[23] | 1 | #!/usr/bin/env python |
---|
[225] | 2 | # |
---|
| 3 | # This file is part of Jobmonarch |
---|
| 4 | # |
---|
[363] | 5 | # Copyright (C) 2006-2007 Ramon Bastiaans |
---|
[622] | 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 661 2012-09-04 09:23:16Z ramonb $ |
---|
[227] | 23 | # |
---|
[23] | 24 | |
---|
[471] | 25 | import sys, getopt, ConfigParser, time, os, socket, string, re |
---|
| 26 | import xdrlib, socket, syslog, xml, xml.sax |
---|
[318] | 27 | from xml.sax.handler import feature_namespaces |
---|
[622] | 28 | from collections import deque |
---|
[660] | 29 | from types import * |
---|
[318] | 30 | |
---|
[500] | 31 | VERSION='0.3.1' |
---|
[307] | 32 | |
---|
[471] | 33 | def usage( ver ): |
---|
| 34 | |
---|
[659] | 35 | print 'jobmond %s' %VERSION |
---|
[471] | 36 | |
---|
[659] | 37 | if ver: |
---|
| 38 | return 0 |
---|
[471] | 39 | |
---|
[659] | 40 | print |
---|
| 41 | print 'Purpose:' |
---|
| 42 | print ' The Job Monitoring Daemon (jobmond) reports batch jobs information and statistics' |
---|
| 43 | print ' to Ganglia, which can be viewed with Job Monarch web frontend' |
---|
| 44 | print |
---|
| 45 | print 'Usage: jobmond [OPTIONS]' |
---|
| 46 | print |
---|
| 47 | print ' -c, --config=FILE The configuration file to use (default: /etc/jobmond.conf)' |
---|
| 48 | print ' -p, --pidfile=FILE Use pid file to store the process id' |
---|
| 49 | print ' -h, --help Print help and exit' |
---|
| 50 | print ' -v, --version Print version and exit' |
---|
| 51 | print |
---|
[307] | 52 | |
---|
[212] | 53 | def processArgs( args ): |
---|
[26] | 54 | |
---|
[659] | 55 | SHORT_L = 'p:hvc:' |
---|
| 56 | LONG_L = [ 'help', 'config=', 'pidfile=', 'version' ] |
---|
[165] | 57 | |
---|
[659] | 58 | global PIDFILE |
---|
| 59 | PIDFILE = None |
---|
[61] | 60 | |
---|
[659] | 61 | config_filename = '/etc/jobmond.conf' |
---|
[354] | 62 | |
---|
[659] | 63 | try: |
---|
[68] | 64 | |
---|
[659] | 65 | opts, args = getopt.getopt( args, SHORT_L, LONG_L ) |
---|
[185] | 66 | |
---|
[659] | 67 | except getopt.GetoptError, detail: |
---|
[212] | 68 | |
---|
[659] | 69 | print detail |
---|
| 70 | usage() |
---|
| 71 | sys.exit( 1 ) |
---|
[212] | 72 | |
---|
[659] | 73 | for opt, value in opts: |
---|
[212] | 74 | |
---|
[659] | 75 | if opt in [ '--config', '-c' ]: |
---|
| 76 | |
---|
| 77 | config_filename = value |
---|
[212] | 78 | |
---|
[659] | 79 | if opt in [ '--pidfile', '-p' ]: |
---|
[212] | 80 | |
---|
[659] | 81 | PIDFILE = value |
---|
| 82 | |
---|
| 83 | if opt in [ '--help', '-h' ]: |
---|
[307] | 84 | |
---|
[659] | 85 | usage( False ) |
---|
| 86 | sys.exit( 0 ) |
---|
[212] | 87 | |
---|
[659] | 88 | if opt in [ '--version', '-v' ]: |
---|
[471] | 89 | |
---|
[659] | 90 | usage( True ) |
---|
| 91 | sys.exit( 0 ) |
---|
[471] | 92 | |
---|
[659] | 93 | return loadConfig( config_filename ) |
---|
[212] | 94 | |
---|
[622] | 95 | # Fixme: This doesn't DTRT with commented-out bits of the file. E.g. |
---|
| 96 | # it picked up a commented-out `mcast_join' and tried to use a |
---|
| 97 | # multicast channel when it shouldn't have done. |
---|
[520] | 98 | class GangliaConfigParser: |
---|
| 99 | |
---|
[659] | 100 | def __init__( self, config_file ): |
---|
[520] | 101 | |
---|
[659] | 102 | self.config_file = config_file |
---|
[520] | 103 | |
---|
[659] | 104 | if not os.path.exists( self.config_file ): |
---|
[520] | 105 | |
---|
[659] | 106 | debug_msg( 0, "FATAL ERROR: gmond config '" + self.config_file + "' not found!" ) |
---|
| 107 | sys.exit( 1 ) |
---|
[520] | 108 | |
---|
[659] | 109 | def removeQuotes( self, value ): |
---|
[520] | 110 | |
---|
[659] | 111 | clean_value = value |
---|
| 112 | clean_value = clean_value.replace( "'", "" ) |
---|
| 113 | clean_value = clean_value.replace( '"', '' ) |
---|
| 114 | clean_value = clean_value.strip() |
---|
[520] | 115 | |
---|
[659] | 116 | return clean_value |
---|
[520] | 117 | |
---|
[659] | 118 | def getVal( self, section, valname ): |
---|
[520] | 119 | |
---|
[659] | 120 | cfg_fp = open( self.config_file ) |
---|
| 121 | section_start = False |
---|
| 122 | section_found = False |
---|
| 123 | value = None |
---|
[520] | 124 | |
---|
[659] | 125 | for line in cfg_fp.readlines(): |
---|
[520] | 126 | |
---|
[659] | 127 | if line.find( section ) != -1: |
---|
[520] | 128 | |
---|
[659] | 129 | section_found = True |
---|
[520] | 130 | |
---|
[659] | 131 | if line.find( '{' ) != -1 and section_found: |
---|
[520] | 132 | |
---|
[659] | 133 | section_start = True |
---|
[520] | 134 | |
---|
[659] | 135 | if line.find( '}' ) != -1 and section_found: |
---|
[520] | 136 | |
---|
[659] | 137 | section_start = False |
---|
| 138 | section_found = False |
---|
[520] | 139 | |
---|
[659] | 140 | if line.find( valname ) != -1 and section_start: |
---|
[520] | 141 | |
---|
[659] | 142 | value = string.join( line.split( '=' )[1:], '' ).strip() |
---|
[520] | 143 | |
---|
[659] | 144 | cfg_fp.close() |
---|
[520] | 145 | |
---|
[659] | 146 | return value |
---|
[520] | 147 | |
---|
[659] | 148 | def getInt( self, section, valname ): |
---|
[520] | 149 | |
---|
[659] | 150 | value = self.getVal( section, valname ) |
---|
[520] | 151 | |
---|
[659] | 152 | if not value: |
---|
| 153 | return False |
---|
[520] | 154 | |
---|
[659] | 155 | value = self.removeQuotes( value ) |
---|
[520] | 156 | |
---|
[659] | 157 | return int( value ) |
---|
[520] | 158 | |
---|
[659] | 159 | def getStr( self, section, valname ): |
---|
[520] | 160 | |
---|
[659] | 161 | value = self.getVal( section, valname ) |
---|
[520] | 162 | |
---|
[659] | 163 | if not value: |
---|
| 164 | return False |
---|
[520] | 165 | |
---|
[659] | 166 | value = self.removeQuotes( value ) |
