than our |
---|
| 1281 | # current date, that must mean |
---|
| 1282 | # the SGE server's time is later |
---|
| 1283 | # than our local time. |
---|
| 1284 | start_timestamp = \ |
---|
| 1285 | int (job["start_timestamp"]) |
---|
| 1286 | if start_timestamp > \ |
---|
| 1287 | int(self.cur_time) + \ |
---|
| 1288 | int(self.timeoffset): |
---|
[318] | 1289 | |
---|
[691] | 1290 | self.timeoffset = \ |
---|
| 1291 | start_timestamp - \ |
---|
| 1292 | int(self.cur_time) |
---|
| 1293 | else: |
---|
| 1294 | # fixme: Note sure what this should be: |
---|
| 1295 | job["ppn"] = job["RN_max"] |
---|
| 1296 | job["nodes"] = "1" |
---|
[318] | 1297 | |
---|
[691] | 1298 | myAttrs = {} |
---|
| 1299 | for attr in ["name", "queue", "owner", |
---|
| 1300 | "requested_time", "status", |
---|
| 1301 | "requested_memory", "ppn", |
---|
| 1302 | "start_timestamp", "queued_timestamp"]: |
---|
| 1303 | myAttrs[attr] = str(job[attr]) |
---|
| 1304 | myAttrs["nodes"] = job["nodes"] |
---|
| 1305 | myAttrs["reported"] = str(int(self.cur_time) + \ |
---|
| 1306 | int(self.timeoffset)) |
---|
| 1307 | myAttrs["domain"] = fqdn_parts(socket.getfqdn())[1] |
---|
| 1308 | myAttrs["poll_interval"] = str(BATCH_POLL_INTERVAL) |
---|
[318] | 1309 | |
---|
[691] | 1310 | if self.jobDataChanged(self.jobs, job_id, myAttrs) \ |
---|
| 1311 | and myAttrs["status"] in ["R", "Q"]: |
---|
| 1312 | self.jobs[job_id] = myAttrs |
---|
| 1313 | for id, attrs in self.jobs.items(): |
---|
| 1314 | if id not in jobs_processed: |
---|
| 1315 | del self.jobs[id] |
---|
[318] | 1316 | |
---|
[524] | 1317 | # LSF code by Mahmoud Hanafi <hanafim@users.sourceforge.nt> |
---|
| 1318 | # Requres LSFObject http://sourceforge.net/projects/lsfobject |
---|
| 1319 | # |
---|
| 1320 | class LsfDataGatherer(DataGatherer): |
---|
[525] | 1321 | |
---|
[691] | 1322 | """This is the DataGatherer for LSf""" |
---|
[524] | 1323 | |
---|
[691] | 1324 | global lsfObject |
---|
[524] | 1325 | |
---|
[691] | 1326 | def __init__( self ): |
---|
[525] | 1327 | |
---|
[691] | 1328 | self.jobs = { } |
---|
| 1329 | self.timeoffset = 0 |
---|
| 1330 | self.dp = DataProcessor() |
---|
| 1331 | self.initLsfQuery() |
---|
[524] | 1332 | |
---|
[691] | 1333 | def _countDuplicatesInList( self, dupedList ): |
---|
[525] | 1334 | |
---|
[691] | 1335 | countDupes = { } |
---|
[525] | 1336 | |
---|
[691] | 1337 | for item in dupedList: |
---|
[525] | 1338 | |
---|
[691] | 1339 | if not countDupes.has_key( item ): |
---|
[525] | 1340 | |
---|
[691] | 1341 | countDupes[ item ] = 1 |
---|
| 1342 | else: |
---|
| 1343 | countDupes[ item ] = countDupes[ item ] + 1 |
---|
[525] | 1344 | |
---|
[691] | 1345 | dupeCountList = [ ] |
---|
[525] | 1346 | |
---|
[691] | 1347 | for item, count in countDupes.items(): |
---|
[525] | 1348 | |
---|
[691] | 1349 | dupeCountList.append( ( item, count ) ) |
---|
[525] | 1350 | |
---|
[691] | 1351 | return dupeCountList |
---|
[524] | 1352 | # |
---|
| 1353 | #lst = ['I1','I2','I1','I3','I4','I4','I7','I7','I7','I7','I7'] |
---|
| 1354 | #print _countDuplicatesInList(lst) |
---|
| 1355 | #[('I1', 2), ('I3', 1), ('I2', 1), ('I4', 2), ('I7', 5)] |
---|
| 1356 | ######################## |
---|
| 1357 | |
---|
[691] | 1358 | def initLsfQuery( self ): |
---|
| 1359 | self.pq = None |
---|
| 1360 | self.pq = lsfObject.jobInfoEntObject() |
---|
[524] | 1361 | |
---|