---|
[520] | 167 | |
---|
[659] | 168 | return str( value ) |
---|
[520] | 169 | |
---|
| 170 | def findGmetric(): |
---|
| 171 | |
---|
[659] | 172 | for dir in os.path.expandvars( '$PATH' ).split( ':' ): |
---|
[520] | 173 | |
---|
[659] | 174 | guess = '%s/%s' %( dir, 'gmetric' ) |
---|
[520] | 175 | |
---|
[659] | 176 | if os.path.exists( guess ): |
---|
[520] | 177 | |
---|
[659] | 178 | return guess |
---|
[520] | 179 | |
---|
[659] | 180 | return False |
---|
[520] | 181 | |
---|
[212] | 182 | def loadConfig( filename ): |
---|
| 183 | |
---|
[660] | 184 | def getlist( cfg_string ): |
---|
[215] | 185 | |
---|
[660] | 186 | my_list = [ ] |
---|
[215] | 187 | |
---|
[660] | 188 | for item_txt in cfg_string.split( ',' ): |
---|
[215] | 189 | |
---|
[660] | 190 | sep_char = None |
---|
[215] | 191 | |
---|
[660] | 192 | item_txt = item_txt.strip() |
---|
[215] | 193 | |
---|
[660] | 194 | for s_char in [ "'", '"' ]: |
---|
[215] | 195 | |
---|
[660] | 196 | if item_txt.find( s_char ) != -1: |
---|
[215] | 197 | |
---|
[660] | 198 | if item_txt.count( s_char ) != 2: |
---|
[215] | 199 | |
---|
[660] | 200 | print 'Missing quote: %s' %item_txt |
---|
| 201 | sys.exit( 1 ) |
---|
[215] | 202 | |
---|
[660] | 203 | else: |
---|
[215] | 204 | |
---|
[660] | 205 | sep_char = s_char |
---|
| 206 | break |
---|
[215] | 207 | |
---|
[660] | 208 | if sep_char: |
---|
[215] | 209 | |
---|
[660] | 210 | item_txt = item_txt.split( sep_char )[1] |
---|
[215] | 211 | |
---|
[660] | 212 | my_list.append( item_txt ) |
---|
[215] | 213 | |
---|
[660] | 214 | return my_list |
---|
[215] | 215 | |
---|
[659] | 216 | cfg = ConfigParser.ConfigParser() |
---|
[212] | 217 | |
---|
[659] | 218 | cfg.read( filename ) |
---|
[212] | 219 | |
---|
[659] | 220 | global DEBUG_LEVEL, DAEMONIZE, BATCH_SERVER, BATCH_POLL_INTERVAL |
---|
| 221 | global GMOND_CONF, DETECT_TIME_DIFFS, BATCH_HOST_TRANSLATE |
---|
| 222 | global BATCH_API, QUEUE, GMETRIC_TARGET, USE_SYSLOG |
---|
| 223 | global SYSLOG_LEVEL, SYSLOG_FACILITY, GMETRIC_BINARY |
---|
[212] | 224 | |
---|
[659] | 225 | DEBUG_LEVEL = cfg.getint( 'DEFAULT', 'DEBUG_LEVEL' ) |
---|
[212] | 226 | |
---|
[659] | 227 | DAEMONIZE = cfg.getboolean( 'DEFAULT', 'DAEMONIZE' ) |
---|
[212] | 228 | |
---|
[659] | 229 | SYSLOG_LEVEL = -1 |
---|
| 230 | SYSLOG_FACILITY = None |
---|
[377] | 231 | |
---|
[659] | 232 | try: |
---|
| 233 | USE_SYSLOG = cfg.getboolean( 'DEFAULT', 'USE_SYSLOG' ) |
---|
[212] | 234 | |
---|
[659] | 235 | except ConfigParser.NoOptionError: |
---|
[373] | 236 | |
---|
[659] | 237 | USE_SYSLOG = True |
---|
[373] | 238 | |
---|
[659] | 239 | debug_msg( 0, 'ERROR: no option USE_SYSLOG found: assuming yes' ) |
---|
[373] | 240 | |
---|
[659] | 241 | if USE_SYSLOG: |
---|
[373] | 242 | |
---|
[659] | 243 | try: |
---|
| 244 | SYSLOG_LEVEL = cfg.getint( 'DEFAULT', 'SYSLOG_LEVEL' ) |
---|
[373] | 245 | |
---|
[659] | 246 | except ConfigParser.NoOptionError: |
---|
[373] | 247 | |
---|
[659] | 248 | debug_msg( 0, 'ERROR: no option SYSLOG_LEVEL found: assuming level 0' ) |
---|
| 249 | SYSLOG_LEVEL = 0 |
---|
[373] | 250 | |
---|
[659] | 251 | try: |
---|
[373] | 252 | |
---|
[659] | 253 | SYSLOG_FACILITY = eval( 'syslog.LOG_' + cfg.get( 'DEFAULT', 'SYSLOG_FACILITY' ) ) |
---|
[373] | 254 | |
---|
[659] | 255 | except ConfigParser.NoOptionError: |
---|
[373] | 256 | |
---|
[659] | 257 | SYSLOG_FACILITY = syslog.LOG_DAEMON |
---|
[373] | 258 | |
---|
[659] | 259 | debug_msg( 0, 'ERROR: no option SYSLOG_FACILITY found: assuming facility DAEMON' ) |
---|
[373] | 260 | |
---|
[659] | 261 | try: |
---|
[373] | 262 | |
---|
[659] | 263 | BATCH_SERVER = cfg.get( 'DEFAULT', 'BATCH_SERVER' ) |
---|
[212] | 264 | |
---|
[659] | 265 | except ConfigParser.NoOptionError: |
---|
[265] | 266 | |
---|
[659] | 267 | # Backwards compatibility for old configs |
---|
| 268 | # |
---|
[265] | 269 | |
---|
[659] | 270 | BATCH_SERVER = cfg.get( 'DEFAULT', 'TORQUE_SERVER' ) |
---|
| 271 | api_guess = 'pbs' |
---|
| 272 | |
---|
| 273 | try: |
---|
| 274 | |
---|
| 275 | BATCH_POLL_INTERVAL = cfg.getint( 'DEFAULT', 'BATCH_POLL_INTERVAL' ) |
---|
[265] | 276 | |
---|
[659] | 277 | except ConfigParser.NoOptionError: |
---|
[265] | 278 | |
---|
[659] | 279 | # Backwards compatibility for old configs |
---|
| 280 | # |
---|
[265] | 281 | |
---|
[659] | 282 | BATCH_POLL_INTERVAL = cfg.getint( 'DEFAULT', 'TORQUE_POLL_INTERVAL' ) |
---|
| 283 | api_guess = 'pbs' |
---|
| 284 | |
---|
| 285 | try: |
---|
[212] | 286 | |
---|
[659] | 287 | GMOND_CONF = cfg.get( 'DEFAULT', 'GMOND_CONF' ) |
---|
[353] | 288 | |
---|
[659] | 289 | except ConfigParser.NoOptionError: |
---|
[353] | 290 | |
---|
[661] | 291 | # Not specified: assume /etc/ganglia/gmond.conf |
---|
[659] | 292 | # |
---|
[661] | 293 | GMOND_CONF = '/etc/ganglia/gmond.conf' |
---|
[353] | 294 | |
---|
[659] | 295 | ganglia_cfg = GangliaConfigParser( GMOND_CONF ) |
---|
[449] | 296 | |
---|
[659] | 297 | # Let's try to find the GMETRIC_TARGET ourselves first from GMOND_CONF |
---|
| 298 | # |
---|
| 299 | gmetric_dest_ip = ganglia_cfg.getStr( 'udp_send_channel', 'mcast_join' ) |
---|
[449] | 300 | |
---|
[659] | 301 | if not gmetric_dest_ip: |
---|
[449] | 302 | |
---|
[659] | 303 | # Maybe unicast target then |
---|
| 304 | # |
---|
| 305 | gmetric_dest_ip = ganglia_cfg.getStr( 'udp_send_channel', 'host' ) |
---|
[449] | 306 | |
---|
[660] | 307 | gmetric_dest_port = ganglia_cfg.getStr( 'udp_send_channel', 'port' ) |
---|
[520] | 308 | |
---|
[659] | 309 | if gmetric_dest_ip and gmetric_dest_port: |
---|
[520] | 310 | |
---|
[659] | 311 | GMETRIC_TARGET = '%s:%s' %( gmetric_dest_ip, gmetric_dest_port ) |
---|
| 312 | else: |
---|
[520] | 313 | |
---|
[659] | 314 | debug_msg( 0, "WARNING: Can't parse udp_send_channel from: '%s'" %GMOND_CONF ) |
---|
[520] | 315 | |
---|
[659] | 316 | # Couldn't figure it out: let's see if it's in our jobmond.conf |
---|
| 317 | # |
---|
| 318 | try: |
---|
[520] | 319 | |
---|
[659] | 320 | GMETRIC_TARGET = cfg.get( 'DEFAULT', 'GMETRIC_TARGET' ) |
---|
[520] | 321 | |
---|
[659] | 322 | # Guess not: now just give up |
---|
| 323 | # |
---|
| 324 | except ConfigParser.NoOptionError: |
---|
[520] | 325 | |
---|
[659] | 326 | GMETRIC_TARGET = None |
---|
[520] | 327 | |
---|
[659] | 328 | 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] | 329 | |
---|
[659] | 330 | gmetric_bin = findGmetric() |
---|
[520] | 331 | |
---|
[659] | 332 | if gmetric_bin: |
---|
[520] | 333 | |
---|
[659] | 334 | GMETRIC_BINARY = gmetric_bin |
---|
| 335 | else: |
---|
| 336 | debug_msg( 0, "WARNING: Can't find gmetric binary anywhere in $PATH" ) |
---|
[520] | 337 | |
---|
[659] | 338 | try: |
---|
[520] | 339 | |
---|
[659] | 340 | GMETRIC_BINARY = cfg.get( 'DEFAULT', 'GMETRIC_BINARY' ) |
---|
[520] | 341 | |
---|
[659] | 342 | except ConfigParser.NoOptionError: |
---|
[520] | 343 | |
---|
[659] | 344 | debug_msg( 0, "FATAL ERROR: GMETRIC_BINARY not set and not in $PATH" ) |
---|
| 345 | sys.exit( 1 ) |
---|
[520] | 346 | |
---|
[659] | 347 | DETECT_TIME_DIFFS = cfg.getboolean( 'DEFAULT', 'DETECT_TIME_DIFFS' ) |
---|
[212] | 348 | |
---|
[659] | 349 | BATCH_HOST_TRANSLATE = getlist( cfg.get( 'DEFAULT', 'BATCH_HOST_TRANSLATE' ) ) |
---|
[215] | 350 | |
---|
[659] | 351 | try: |
---|
[256] | 352 | |
---|
[659] | 353 | BATCH_API = cfg.get( 'DEFAULT', 'BATCH_API' ) |
---|
[266] | 354 | |
---|
[659] | 355 | except ConfigParser.NoOptionError, detail: |
---|
[266] | 356 | |
---|
[659] | 357 | if BATCH_SERVER and api_guess: |
---|
[354] | 358 | |
---|
[659] | 359 | BATCH_API = api_guess |
---|
| 360 | else: |
---|
| 361 | debug_msg( 0, "FATAL ERROR: BATCH_API not set and can't make guess" ) |
---|
| 362 | sys.exit( 1 ) |
---|
[317] | 363 | |
---|
[659] | 364 | try: |
---|
[317] | 365 | |
---|
[659] | 366 | QUEUE = getlist( cfg.get( 'DEFAULT', 'QUEUE' ) ) |
---|
[317] | 367 | |
---|
[659] | 368 | except ConfigParser.NoOptionError, detail: |
---|
[317] | 369 | |
---|
[659] | 370 | QUEUE = None |
---|
[353] | 371 | |
---|
[659] | 372 | return True |
---|
[212] | 373 | |
---|
[507] | 374 | def fqdn_parts (fqdn): |
---|
[520] | 375 | |
---|
[659] | 376 | """Return pair of host and domain for fully-qualified domain name arg.""" |
---|
[520] | 377 | |
---|
[659] | 378 | parts = fqdn.split (".") |
---|
[520] | 379 | |
---|
[659] | 380 | return (parts[0], string.join(parts[1:], ".")) |
---|
[507] | 381 | |
---|
[253] | 382 | METRIC_MAX_VAL_LEN = 900 |
---|
| 383 | |
---|
[61] | 384 | class DataProcessor: |
---|
[355] | 385 | |
---|
[659] | 386 | """Class for processing of data""" |
---|
[61] | 387 | |
---|
[659] | 388 | binary = None |
---|
[61] | 389 | |
---|
[659] | 390 | def __init__( self, binary=None ): |
---|
[355] | 391 | |
---|
[659] | 392 | """Remember alternate binary location if supplied""" |
---|
[61] | 393 | |
---|
[659] | 394 | global GMETRIC_BINARY |
---|
[449] | 395 | |
---|
[659] | 396 | if binary: |
---|
| 397 | self.binary = binary |
---|
[61] | 398 | |
---|
[659] | 399 | if not self.binary: |
---|
| 400 | self.binary = GMETRIC_BINARY |
---|
[449] | 401 | |
---|
[659] | 402 | # Timeout for XML |
---|
| 403 | # |
---|
| 404 | # From ganglia's documentation: |
---|
| 405 | # |
---|
| 406 | # 'A metric will be deleted DMAX seconds after it is received, and |
---|
| 407 | # DMAX=0 means eternal life.' |
---|
[61] | 408 | |
---|
[659] | 409 | self.dmax = str( int( int( BATCH_POLL_INTERVAL ) * 2 ) ) |
---|
[80] | 410 | |
---|
[659] | 411 | if GMOND_CONF: |
---|
[354] | 412 | |
---|
[659] | 413 | incompatible = self.checkGmetricVersion() |
---|
[61] | 414 | |
---|
[659] | 415 | if incompatible: |
---|
[355] | 416 | |
---|
[659] | 417 | debug_msg( 0, 'Gmetric version not compatible, please upgrade to at least 3.0.1' ) |
---|
| 418 | sys.exit( 1 ) |
---|
[65] | 419 | |
---|
[659] | 420 | def checkGmetricVersion( self ): |
---|
[355] | 421 | |
---|
[659] | 422 | """ |
---|
| 423 | Check version of gmetric is at least 3.0.1 |
---|
| 424 | for the syntax we use |
---|
| 425 | """ |
---|
[65] | 426 | |
---|
[659] | 427 | global METRIC_MAX_VAL_LEN |
---|
[255] | 428 | |
---|
[659] | 429 | incompatible = 0 |
---|
[341] | 430 | |
---|
[659] | 431 | gfp = os.popen( self.binary + ' --version' ) |
---|
| 432 | lines = gfp.readlines() |
---|
[65] | 433 | |
---|
[659] | 434 | gfp.close() |
---|
[355] | 435 | |
---|
[659] | 436 | for line in lines: |
---|
[355] | 437 | |
---|
[659] | 438 | line = line.split( ' ' ) |
---|
[65] | 439 | |
---|
[659] | 440 | if len( line ) == 2 and str( line ).find( 'gmetric' ) != -1: |
---|
| 441 | |
---|
| 442 | gmetric_version = line[1].split( '\n' )[0] |
---|
[65] | 443 | |
---|
[659] | 444 | version_major = int( gmetric_version.split( '.' )[0] ) |
---|
| 445 | version_minor = int( gmetric_version.split( '.' )[1] ) |
---|
| 446 | version_patch = int( gmetric_version.split( '.' )[2] ) |
---|
[65] | 447 | |
---|
[659] | 448 | incompatible = 0 |
---|
[65] | 449 | |
---|
[659] | 450 | if version_major < 3: |
---|
[65] | 451 | |
---|
[659] | 452 | incompatible = 1 |
---|
| 453 | |
---|
| 454 | elif version_major == 3: |
---|
[65] | 455 | |
---|
[659] | 456 | if version_minor == 0: |
---|
[65] | 457 | |
---|
[659] | 458 | if version_patch < 1: |
---|
| 459 | |
---|
| 460 | incompatible = 1 |
---|
[65] | 461 | |
---|
[659] | 462 | # Gmetric 3.0.1 >< 3.0.3 had a bug in the max metric length |
---|
| 463 | # |
---|
| 464 | if version_patch < 3: |
---|
[255] | 465 | |
---|
[659] | 466 | METRIC_MAX_VAL_LEN = 900 |