[691] | 1362 | def getJobData( self, known_jobs="" ): |
---|
| 1363 | """Gather all data on current jobs in LSF""" |
---|
| 1364 | if len( known_jobs ) > 0: |
---|
| 1365 | jobs = known_jobs |
---|
| 1366 | else: |
---|
| 1367 | jobs = { } |
---|
| 1368 | joblist = {} |
---|
| 1369 | joblist = self.pq.getJobInfo() |
---|
| 1370 | nodelist = '' |
---|
[524] | 1371 | |
---|
[691] | 1372 | self.cur_time = time.time() |
---|
[524] | 1373 | |
---|
[691] | 1374 | jobs_processed = [ ] |
---|
[524] | 1375 | |
---|
[691] | 1376 | for name, attrs in joblist.items(): |
---|
| 1377 | job_id = str(name) |
---|
| 1378 | jobs_processed.append( job_id ) |
---|
| 1379 | name = self.getAttr( attrs, 'jobName' ) |
---|
| 1380 | queue = self.getAttr( self.getAttr( attrs, 'submit') , 'queue' ) |
---|
| 1381 | owner = self.getAttr( attrs, 'user' ) |
---|
[524] | 1382 | |
---|
| 1383 | ### THIS IS THE rLimit List index values |
---|
[691] | 1384 | #define LSF_RLIMIT_CPU 0 /* cpu time in milliseconds */ |
---|
| 1385 | #define LSF_RLIMIT_FSIZE 1 /* maximum file size */ |
---|
| 1386 | #define LSF_RLIMIT_DATA 2 /* data size */ |
---|
| 1387 | #define LSF_RLIMIT_STACK 3 /* stack size */ |
---|
| 1388 | #define LSF_RLIMIT_CORE 4 /* core file size */ |
---|
| 1389 | #define LSF_RLIMIT_RSS 5 /* resident set size */ |
---|
| 1390 | #define LSF_RLIMIT_NOFILE 6 /* open files */ |
---|
| 1391 | #define LSF_RLIMIT_OPEN_MAX 7 /* (from HP-UX) */ |
---|
| 1392 | #define LSF_RLIMIT_VMEM 8 /* maximum swap mem */ |
---|
[524] | 1393 | #define LSF_RLIMIT_SWAP 8 |
---|
[691] | 1394 | #define LSF_RLIMIT_RUN 9 /* max wall-clock time limit */ |
---|
| 1395 | #define LSF_RLIMIT_PROCESS 10 /* process number limit */ |
---|
| 1396 | #define LSF_RLIMIT_THREAD 11 /* thread number limit (introduced in LSF6.0) */ |
---|
| 1397 | #define LSF_RLIM_NLIMITS 12 /* number of resource limits */ |
---|
[524] | 1398 | |
---|
[691] | 1399 | requested_time = self.getAttr( self.getAttr( attrs, 'submit') , 'rLimits' )[9] |
---|
| 1400 | if requested_time == -1: |
---|
| 1401 | requested_time = "" |
---|
| 1402 | requested_memory = self.getAttr( self.getAttr( attrs, 'submit') , 'rLimits' )[8] |
---|
| 1403 | if requested_memory == -1: |
---|
| 1404 | requested_memory = "" |
---|
[524] | 1405 | # This tries to get proc per node. We don't support this right now |
---|
[691] | 1406 | ppn = 0 #self.getAttr( self.getAttr( attrs, 'SubmitList') , 'numProessors' ) |
---|
| 1407 | requested_cpus = self.getAttr( self.getAttr( attrs, 'submit') , 'numProcessors' ) |
---|
| 1408 | if requested_cpus == None or requested_cpus == "": |
---|
| 1409 | requested_cpus = 1 |
---|
[524] | 1410 | |
---|
[691] | 1411 | if QUEUE: |
---|
| 1412 | for q in QUEUE: |
---|
| 1413 | if q == queue: |
---|
| 1414 | display_queue = 1 |
---|
| 1415 | break |
---|
| 1416 | else: |
---|
| 1417 | display_queue = 0 |
---|
| 1418 | continue |
---|
| 1419 | if display_queue == 0: |
---|
| 1420 | continue |
---|
[524] | 1421 | |
---|
[691] | 1422 | runState = self.getAttr( attrs, 'status' ) |
---|
| 1423 | if runState == 4: |
---|
| 1424 | status = 'R' |
---|
| 1425 | else: |
---|
| 1426 | status = 'Q' |
---|
| 1427 | queued_timestamp = self.getAttr( attrs, 'submitTime' ) |
---|
[524] | 1428 | |
---|
[691] | 1429 | if status == 'R': |
---|
| 1430 | start_timestamp = self.getAttr( attrs, 'startTime' ) |
---|
| 1431 | nodesCpu = dict(self._countDuplicatesInList(self.getAttr( attrs, 'exHosts' ))) |
---|
| 1432 | nodelist = nodesCpu.keys() |
---|
[524] | 1433 | |
---|
[691] | 1434 | if DETECT_TIME_DIFFS: |
---|
[524] | 1435 | |
---|
[691] | 1436 | # If a job start if later than our current date, |
---|
| 1437 | # that must mean the Torque server's time is later |
---|
| 1438 | # than our local time. |
---|
[524] | 1439 | |
---|
[691] | 1440 | if int(start_timestamp) > int( int(self.cur_time) + int(self.timeoffset) ): |
---|
[524] | 1441 | |
---|
[691] | 1442 | self.timeoffset = int( int(start_timestamp) - int(self.cur_time) ) |
---|
[524] | 1443 | |
---|
[691] | 1444 | elif status == 'Q': |
---|
| 1445 | start_timestamp = '' |
---|
| 1446 | count_mynodes = 0 |
---|
| 1447 | numeric_node = 1 |
---|
| 1448 | nodelist = '' |
---|
[524] | 1449 | |
---|
[691] | 1450 | myAttrs = { } |
---|
| 1451 | if name == "": |
---|
| 1452 | myAttrs['name'] = "none" |
---|
| 1453 | else: |
---|
| 1454 | myAttrs['name'] = name |
---|
[524] | 1455 | |
---|
[691] | 1456 | myAttrs[ 'owner' ] = owner |
---|
| 1457 | myAttrs[ 'requested_time' ] = str(requested_time) |
---|
| 1458 | myAttrs[ 'requested_memory' ] = str(requested_memory) |
---|
| 1459 | myAttrs[ 'requested_cpus' ] = str(requested_cpus) |
---|
| 1460 | myAttrs[ 'ppn' ] = str( ppn ) |
---|
| 1461 | myAttrs[ 'status' ] = status |
---|
| 1462 | myAttrs[ 'start_timestamp' ] = str(start_timestamp) |
---|
| 1463 | myAttrs[ 'queue' ] = str(queue) |
---|
| 1464 | myAttrs[ 'queued_timestamp' ] = str(queued_timestamp) |
---|
| 1465 | myAttrs[ 'reported' ] = str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) |
---|
| 1466 | myAttrs[ 'nodes' ] = do_nodelist( nodelist ) |
---|
| 1467 | myAttrs[ 'domain' ] = fqdn_parts( socket.getfqdn() )[1] |
---|
| 1468 | myAttrs[ 'poll_interval' ] = str(BATCH_POLL_INTERVAL) |
---|
[524] | 1469 | |
---|
[691] | 1470 | if self.jobDataChanged( jobs, job_id, myAttrs ) and myAttrs['status'] in [ 'R', 'Q' ]: |
---|
| 1471 | jobs[ job_id ] = myAttrs |
---|
[524] | 1472 | |
---|
[691] | 1473 | debug_msg( 10, printTime() + ' job %s state changed' %(job_id) ) |
---|
[524] | 1474 | |
---|
[691] | 1475 | for id, attrs in jobs.items(): |
---|
| 1476 | if id not in jobs_processed: |
---|
| 1477 | # This one isn't there anymore |
---|
| 1478 | # |
---|
| 1479 | del jobs[ id ] |
---|
| 1480 | self.jobs=jobs |
---|
[524] | 1481 | |
---|
| 1482 | |
---|
[355] | 1483 | class PbsDataGatherer( DataGatherer ): |
---|
[318] | 1484 | |
---|
[691] | 1485 | """This is the DataGatherer for PBS and Torque""" |
---|
[318] | 1486 | |
---|
[691] | 1487 | global PBSQuery, PBSError |
---|
[256] | 1488 | |
---|
[691] | 1489 | def __init__( self ): |
---|
[354] | 1490 | |
---|
[691] | 1491 | """Setup appropriate variables""" |
---|
[23] | 1492 | |
---|
[691] | 1493 | self.jobs = { } |
---|
| 1494 | self.timeoffset = 0 |
---|
| 1495 | self.dp = DataProcessor() |
---|
[354] | 1496 | |
---|
[691] | 1497 | self.initPbsQuery() |
---|
[23] | 1498 | |
---|
[691] | 1499 | def initPbsQuery( self ): |
---|
[91] | 1500 | |
---|
[691] | 1501 | self.pq = None |
---|
[354] | 1502 | |
---|
[691] | 1503 | if( BATCH_SERVER ): |
---|
[354] | 1504 | |
---|
[691] | 1505 | self.pq = PBSQuery( BATCH_SERVER ) |
---|
| 1506 | else: |
---|
| 1507 | self.pq = PBSQuery() |
---|
[91] | 1508 | |
---|
[691] | 1509 | try: |
---|