---|
[255] | 467 | |
---|
[659] | 468 | elif version_patch >= 3: |
---|
[255] | 469 | |
---|
[659] | 470 | METRIC_MAX_VAL_LEN = 1400 |
---|
[255] | 471 | |
---|
[659] | 472 | return incompatible |
---|
[65] | 473 | |
---|
[659] | 474 | def multicastGmetric( self, metricname, metricval, valtype='string', units='' ): |
---|
[355] | 475 | |
---|
[659] | 476 | """Call gmetric binary and multicast""" |
---|
[65] | 477 | |
---|
[659] | 478 | cmd = self.binary |
---|
[65] | 479 | |
---|
[659] | 480 | if GMETRIC_TARGET: |
---|
[61] | 481 | |
---|
[659] | 482 | GMETRIC_TARGET_HOST = GMETRIC_TARGET.split( ':' )[0] |
---|
| 483 | GMETRIC_TARGET_PORT = GMETRIC_TARGET.split( ':' )[1] |
---|
[353] | 484 | |
---|
[659] | 485 | metric_debug = "[gmetric] name: %s - val: %s - dmax: %s" %( str( metricname ), str( metricval ), str( self.dmax ) ) |
---|
[353] | 486 | |
---|
[659] | 487 | debug_msg( 10, printTime() + ' ' + metric_debug) |
---|
[353] | 488 | |
---|
[659] | 489 | gm = Gmetric( GMETRIC_TARGET_HOST, GMETRIC_TARGET_PORT ) |
---|
[353] | 490 | |
---|
[659] | 491 | gm.send( str( metricname ), str( metricval ), str( self.dmax ), valtype, units ) |
---|
[353] | 492 | |
---|
[659] | 493 | else: |
---|
| 494 | try: |
---|
| 495 | cmd = cmd + ' -c' + GMOND_CONF |
---|
[353] | 496 | |
---|
[659] | 497 | except NameError: |
---|
[353] | 498 | |
---|
[659] | 499 | debug_msg( 10, 'Assuming /etc/gmond.conf for gmetric cmd (omitting)' ) |
---|
[353] | 500 | |
---|
[659] | 501 | cmd = cmd + ' -n' + str( metricname )+ ' -v"' + str( metricval )+ '" -t' + str( valtype ) + ' -d' + str( self.dmax ) |
---|
[353] | 502 | |
---|
[659] | 503 | if len( units ) > 0: |
---|
[409] | 504 | |
---|
[659] | 505 | cmd = cmd + ' -u"' + units + '"' |
---|
[409] | 506 | |
---|
[659] | 507 | debug_msg( 10, printTime() + ' ' + cmd ) |
---|
[353] | 508 | |
---|
[659] | 509 | os.system( cmd ) |
---|
[353] | 510 | |
---|
[318] | 511 | class DataGatherer: |
---|
[23] | 512 | |
---|
[659] | 513 | """Skeleton class for batch system DataGatherer""" |
---|
[256] | 514 | |
---|
[659] | 515 | def printJobs( self, jobs ): |
---|
[355] | 516 | |
---|
[659] | 517 | """Print a jobinfo overview""" |
---|
[318] | 518 | |
---|
[659] | 519 | for name, attrs in self.jobs.items(): |
---|
[318] | 520 | |
---|
[659] | 521 | print 'job %s' %(name) |
---|
[318] | 522 | |
---|
[659] | 523 | for name, val in attrs.items(): |
---|
[318] | 524 | |
---|
[659] | 525 | print '\t%s = %s' %( name, val ) |
---|
[318] | 526 | |
---|
[659] | 527 | def printJob( self, jobs, job_id ): |
---|
[355] | 528 | |
---|
[659] | 529 | """Print job with job_id from jobs""" |
---|
[318] | 530 | |
---|
[659] | 531 | print 'job %s' %(job_id) |
---|
[318] | 532 | |
---|
[659] | 533 | for name, val in jobs[ job_id ].items(): |
---|
[318] | 534 | |
---|
[659] | 535 | print '\t%s = %s' %( name, val ) |
---|
[318] | 536 | |
---|
[660] | 537 | def getAttr( self, d, name ): |
---|
[507] | 538 | |
---|
[659] | 539 | """Return certain attribute from dictionary, if exists""" |
---|
[507] | 540 | |
---|
[660] | 541 | if d.has_key( name ): |
---|
[507] | 542 | |
---|
[660] | 543 | if type( d[ name ] ) == ListType: |
---|
[507] | 544 | |
---|
[660] | 545 | return string.join( d[ name ], ' ' ) |
---|
| 546 | |
---|
| 547 | return d[ name ] |
---|
| 548 | |
---|
| 549 | return '' |
---|
| 550 | |
---|
[659] | 551 | def jobDataChanged( self, jobs, job_id, attrs ): |
---|
[507] | 552 | |
---|
[659] | 553 | """Check if job with attrs and job_id in jobs has changed""" |
---|
[507] | 554 | |
---|
[659] | 555 | if jobs.has_key( job_id ): |
---|
[507] | 556 | |
---|
[659] | 557 | oldData = jobs[ job_id ] |
---|
| 558 | else: |
---|
| 559 | return 1 |
---|
[507] | 560 | |
---|
[659] | 561 | for name, val in attrs.items(): |
---|
[507] | 562 | |
---|
[659] | 563 | if oldData.has_key( name ): |
---|
[507] | 564 | |
---|
[659] | 565 | if oldData[ name ] != attrs[ name ]: |
---|
[507] | 566 | |
---|
[659] | 567 | return 1 |
---|
[507] | 568 | |
---|
[659] | 569 | else: |
---|
| 570 | return 1 |
---|
[507] | 571 | |
---|
[659] | 572 | return 0 |
---|
[507] | 573 | |
---|
[659] | 574 | def submitJobData( self ): |
---|
[507] | 575 | |
---|
[659] | 576 | """Submit job info list""" |
---|
[507] | 577 | |
---|
[659] | 578 | global BATCH_API |
---|
[512] | 579 | |
---|
[659] | 580 | self.dp.multicastGmetric( 'MONARCH-HEARTBEAT', str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
[507] | 581 | |
---|
[659] | 582 | running_jobs = 0 |
---|
| 583 | queued_jobs = 0 |
---|
[507] | 584 | |
---|
[659] | 585 | # Count how many running/queued jobs we found |
---|
[507] | 586 | # |
---|
[659] | 587 | for jobid, jobattrs in self.jobs.items(): |
---|
[507] | 588 | |
---|
[659] | 589 | if jobattrs[ 'status' ] == 'Q': |
---|
[507] | 590 | |
---|
[659] | 591 | queued_jobs += 1 |
---|
[507] | 592 | |
---|
[659] | 593 | elif jobattrs[ 'status' ] == 'R': |
---|
[507] | 594 | |
---|
[659] | 595 | running_jobs += 1 |
---|
[507] | 596 | |
---|
[659] | 597 | # Report running/queued jobs as seperate metric for a nice RRD graph |
---|
[507] | 598 | # |
---|
[659] | 599 | self.dp.multicastGmetric( 'MONARCH-RJ', str( running_jobs ), 'uint32', 'jobs' ) |
---|
| 600 | self.dp.multicastGmetric( 'MONARCH-QJ', str( queued_jobs ), 'uint32', 'jobs' ) |
---|
[507] | 601 | |
---|
[659] | 602 | # Report down/offline nodes in batch (PBS only ATM) |
---|
| 603 | # |
---|
| 604 | if BATCH_API == 'pbs': |
---|
[512] | 605 | |
---|
[659] | 606 | domain = fqdn_parts( socket.getfqdn() )[1] |
---|
[514] | 607 | |
---|
[659] | 608 | downed_nodes = list() |
---|
| 609 | offline_nodes = list() |
---|
| 610 | |
---|
| 611 | l = ['state'] |
---|
| 612 | |
---|
| 613 | for name, node in self.pq.getnodes().items(): |
---|
[512] | 614 | |
---|
[660] | 615 | if 'down' in node[ 'state' ]: |
---|
[512] | 616 | |
---|
[659] | 617 | downed_nodes.append( name ) |
---|
[512] | 618 | |
---|
[660] | 619 | if 'offline' in node[ 'state' ]: |
---|
[512] | 620 | |
---|
[659] | 621 | offline_nodes.append( name ) |
---|
[512] | 622 | |
---|
[659] | 623 | downnodeslist = do_nodelist( downed_nodes ) |
---|
| 624 | offlinenodeslist = do_nodelist( offline_nodes ) |
---|
[512] | 625 | |
---|
[659] | 626 | down_str = 'nodes=%s domain=%s reported=%s' %( string.join( downnodeslist, ';' ), domain, str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
| 627 | offl_str = 'nodes=%s domain=%s reported=%s' %( string.join( offlinenodeslist, ';' ), domain, str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
| 628 | self.dp.multicastGmetric( 'MONARCH-DOWN' , down_str ) |
---|
| 629 | self.dp.multicastGmetric( 'MONARCH-OFFLINE', offl_str ) |
---|
[514] | 630 | |
---|
[659] | 631 | # Now let's spread the knowledge |
---|
| 632 | # |
---|
| 633 | for jobid, jobattrs in self.jobs.items(): |
---|
[507] | 634 | |
---|
[659] | 635 | # Make gmetric values for each job: respect max gmetric value length |
---|
| 636 | # |
---|
[660] | 637 | gmetrics = self.compileGmetricVal( jobid, jobattrs ) |
---|
[507] | 638 | |
---|
[660] | 639 | for g_name, g_val in gmetrics.items(): |
---|
[507] | 640 | |
---|
[660] | 641 | self.dp.multicastGmetric( g_name, g_val ) |
---|
[507] | 642 | |
---|
[659] | 643 | def compileGmetricVal( self, jobid, jobattrs ): |
---|
[507] | 644 | |
---|
[660] | 645 | """Create gmetric name/value pairs of jobinfo""" |
---|
[507] | 646 | |
---|
[660] | 647 | gmetrics = { } |
---|
[507] | 648 | |
---|
[659] | 649 | for val_name, val_value in jobattrs.items(): |
---|
[507] | 650 | |
---|
[660] | 651 | gmetric_sequence = 0 |
---|
[507] | 652 | |
---|
[660] | 653 | if len( val_value ) > METRIC_MAX_VAL_LEN: |
---|
[507] | 654 | |
---|
[660] | 655 | while len( val_value ) > METRIC_MAX_VAL_LEN: |
---|
[507] | 656 | |
---|
[660] | 657 | gmetric_value = val_value[:METRIC_MAX_VAL_LEN] |
---|
| 658 | val_value = val_value[METRIC_MAX_VAL_LEN:] |
---|
[507] | 659 | |
---|
[660] | 660 | gmetric_name = 'MONARCHJOB$%s$%s$%s' %( jobid, string.upper(val_name), gmetric_sequence ) |
---|
[507] | 661 | |
---|
[660] | 662 | gmetrics[ gmetric_name ] = gmetric_value |
---|
[507] | 663 | |
---|
[660] | 664 | gmetric_sequence = gmetric_sequence + 1 |
---|
| 665 | else: |
---|
| 666 | gmetric_value = val_value |
---|
[507] | 667 | |
---|
[660] | 668 | gmetric_name = 'MONARCH$%s$%s$%s' %( jobid, string.upper(val_name), gmetric_sequence ) |
---|
[507] | 669 | |
---|
[660] | 670 | gmetrics[ gmetric_name ] = gmetric_value |
---|
[507] | 671 | |
---|
[660] | 672 | return gmetrics |
---|
[507] | 673 | |
---|
[660] | 674 | def daemon( self ): |
---|
[507] | 675 | |
---|
[660] | 676 | """Run as daemon forever""" |
---|
[507] | 677 | |
---|
[660] | 678 | # Fork the first child |
---|
| 679 | # |
---|
| 680 | pid = os.fork() |
---|
| 681 | if pid > 0: |
---|
| 682 | sys.exit(0) # end parent |
---|
[507] | 683 | |
---|
[660] | 684 | # creates a session and sets the process group ID |
---|
| 685 | # |
---|
| 686 | os.setsid() |
---|
[507] | 687 | |
---|
[660] | 688 | # Fork the second child |
---|
| 689 | # |
---|
| 690 | pid = os.fork() |
---|
| 691 | if pid > 0: |
---|
| 692 | sys.exit(0) # end parent |
---|
[507] | 693 | |
---|
[659] | 694 | write_pidfile() |
---|
[318] | 695 | |
---|
[660] | 696 | # Go to the root directory and set the umask |
---|
| 697 | # |
---|
| 698 | os.chdir('/') |
---|
| 699 | os.umask(0) |
---|
[318] | 700 | |
---|
[660] | 701 | sys.stdin.close() |
---|
| 702 | sys.stdout.close() |
---|
| 703 | sys.stderr.close() |
---|
[318] | 704 | |
---|
[660] | 705 | os.open('/dev/null', os.O_RDWR) |
---|
| 706 | os.dup2(0, 1) |
---|
| 707 | os.dup2(0, 2) |
---|
[318] | 708 | |
---|
[660] | 709 | self.run() |
---|
[318] | 710 | |
---|
[660] | 711 | def run( self ): |
---|
[355] | 712 | |
---|
[660] | 713 | """Main thread""" |
---|
[256] | 714 | |
---|
[660] | 715 | while ( 1 ): |
---|
[659] | 716 | |
---|
| 717 | self.getJobData() |
---|
| 718 | self.submitJobData() |
---|
| 719 | time.sleep( BATCH_POLL_INTERVAL ) |
---|
[256] | 720 | |
---|
[622] | 721 | # SGE code by Dave Love <fx@gnu.org>. Tested with SGE 6.0u8 and 6.0u11. May |
---|
| 722 | # work with SGE 6.1 (else should be easily fixable), but definitely doesn't |
---|
| 723 | # with 6.2. See also the fixmes. |
---|
[256] | 724 | |
---|
[507] | 725 | class NoJobs (Exception): |
---|
[659] | 726 | """Exception raised by empty job list in qstat output.""" |
---|
| 727 | pass |
---|
[256] | 728 | |
---|
[507] | 729 | class SgeQstatXMLParser(xml.sax.handler.ContentHandler): |
---|
[659] | 730 | """SAX handler for XML output from Sun Grid Engine's `qstat'.""" |
---|
[318] | 731 | |
---|
[659] | 732 | def __init__(self): |
---|
| 733 | self.value = "" |
---|
| 734 | self.joblist = [] |
---|
| 735 | self.job = {} |
---|
| 736 | self.queue = "" |
---|
| 737 | self.in_joblist = False |
---|
| 738 | self.lrequest = False |
---|
| 739 | self.eltq = deque() |
---|
| 740 | xml.sax.handler.ContentHandler.__init__(self) |
---|
[318] | 741 | |
---|
[659] | 742 | # The structure of the output is as follows (for SGE 6.0). It's |
---|
| 743 | # similar for 6.1, but radically different for 6.2, and is |
---|
| 744 | # undocumented generally. Unfortunately it's voluminous, and probably |
---|
| 745 | # doesn't scale to large clusters/queues. |
---|
[318] | 746 | |
---|
[659] | 747 | # <detailed_job_info xmlns:xsd="http://www.w3.org/2001/XMLSchema"> |
---|
| 748 | # <djob_info> |
---|
| 749 | # <qmaster_response> <!-- job --> |
---|