| 1510 | self.pq.old_data_structure() |
---|
[656] | 1511 | |
---|
[691] | 1512 | except AttributeError: |
---|
[656] | 1513 | |
---|
[691] | 1514 | # pbs_query is older |
---|
| 1515 | # |
---|
| 1516 | pass |
---|
[656] | 1517 | |
---|
[691] | 1518 | def getJobData( self ): |
---|
[354] | 1519 | |
---|
[691] | 1520 | """Gather all data on current jobs in Torque""" |
---|
[26] | 1521 | |
---|
[691] | 1522 | joblist = {} |
---|
| 1523 | self.cur_time = 0 |
---|
[349] | 1524 | |
---|
[691] | 1525 | try: |
---|
| 1526 | joblist = self.pq.getjobs() |
---|
| 1527 | self.cur_time = time.time() |
---|
[354] | 1528 | |
---|
[691] | 1529 | except PBSError, detail: |
---|
[354] | 1530 | |
---|
[691] | 1531 | debug_msg( 10, "Caught PBS unavailable, skipping until next polling interval: " + str( detail ) ) |
---|
| 1532 | return None |
---|
[354] | 1533 | |
---|
[691] | 1534 | jobs_processed = [ ] |
---|
[26] | 1535 | |
---|
[691] | 1536 | for name, attrs in joblist.items(): |
---|
| 1537 | display_queue = 1 |
---|
| 1538 | job_id = name.split( '.' )[0] |
---|
[26] | 1539 | |
---|
[691] | 1540 | name = self.getAttr( attrs, 'Job_Name' ) |
---|
| 1541 | queue = self.getAttr( attrs, 'queue' ) |
---|
[317] | 1542 | |
---|
[691] | 1543 | if QUEUE: |
---|
| 1544 | for q in QUEUE: |
---|
| 1545 | if q == queue: |
---|
| 1546 | display_queue = 1 |
---|
| 1547 | break |
---|
| 1548 | else: |
---|
| 1549 | display_queue = 0 |
---|
| 1550 | continue |
---|
| 1551 | if display_queue == 0: |
---|
| 1552 | continue |
---|
[317] | 1553 | |
---|
| 1554 | |
---|
[691] | 1555 | owner = self.getAttr( attrs, 'Job_Owner' ).split( '@' )[0] |
---|
| 1556 | requested_time = self.getAttr( attrs, 'Resource_List.walltime' ) |
---|
| 1557 | requested_memory = self.getAttr( attrs, 'Resource_List.mem' ) |
---|
[95] | 1558 | |
---|
[691] | 1559 | mynoderequest = self.getAttr( attrs, 'Resource_List.nodes' ) |
---|
[95] | 1560 | |
---|
[691] | 1561 | ppn = '' |
---|
[281] | 1562 | |
---|
[691] | 1563 | if mynoderequest.find( ':' ) != -1 and mynoderequest.find( 'ppn' ) != -1: |
---|
[95] | 1564 | |
---|
[691] | 1565 | mynoderequest_fields = mynoderequest.split( ':' ) |
---|
[281] | 1566 | |
---|
[691] | 1567 | for mynoderequest_field in mynoderequest_fields: |
---|
[281] | 1568 | |
---|
[691] | 1569 | if mynoderequest_field.find( 'ppn' ) != -1: |
---|
[281] | 1570 | |
---|
[691] | 1571 | ppn = mynoderequest_field.split( 'ppn=' )[1] |
---|
[281] | 1572 | |
---|
[691] | 1573 | status = self.getAttr( attrs, 'job_state' ) |
---|
[25] | 1574 | |
---|
[691] | 1575 | if status in [ 'Q', 'R' ]: |
---|
[450] | 1576 | |
---|
[691] | 1577 | jobs_processed.append( job_id ) |
---|
[450] | 1578 | |
---|
[691] | 1579 | queued_timestamp = self.getAttr( attrs, 'ctime' ) |
---|
[243] | 1580 | |
---|
[691] | 1581 | if status == 'R': |
---|
[133] | 1582 | |
---|
[691] | 1583 | start_timestamp = self.getAttr( attrs, 'mtime' ) |
---|
| 1584 | nodes = self.getAttr( attrs, 'exec_host' ).split( '+' ) |
---|
[133] | 1585 | |
---|
[691] | 1586 | nodeslist = do_nodelist( nodes ) |
---|
[354] | 1587 | |
---|
[691] | 1588 | if DETECT_TIME_DIFFS: |
---|
[185] | 1589 | |
---|
[691] | 1590 | # If a job start if later than our current date, |
---|
| 1591 | # that must mean the Torque server's time is later |
---|
| 1592 | # than our local time. |
---|
| 1593 | |
---|
| 1594 | if int( start_timestamp ) > int( int( self.cur_time ) + int( self.timeoffset ) ): |
---|
[185] | 1595 | |
---|
[691] | 1596 | self.timeoffset = int( int(start_timestamp) - int(self.cur_time) ) |
---|
[185] | 1597 | |
---|
[691] | 1598 | elif status == 'Q': |
---|
[95] | 1599 | |
---|
[691] | 1600 | # 'mynodequest' can be a string in the following syntax according to the |
---|
| 1601 | # Torque Administator's manual: |
---|
| 1602 | # |
---|
| 1603 | # {<node_count> | <hostname>}[:ppn=<ppn>][:<property>[:<property>]...][+ ...] |
---|
| 1604 | # {<node_count> | <hostname>}[:ppn=<ppn>][:<property>[:<property>]...][+ ...] |
---|
| 1605 | # etc |
---|
| 1606 | # |
---|
[451] | 1607 | |
---|
[691] | 1608 | # |
---|
| 1609 | # For now we only count the amount of nodes request and ignore properties |
---|
| 1610 | # |
---|
[451] | 1611 | |
---|
[691] | 1612 | start_timestamp = '' |
---|
| 1613 | count_mynodes = 0 |
---|
[354] | 1614 | |
---|
[691] | 1615 | for node in mynoderequest.split( '+' ): |
---|
[67] | 1616 | |
---|
[691] | 1617 | # Just grab the {node_count|hostname} part and ignore properties |
---|
| 1618 | # |
---|
| 1619 | nodepart = node.split( ':' )[0] |
---|
[67] | 1620 | |
---|
[691] | 1621 | # Let's assume a node_count value |
---|
| 1622 | # |
---|
| 1623 | numeric_node = 1 |
---|
[451] | 1624 | |
---|
[691] | 1625 | # Chop the value up into characters |
---|
| 1626 | # |
---|
| 1627 | for letter in nodepart: |
---|
[67] | 1628 | |
---|
[691] | 1629 | # If this char is not a digit (0-9), this must be a hostname |
---|
| 1630 | # |
---|
| 1631 | if letter not in string.digits: |
---|
[133] | 1632 | |
---|
[691] | 1633 | numeric_node = 0 |
---|
[133] | 1634 | |
---|
[691] | 1635 | # If this is a hostname, just count this as one (1) node |
---|
| 1636 | # |
---|
| 1637 | if not numeric_node: |
---|
[354] | 1638 | |
---|
[691] | 1639 | count_mynodes = count_mynodes + 1 |
---|
| 1640 | else: |
---|
[451] | 1641 | |
---|
[691] | 1642 | # If this a number, it must be the node_count |
---|
| 1643 | # and increase our count with it's value |
---|
| 1644 | # |
---|
| 1645 | try: |
---|
| 1646 | count_mynodes = count_mynodes + int( nodepart ) |
---|
[354] | 1647 | |
---|
[691] | 1648 | except ValueError, detail: |
---|
[354] | 1649 | |
---|
[691] | 1650 | # When we arrive here I must be bugged or very confused |
---|
| 1651 | # THIS SHOULD NOT HAPPEN! |
---|
| 1652 | # |
---|
| 1653 | debug_msg( 10, str( detail ) ) |
---|
| 1654 | debug_msg( 10, "Encountered weird node in Resources_List?!" ) |
---|
| 1655 | debug_msg( 10, 'nodepart = ' + str( nodepart ) ) |
---|
| 1656 | debug_msg( 10, 'job = ' + str( name ) ) |
---|
| 1657 | debug_msg( 10, 'attrs = ' + str( attrs ) ) |
---|
| 1658 | |
---|
| 1659 | nodeslist = str( count_mynodes ) |
---|
| 1660 | else: |
---|
| 1661 | start_timestamp = '' |
---|
| 1662 | nodeslist = '' |
---|
[133] | 1663 | |
---|
[691] | 1664 | myAttrs = { } |
---|
[26] | 1665 | |
---|
[691] | 1666 | myAttrs[ 'name' ] = str( name ) |
---|
| 1667 | myAttrs[ 'queue' ] = str( queue ) |
---|
| 1668 | myAttrs[ 'owner' ] = str( owner ) |
---|
| 1669 | myAttrs[ 'requested_time' ] = str( requested_time ) |
---|
| 1670 | myAttrs[ 'requested_memory' ] = str( requested_memory ) |
---|
| 1671 | myAttrs[ 'ppn' ] = str( ppn ) |
---|
| 1672 | myAttrs[ 'status' ] = str( status ) |
---|
| 1673 | myAttrs[ 'start_timestamp' ] = str( start_timestamp ) |