| 750 | # ... |
---|
| 751 | # <JB_ja_template> |
---|
| 752 | # <ulong_sublist> |
---|
| 753 | # ... <!-- start_time, state ... --> |
---|
| 754 | # </ulong_sublist> |
---|
| 755 | # </JB_ja_template> |
---|
| 756 | # <JB_ja_tasks> |
---|
| 757 | # <ulong_sublist> |
---|
| 758 | # ... <!-- task info |
---|
| 759 | # </ulong_sublist> |
---|
| 760 | # ... |
---|
| 761 | # </JB_ja_tasks> |
---|
| 762 | # ... |
---|
| 763 | # </qmaster_response> |
---|
| 764 | # </djob_info> |
---|
| 765 | # <messages> |
---|
| 766 | # ... |
---|
[318] | 767 | |
---|
[659] | 768 | # NB. We might treat each task as a separate job, like |
---|
| 769 | # straight qstat output, but the web interface expects jobs to |
---|
| 770 | # be identified by integers, not, say, <job number>.<task>. |
---|
[318] | 771 | |
---|
[659] | 772 | # So, I lied. If the job list is empty, we get invalid XML |
---|
| 773 | # like this, which we need to defend against: |
---|
[318] | 774 | |
---|
[659] | 775 | # <unknown_jobs xmlns:xsd="http://www.w3.org/2001/XMLSchema"> |
---|
| 776 | # <> |
---|
| 777 | # <ST_name>*</ST_name> |
---|
| 778 | # </> |
---|
| 779 | # </unknown_jobs> |
---|
[318] | 780 | |
---|
[659] | 781 | def startElement(self, name, attrs): |
---|
| 782 | self.value = "" |
---|
| 783 | if name == "djob_info": # job list |
---|
| 784 | self.in_joblist = True |
---|
| 785 | # The job container is "qmaster_response" in SGE 6.0 |
---|
| 786 | # and 6.1, but "element" in 6.2. This is only the very |
---|
| 787 | # start of what's necessary for 6.2, though (sigh). |
---|
| 788 | elif (name == "qmaster_response" or name == "element") \ |
---|
| 789 | and self.eltq[-1] == "djob_info": # job |
---|
| 790 | self.job = {"job_state": "U", "slots": 0, |
---|
| 791 | "nodes": [], "queued_timestamp": "", |
---|
| 792 | "queued_timestamp": "", "queue": "", |
---|
| 793 | "ppn": "0", "RN_max": 0, |
---|
| 794 | # fixme in endElement |
---|
| 795 | "requested_memory": 0, "requested_time": 0 |
---|
| 796 | } |
---|
| 797 | self.joblist.append(self.job) |
---|
| 798 | elif name == "qstat_l_requests": # resource request |
---|
| 799 | self.lrequest = True |
---|
| 800 | elif name == "unknown_jobs": |
---|
| 801 | raise NoJobs |
---|
| 802 | self.eltq.append (name) |
---|
[318] | 803 | |
---|
[659] | 804 | def characters(self, ch): |
---|
| 805 | self.value += ch |
---|
[318] | 806 | |
---|
[659] | 807 | def endElement(self, name): |
---|
| 808 | """Snarf job elements contents into job dictionary. |
---|
| 809 | Translate keys if appropriate.""" |
---|
[318] | 810 | |
---|
[659] | 811 | name_trans = { |
---|
| 812 | "JB_job_number": "number", |
---|
| 813 | "JB_job_name": "name", "JB_owner": "owner", |
---|
| 814 | "queue_name": "queue", "JAT_start_time": "start_timestamp", |
---|
| 815 | "JB_submission_time": "queued_timestamp" |
---|
| 816 | } |
---|
| 817 | value = self.value |
---|
| 818 | self.eltq.pop () |
---|
[318] | 819 | |
---|
[659] | 820 | if name == "djob_info": |
---|
| 821 | self.in_joblist = False |
---|
| 822 | self.job = {} |
---|
| 823 | elif name == "JAT_master_queue": |
---|
| 824 | self.job["queue"] = value.split("@")[0] |
---|
| 825 | elif name == "JG_qhostname": |
---|
| 826 | if not (value in self.job["nodes"]): |
---|
| 827 | self.job["nodes"].append(value) |
---|
| 828 | elif name == "JG_slots": # slots in use |
---|
| 829 | self.job["slots"] += int(value) |
---|
| 830 | elif name == "RN_max": # requested slots (tasks or parallel) |
---|
| 831 | self.job["RN_max"] = max (self.job["RN_max"], |
---|
| 832 | int(value)) |
---|
| 833 | elif name == "JAT_state": # job state (bitwise or) |
---|
| 834 | value = int (value) |
---|
| 835 | # Status values from sge_jobL.h |
---|
| 836 | #define JIDLE 0x00000000 |
---|
| 837 | #define JHELD 0x00000010 |
---|
| 838 | #define JMIGRATING 0x00000020 |
---|
| 839 | #define JQUEUED 0x00000040 |
---|
| 840 | #define JRUNNING 0x00000080 |
---|
| 841 | #define JSUSPENDED 0x00000100 |
---|
| 842 | #define JTRANSFERING 0x00000200 |
---|
| 843 | #define JDELETED 0x00000400 |
---|
| 844 | #define JWAITING 0x00000800 |
---|
| 845 | #define JEXITING 0x00001000 |
---|
| 846 | #define JWRITTEN 0x00002000 |
---|
| 847 | #define JSUSPENDED_ON_THRESHOLD 0x00010000 |
---|
| 848 | #define JFINISHED 0x00010000 |
---|
| 849 | if value & 0x80: |
---|
| 850 | self.job["status"] = "R" |
---|
| 851 | elif value & 0x40: |
---|
| 852 | self.job["status"] = "Q" |
---|
| 853 | else: |
---|
| 854 | self.job["status"] = "O" # `other' |
---|
| 855 | elif name == "CE_name" and self.lrequest and self.value in \ |
---|
| 856 | ("h_cpu", "s_cpu", "cpu", "h_core", "s_core"): |
---|
| 857 | # We're in a container for an interesting resource |
---|
| 858 | # request; record which type. |
---|
| 859 | self.lrequest = self.value |
---|
| 860 | elif name == "CE_doubleval" and self.lrequest: |
---|
| 861 | # if we're in a container for an interesting |
---|
| 862 | # resource request, use the maxmimum of the hard |
---|
| 863 | # and soft requests to record the requested CPU |
---|
| 864 | # or core. Fixme: I'm not sure if this logic is |
---|
| 865 | # right. |
---|
| 866 | if self.lrequest in ("h_core", "s_core"): |
---|
| 867 | self.job["requested_memory"] = \ |
---|
| 868 | max (float (value), |
---|
| 869 | self.job["requested_memory"]) |
---|
| 870 | # Fixme: Check what cpu means, c.f [hs]_cpu. |
---|
| 871 | elif self.lrequest in ("h_cpu", "s_cpu", "cpu"): |
---|
| 872 | self.job["requested_time"] = \ |
---|
| 873 | max (float (value), |
---|
| 874 | self.job["requested_time"]) |
---|
| 875 | elif name == "qstat_l_requests": |
---|
| 876 | self.lrequest = False |
---|
| 877 | elif self.job and self.in_joblist: |
---|
| 878 | if name in name_trans: |
---|