---|
| 1674 | myAttrs[ 'queued_timestamp' ] = str( queued_timestamp ) |
---|
| 1675 | myAttrs[ 'reported' ] = str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) |
---|
| 1676 | myAttrs[ 'nodes' ] = nodeslist |
---|
| 1677 | myAttrs[ 'domain' ] = fqdn_parts( socket.getfqdn() )[1] |
---|
| 1678 | myAttrs[ 'poll_interval' ] = str( BATCH_POLL_INTERVAL ) |
---|
[354] | 1679 | |
---|
[691] | 1680 | if self.jobDataChanged( self.jobs, job_id, myAttrs ) and myAttrs['status'] in [ 'R', 'Q' ]: |
---|
[61] | 1681 | |
---|
[691] | 1682 | self.jobs[ job_id ] = myAttrs |
---|
[26] | 1683 | |
---|
[691] | 1684 | for id, attrs in self.jobs.items(): |
---|
[76] | 1685 | |
---|
[691] | 1686 | if id not in jobs_processed: |
---|
[76] | 1687 | |
---|
[691] | 1688 | # This one isn't there anymore; toedeledoki! |
---|
| 1689 | # |
---|
| 1690 | del self.jobs[ id ] |
---|
[76] | 1691 | |
---|
[362] | 1692 | GMETRIC_DEFAULT_TYPE = 'string' |
---|
| 1693 | GMETRIC_DEFAULT_HOST = '127.0.0.1' |
---|
| 1694 | GMETRIC_DEFAULT_PORT = '8649' |
---|
[700] | 1695 | GMETRIC_DEFAULT_UNITS = '' |
---|
[362] | 1696 | |
---|
| 1697 | class Gmetric: |
---|
| 1698 | |
---|
[691] | 1699 | global GMETRIC_DEFAULT_HOST, GMETRIC_DEFAULT_PORT |
---|
[362] | 1700 | |
---|
[700] | 1701 | slope = { 'zero' : 0, 'positive' : 1, 'negative' : 2, 'both' : 3, 'unspecified' : 4 } |
---|
| 1702 | type = ( '', 'string', 'uint16', 'int16', 'uint32', 'int32', 'float', 'double', 'timestamp' ) |
---|
| 1703 | protocol = ( 'udp', 'multicast' ) |
---|
[362] | 1704 | |
---|
[691] | 1705 | def __init__( self, host=GMETRIC_DEFAULT_HOST, port=GMETRIC_DEFAULT_PORT ): |
---|
[700] | 1706 | |
---|
[691] | 1707 | global GMETRIC_DEFAULT_TYPE |
---|
[362] | 1708 | |
---|
[691] | 1709 | self.prot = self.checkHostProtocol( host ) |
---|
[700] | 1710 | self.data_msg = xdrlib.Packer() |
---|
| 1711 | self.meta_msg = xdrlib.Packer() |
---|
[691] | 1712 | self.socket = socket.socket( socket.AF_INET, socket.SOCK_DGRAM ) |
---|
[362] | 1713 | |
---|
[691] | 1714 | if self.prot not in self.protocol: |
---|
[362] | 1715 | |
---|
[691] | 1716 | raise ValueError( "Protocol must be one of: " + str( self.protocol ) ) |
---|
[362] | 1717 | |
---|
[691] | 1718 | if self.prot == 'multicast': |
---|
[362] | 1719 | |
---|
[691] | 1720 | # Set multicast options |
---|
| 1721 | # |
---|
| 1722 | self.socket.setsockopt( socket.IPPROTO_IP, socket.IP_MULTICAST_TTL, 20 ) |
---|
[362] | 1723 | |
---|
[691] | 1724 | self.hostport = ( host, int( port ) ) |
---|
| 1725 | self.slopestr = 'both' |
---|
| 1726 | self.tmax = 60 |
---|
[362] | 1727 | |
---|
[691] | 1728 | def checkHostProtocol( self, ip ): |
---|
[362] | 1729 | |
---|
[691] | 1730 | """Detect if a ip adress is a multicast address""" |
---|
[471] | 1731 | |
---|
[691] | 1732 | MULTICAST_ADDRESS_MIN = ( "224", "0", "0", "0" ) |
---|
| 1733 | MULTICAST_ADDRESS_MAX = ( "239", "255", "255", "255" ) |
---|
[362] | 1734 | |
---|
[700] | 1735 | ip_fields = ip.split( '.' ) |
---|
[362] | 1736 | |
---|
[691] | 1737 | if ip_fields >= MULTICAST_ADDRESS_MIN and ip_fields <= MULTICAST_ADDRESS_MAX: |
---|
[362] | 1738 | |
---|
[691] | 1739 | return 'multicast' |
---|
| 1740 | else: |
---|
| 1741 | return 'udp' |
---|
[362] | 1742 | |
---|
[691] | 1743 | def send( self, name, value, dmax, typestr = '', units = '' ): |
---|
[362] | 1744 | |
---|