| 879 | name = name_trans[name] |
---|
| 880 | self.job[name] = value |
---|
[318] | 881 | |
---|
[507] | 882 | # Abstracted from PBS original. |
---|
[520] | 883 | # |
---|
| 884 | def do_nodelist( nodes ): |
---|
| 885 | |
---|
[659] | 886 | """Translate node list as appropriate.""" |
---|
[520] | 887 | |
---|
[659] | 888 | nodeslist = [ ] |
---|
| 889 | my_domain = fqdn_parts( socket.getfqdn() )[1] |
---|
[520] | 890 | |
---|
[659] | 891 | for node in nodes: |
---|
[520] | 892 | |
---|
[659] | 893 | host = node.split( '/' )[0] # not relevant for SGE |
---|
| 894 | h, host_domain = fqdn_parts(host) |
---|
[520] | 895 | |
---|
[659] | 896 | if host_domain == my_domain: |
---|
[520] | 897 | |
---|
[659] | 898 | host = h |
---|
[520] | 899 | |
---|
[659] | 900 | if nodeslist.count( host ) == 0: |
---|
[520] | 901 | |
---|
[659] | 902 | for translate_pattern in BATCH_HOST_TRANSLATE: |
---|
[520] | 903 | |
---|
[659] | 904 | if translate_pattern.find( '/' ) != -1: |
---|
[520] | 905 | |
---|
[659] | 906 | translate_orig = \ |
---|
| 907 | translate_pattern.split( '/' )[1] |
---|
| 908 | translate_new = \ |
---|
| 909 | translate_pattern.split( '/' )[2] |
---|
| 910 | host = re.sub( translate_orig, |
---|
| 911 | translate_new, host ) |
---|
| 912 | if not host in nodeslist: |
---|
| 913 | nodeslist.append( host ) |
---|
| 914 | return nodeslist |
---|
[318] | 915 | |
---|
| 916 | class SgeDataGatherer(DataGatherer): |
---|
| 917 | |
---|
[659] | 918 | jobs = {} |
---|
[61] | 919 | |
---|
[659] | 920 | def __init__( self ): |
---|
| 921 | self.jobs = {} |
---|
| 922 | self.timeoffset = 0 |
---|
| 923 | self.dp = DataProcessor() |
---|
[318] | 924 | |
---|
[659] | 925 | def getJobData( self ): |
---|
| 926 | """Gather all data on current jobs in SGE""" |
---|
[318] | 927 | |
---|
[659] | 928 | import popen2 |
---|
[318] | 929 | |
---|
[659] | 930 | self.cur_time = 0 |
---|
| 931 | queues = "" |
---|
| 932 | if QUEUE: # only for specific queues |
---|
| 933 | # Fixme: assumes queue names don't contain single |
---|
| 934 | # quote or comma. Don't know what the SGE rules are. |
---|
| 935 | queues = " -q '" + string.join (QUEUE, ",") + "'" |
---|
| 936 | # Note the comment in SgeQstatXMLParser about scaling with |
---|
| 937 | # this method of getting data. I haven't found better one. |
---|
| 938 | # Output with args `-xml -ext -f -r' is easier to parse |
---|
| 939 | # in some ways, harder in others, but it doesn't provide |
---|
| 940 | # the submission time (at least SGE 6.0). The pipeline |
---|
| 941 | # into sed corrects bogus XML observed with a configuration |
---|
| 942 | # of SGE 6.0u8, which otherwise causes the parsing to hang. |
---|
| 943 | piping = popen2.Popen3("qstat -u '*' -j '*' -xml | \ |
---|
[622] | 944 | sed -e 's/reported usage>/reported_usage>/g' -e 's;<\/*JATASK:.*>;;'" \ |
---|
[659] | 945 | + queues, True) |
---|
| 946 | qstatparser = SgeQstatXMLParser() |
---|
| 947 | parse_err = 0 |
---|
| 948 | try: |
---|
| 949 | xml.sax.parse(piping.fromchild, qstatparser) |
---|
| 950 | except NoJobs: |
---|
| 951 | pass |
---|
| 952 | except: |
---|
| 953 | parse_err = 1 |
---|
| 954 | if piping.wait(): |
---|
[660] | 955 | debug_msg(10, |
---|
[659] | 956 | "qstat error, skipping until next polling interval: " |
---|
| 957 | + piping.childerr.readline()) |
---|
[660] | 958 | return None |
---|
| 959 | elif parse_err: |
---|
| 960 | debug_msg(10, "Bad XML output from qstat"()) |
---|
| 961 | exit (1) |
---|
[659] | 962 | for f in piping.fromchild, piping.tochild, piping.childerr: |
---|
| 963 | f.close() |
---|
| 964 | self.cur_time = time.time() |
---|
| 965 | jobs_processed = [] |
---|
| 966 | for job in qstatparser.joblist: |
---|
| 967 | job_id = job["number"] |
---|
| 968 | if job["status"] in [ 'Q', 'R' ]: |
---|
| 969 | jobs_processed.append(job_id) |
---|
| 970 | if job["status"] == "R": |
---|
| 971 | job["nodes"] = do_nodelist (job["nodes"]) |
---|
| 972 | # Fixme: why is job["nodes"] sometimes null? |
---|
| 973 | try: |
---|
| 974 | # Fixme: Is this sensible? The |
---|
| 975 | # PBS-type PPN isn't something you use |
---|
| 976 | # with SGE. |
---|
| 977 | job["ppn"] = float(job["slots"]) / \ |
---|
| 978 | len(job["nodes"]) |
---|
| 979 | except: |
---|
| 980 | job["ppn"] = 0 |
---|
| 981 | if DETECT_TIME_DIFFS: |
---|
| 982 | # If a job start is later than our |
---|
| 983 | # current date, that must mean |
---|
| 984 | # the SGE server's time is later |
---|
| 985 | # than our local time. |
---|
| 986 | start_timestamp = \ |
---|
| 987 | int (job["start_timestamp"]) |
---|
| 988 | if start_timestamp > \ |
---|
| 989 | int(self.cur_time) + \ |
---|
| 990 | int(self.timeoffset): |
---|
[318] | 991 | |
---|
[659] | 992 | self.timeoffset = \ |
---|
| 993 | start_timestamp - \ |
---|
| 994 | int(self.cur_time) |
---|
| 995 | else: |
---|
| 996 | # fixme: Note sure what this should be: |
---|
| 997 | job["ppn"] = job["RN_max"] |
---|
| 998 | job["nodes"] = "1" |
---|
[318] | 999 | |
---|
[659] | 1000 | myAttrs = {} |
---|
| 1001 | for attr in ["name", "queue", "owner", |
---|
| 1002 | "requested_time", "status", |
---|
| 1003 | "requested_memory", "ppn", |
---|
| 1004 | "start_timestamp", "queued_timestamp"]: |
---|
| 1005 | myAttrs[attr] = str(job[attr]) |
---|
| 1006 | myAttrs["nodes"] = job["nodes"] |
---|
| 1007 | myAttrs["reported"] = str(int(self.cur_time) + \ |
---|
| 1008 | int(self.timeoffset)) |
---|
| 1009 | myAttrs["domain"] = fqdn_parts(socket.getfqdn())[1] |
---|
| 1010 | myAttrs["poll_interval"] = str(BATCH_POLL_INTERVAL) |
---|
[318] | 1011 | |
---|
[659] | 1012 | if self.jobDataChanged(self.jobs, job_id, myAttrs) \ |
---|
| 1013 | and myAttrs["status"] in ["R", "Q"]: |
---|
| 1014 | self.jobs[job_id] = myAttrs |