[691] | 1745 | if len( units ) == 0: |
---|
[700] | 1746 | units = GMETRIC_DEFAULT_UNITS |
---|
[471] | 1747 | |
---|
[691] | 1748 | if len( typestr ) == 0: |
---|
[700] | 1749 | typestr = GMETRIC_DEFAULT_TYPE |
---|
[362] | 1750 | |
---|
[700] | 1751 | (meta_msg, data_msg) = self.makexdr( name, value, typestr, units, self.slopestr, self.tmax, dmax ) |
---|
[409] | 1752 | |
---|
[700] | 1753 | meta_rt = self.socket.sendto( meta_msg, self.hostport ) |
---|
| 1754 | data_rt = self.socket.sendto( data_msg, self.hostport ) |
---|
[362] | 1755 | |
---|
[700] | 1756 | return ( meta_rt, data_rt ) |
---|
[362] | 1757 | |
---|
[700] | 1758 | def makexdr( self, name, value, typestr, unitstr, slopestr, tmax, dmax, group=None, spoof=None ): |
---|
| 1759 | |
---|
| 1760 | hostname = "unset" |
---|
| 1761 | |
---|
[691] | 1762 | if slopestr not in self.slope: |
---|
[362] | 1763 | |
---|
[691] | 1764 | raise ValueError( "Slope must be one of: " + str( self.slope.keys() ) ) |
---|
[362] | 1765 | |
---|
[691] | 1766 | if typestr not in self.type: |
---|
[362] | 1767 | |
---|
[691] | 1768 | raise ValueError( "Type must be one of: " + str( self.type ) ) |
---|
[362] | 1769 | |
---|
[691] | 1770 | if len( name ) == 0: |
---|
[362] | 1771 | |
---|
[691] | 1772 | raise ValueError( "Name must be non-empty" ) |
---|
[362] | 1773 | |
---|
[700] | 1774 | self.meta_msg.reset() |
---|
| 1775 | self.meta_msg.pack_int( 128 ) |
---|
[362] | 1776 | |
---|
[700] | 1777 | if not spoof: |
---|
| 1778 | self.meta_msg.pack_string( hostname ) |
---|
| 1779 | else: |
---|
| 1780 | self.meta_msg.pack_string( spoof ) |
---|
[362] | 1781 | |
---|
[700] | 1782 | self.meta_msg.pack_string( name ) |
---|
| 1783 | |
---|
| 1784 | if not spoof: |
---|
| 1785 | self.meta_msg.pack_int( 0 ) |
---|
| 1786 | else: |
---|
| 1787 | self.meta_msg.pack_int( 1 ) |
---|
| 1788 | |
---|
| 1789 | self.meta_msg.pack_string( typestr ) |
---|
| 1790 | self.meta_msg.pack_string( name ) |
---|
| 1791 | self.meta_msg.pack_string( unitstr ) |
---|
| 1792 | self.meta_msg.pack_int( self.slope[ slopestr ] ) |
---|
| 1793 | self.meta_msg.pack_uint( int( tmax ) ) |
---|
| 1794 | self.meta_msg.pack_uint( int( dmax ) ) |
---|
| 1795 | |
---|
| 1796 | if not group: |
---|
| 1797 | self.meta_msg.pack_int( 0 ) |
---|
| 1798 | else: |
---|
| 1799 | self.meta_msg.pack_int( 1 ) |
---|
| 1800 | self.meta_msg.pack_string( "GROUP" ) |
---|
| 1801 | self.meta_msg.pack_string( group ) |
---|
| 1802 | |
---|
| 1803 | self.data_msg.reset() |
---|
| 1804 | self.data_msg.pack_int( 128+5 ) |
---|
| 1805 | |
---|
| 1806 | if not spoof: |
---|
| 1807 | self.data_msg.pack_string( hostname ) |
---|
| 1808 | else: |
---|
| 1809 | self.data_msg.pack_string( spoof ) |
---|
| 1810 | |
---|
| 1811 | self.data_msg.pack_string( name ) |
---|
| 1812 | |
---|
| 1813 | if not spoof: |
---|
| 1814 | self.data_msg.pack_int( 0 ) |
---|
| 1815 | else: |
---|
| 1816 | self.data_msg.pack_int( 1 ) |
---|
| 1817 | |
---|
| 1818 | self.data_msg.pack_string( "%s" ) |
---|
| 1819 | self.data_msg.pack_string( str( value ) ) |
---|
| 1820 | |
---|
| 1821 | return ( self.meta_msg.get_buffer(), self.data_msg.get_buffer() ) |
---|
| 1822 | |
---|
[26] | 1823 | def printTime( ): |
---|
[354] | 1824 | |
---|
[691] | 1825 | """Print current time/date in human readable format for log/debug""" |
---|
[26] | 1826 | |
---|
[691] | 1827 | return time.strftime("%a, %d %b %Y %H:%M:%S") |
---|
[26] | 1828 | |
---|
| 1829 | def debug_msg( level, msg ): |
---|
[354] | 1830 | |
---|
[691] | 1831 | """Print msg if at or above current debug level""" |
---|
[26] | 1832 | |
---|
[691] | 1833 | global DAEMONIZE, DEBUG_LEVEL, SYSLOG_LEVEL |
---|
[377] | 1834 | |
---|
[691] | 1835 | if (not DAEMONIZE and DEBUG_LEVEL >= level): |
---|
| 1836 | sys.stderr.write( msg + '\n' ) |
---|
[26] | 1837 | |
---|
[691] | 1838 | if (DAEMONIZE and USE_SYSLOG and SYSLOG_LEVEL >= level): |
---|
| 1839 | syslog.syslog( msg ) |
---|
[373] | 1840 | |
---|
[307] | 1841 | def write_pidfile(): |
---|
| 1842 | |
---|
[691] | 1843 | # Write pidfile if PIDFILE is set |
---|
| 1844 | # |
---|
| 1845 | if PIDFILE: |
---|
[307] | 1846 | |
---|
[691] | 1847 | pid = os.getpid() |
---|
[354] | 1848 | |
---|
[691] | 1849 | pidfile = open( PIDFILE, 'w' ) |
---|
[354] | 1850 | |
---|
[691] | 1851 | pidfile.write( str( pid ) ) |
---|
| 1852 | pidfile.close() |
---|
[307] | 1853 | |
---|
[23] | 1854 | def main(): |
---|
[354] | 1855 | |
---|
[691] | 1856 | """Application start""" |
---|
[23] | 1857 | |
---|
[691] | 1858 | global PBSQuery, PBSError, lsfObject |
---|
| 1859 | global SYSLOG_FACILITY, USE_SYSLOG, BATCH_API, DAEMONIZE |
---|
[256] | 1860 | |
---|
[691] | 1861 | if not processArgs( sys.argv[1:] ): |
---|
[354] | 1862 | |
---|
[691] | 1863 | sys.exit( 1 ) |
---|
[212] | 1864 | |
---|
[691] | 1865 | # Load appropriate DataGatherer depending on which BATCH_API is set |
---|
| 1866 | # and any required modules for the Gatherer |
---|
| 1867 | # |
---|
| 1868 | if BATCH_API == 'pbs': |
---|
[256] | 1869 | |
---|
[691] | 1870 | try: |
---|
| 1871 | from PBSQuery import PBSQuery, PBSError |
---|
[256] | 1872 | |
---|
[691] | 1873 | except ImportError: |
---|
[256] | 1874 | |
---|
[691] | 1875 | debug_msg( 0, "FATAL ERROR: BATCH_API set to 'pbs' but python module 'pbs_python' is not installed" ) |
---|
| 1876 | sys.exit( 1 ) |
---|
[256] | 1877 | |
---|
[691] | 1878 | gather = PbsDataGatherer() |
---|
[256] | 1879 | |
---|
[691] | 1880 | elif BATCH_API == 'sge': |
---|
[256] | 1881 | |
---|
[691] | 1882 | # Tested with SGE 6.0u11. |
---|
| 1883 | # |
---|
| 1884 | gather = SgeDataGatherer() |
---|
[368] | 1885 | |
---|
[691] | 1886 | elif BATCH_API == 'lsf': |
---|
[368] | 1887 | |
---|
[691] | 1888 | try: |
---|
| 1889 | from lsfObject import lsfObject |
---|
| 1890 | except: |
---|
| 1891 | debug_msg(0, "fatal error: BATCH_API set to 'lsf' but python module is not found or installed") |
---|
| 1892 | sys.exit( 1) |
---|
[256] | 1893 | |
---|
[691] | 1894 | gather = LsfDataGatherer() |
---|
[524] | 1895 | |
---|
[691] | 1896 | else: |
---|
| 1897 | debug_msg( 0, "FATAL ERROR: unknown BATCH_API '" + BATCH_API + "' is not supported" ) |
---|
[354] | 1898 | |
---|
[691] | 1899 | sys.exit( 1 ) |
---|
[256] | 1900 | |
---|
[691] | 1901 | if( DAEMONIZE and USE_SYSLOG ): |
---|
[373] | 1902 | |
---|
[691] | 1903 | syslog.openlog( 'jobmond', syslog.LOG_NOWAIT, SYSLOG_FACILITY ) |
---|
[373] | 1904 | |
---|
[691] | 1905 | if DAEMONIZE: |
---|
[354] | 1906 | |
---|
[691] | 1907 | gather.daemon() |
---|
| 1908 | else: |
---|
| 1909 | gather.run() |
---|
[23] | 1910 | |
---|
[256] | 1911 | # wh00t? someone started me! :) |
---|
[65] | 1912 | # |
---|
[23] | 1913 | if __name__ == '__main__': |
---|
[691] | 1914 | main() |
---|