---|
| 1015 | for id, attrs in self.jobs.items(): |
---|
| 1016 | if id not in jobs_processed: |
---|
| 1017 | del self.jobs[id] |
---|
[318] | 1018 | |
---|
[524] | 1019 | # LSF code by Mahmoud Hanafi <hanafim@users.sourceforge.nt> |
---|
| 1020 | # Requres LSFObject http://sourceforge.net/projects/lsfobject |
---|
| 1021 | # |
---|
| 1022 | class LsfDataGatherer(DataGatherer): |
---|
[525] | 1023 | |
---|
[524] | 1024 | """This is the DataGatherer for LSf""" |
---|
| 1025 | |
---|
| 1026 | global lsfObject |
---|
| 1027 | |
---|
| 1028 | def __init__( self ): |
---|
[525] | 1029 | |
---|
[524] | 1030 | self.jobs = { } |
---|
| 1031 | self.timeoffset = 0 |
---|
| 1032 | self.dp = DataProcessor() |
---|
| 1033 | self.initLsfQuery() |
---|
| 1034 | |
---|
[525] | 1035 | def _countDuplicatesInList( self, dupedList ): |
---|
| 1036 | |
---|
[660] | 1037 | countDupes = { } |
---|
[525] | 1038 | |
---|
[660] | 1039 | for item in dupedList: |
---|
[525] | 1040 | |
---|
[660] | 1041 | if not countDupes.has_key( item ): |
---|
[525] | 1042 | |
---|
[660] | 1043 | countDupes[ item ] = 1 |
---|
| 1044 | else: |
---|
| 1045 | countDupes[ item ] = countDupes[ item ] + 1 |
---|
[525] | 1046 | |
---|
[660] | 1047 | dupeCountList = [ ] |
---|
[525] | 1048 | |
---|
[660] | 1049 | for item, count in countDupes.items(): |
---|
[525] | 1050 | |
---|
[660] | 1051 | dupeCountList.append( ( item, count ) ) |
---|
[525] | 1052 | |
---|
[660] | 1053 | return dupeCountList |
---|
[524] | 1054 | # |
---|
| 1055 | #lst = ['I1','I2','I1','I3','I4','I4','I7','I7','I7','I7','I7'] |
---|
| 1056 | #print _countDuplicatesInList(lst) |
---|
| 1057 | #[('I1', 2), ('I3', 1), ('I2', 1), ('I4', 2), ('I7', 5)] |
---|
| 1058 | ######################## |
---|
| 1059 | |
---|
| 1060 | def initLsfQuery( self ): |
---|
| 1061 | self.pq = None |
---|
| 1062 | self.pq = lsfObject.jobInfoEntObject() |
---|
| 1063 | |
---|
| 1064 | def getJobData( self, known_jobs="" ): |
---|
| 1065 | """Gather all data on current jobs in LSF""" |
---|
| 1066 | if len( known_jobs ) > 0: |
---|
| 1067 | jobs = known_jobs |
---|
| 1068 | else: |
---|
| 1069 | jobs = { } |
---|
| 1070 | joblist = {} |
---|
| 1071 | joblist = self.pq.getJobInfo() |
---|
| 1072 | nodelist = '' |
---|
| 1073 | |
---|
| 1074 | self.cur_time = time.time() |
---|
| 1075 | |
---|
| 1076 | jobs_processed = [ ] |
---|
| 1077 | |
---|
| 1078 | for name, attrs in joblist.items(): |
---|
| 1079 | job_id = str(name) |
---|
| 1080 | jobs_processed.append( job_id ) |
---|
| 1081 | name = self.getAttr( attrs, 'jobName' ) |
---|
| 1082 | queue = self.getAttr( self.getAttr( attrs, 'submit') , 'queue' ) |
---|
| 1083 | owner = self.getAttr( attrs, 'user' ) |
---|
| 1084 | |
---|
| 1085 | ### THIS IS THE rLimit List index values |
---|
| 1086 | #define LSF_RLIMIT_CPU 0 /* cpu time in milliseconds */ |
---|
| 1087 | #define LSF_RLIMIT_FSIZE 1 /* maximum file size */ |
---|
| 1088 | #define LSF_RLIMIT_DATA 2 /* data size */ |
---|
| 1089 | #define LSF_RLIMIT_STACK 3 /* stack size */ |
---|
| 1090 | #define LSF_RLIMIT_CORE 4 /* core file size */ |
---|
| 1091 | #define LSF_RLIMIT_RSS 5 /* resident set size */ |
---|
| 1092 | #define LSF_RLIMIT_NOFILE 6 /* open files */ |
---|
| 1093 | #define LSF_RLIMIT_OPEN_MAX 7 /* (from HP-UX) */ |
---|
| 1094 | #define LSF_RLIMIT_VMEM 8 /* maximum swap mem */ |
---|
| 1095 | #define LSF_RLIMIT_SWAP 8 |
---|
| 1096 | #define LSF_RLIMIT_RUN 9 /* max wall-clock time limit */ |
---|
| 1097 | #define LSF_RLIMIT_PROCESS 10 /* process number limit */ |
---|
| 1098 | #define LSF_RLIMIT_THREAD 11 /* thread number limit (introduced in LSF6.0) */ |
---|
| 1099 | #define LSF_RLIM_NLIMITS 12 /* number of resource limits */ |
---|
| 1100 | |
---|
| 1101 | requested_time = self.getAttr( self.getAttr( attrs, 'submit') , 'rLimits' )[9] |
---|
| 1102 | if requested_time == -1: |
---|
| 1103 | requested_time = "" |
---|
| 1104 | requested_memory = self.getAttr( self.getAttr( attrs, 'submit') , 'rLimits' )[8] |
---|
| 1105 | if requested_memory == -1: |
---|
| 1106 | requested_memory = "" |
---|
| 1107 | # This tries to get proc per node. We don't support this right now |
---|
| 1108 | ppn = 0 #self.getAttr( self.getAttr( attrs, 'SubmitList') , 'numProessors' ) |
---|
| 1109 | requested_cpus = self.getAttr( self.getAttr( attrs, 'submit') , 'numProcessors' ) |
---|
| 1110 | if requested_cpus == None or requested_cpus == "": |
---|
| 1111 | requested_cpus = 1 |
---|
| 1112 | |
---|
[660] | 1113 | if QUEUE: |
---|
| 1114 | for q in QUEUE: |
---|
| 1115 | if q == queue: |
---|
| 1116 | display_queue = 1 |
---|
| 1117 | break |
---|
| 1118 | else: |
---|
| 1119 | display_queue = 0 |
---|
| 1120 | continue |
---|
| 1121 | if display_queue == 0: |
---|
| 1122 | continue |
---|
[524] | 1123 | |
---|
| 1124 | runState = self.getAttr( attrs, 'status' ) |
---|
| 1125 | if runState == 4: |
---|
| 1126 | status = 'R' |
---|
| 1127 | else: |
---|
| 1128 | status = 'Q' |
---|
| 1129 | queued_timestamp = self.getAttr( attrs, 'submitTime' ) |
---|
| 1130 | |
---|
| 1131 | if status == 'R': |
---|
| 1132 | start_timestamp = self.getAttr( attrs, 'startTime' ) |
---|
| 1133 | nodesCpu = dict(self._countDuplicatesInList(self.getAttr( attrs, 'exHosts' ))) |
---|
| 1134 | nodelist = nodesCpu.keys() |
---|
| 1135 | |
---|
| 1136 | if DETECT_TIME_DIFFS: |
---|
| 1137 | |
---|
| 1138 | # If a job start if later than our current date, |
---|
| 1139 | # that must mean the Torque server's time is later |
---|
| 1140 | # than our local time. |
---|
| 1141 | |
---|
| 1142 | if int(start_timestamp) > int( int(self.cur_time) + int(self.timeoffset) ): |
---|
| 1143 | |
---|
| 1144 | self.timeoffset = int( int(start_timestamp) - int(self.cur_time) ) |
---|
| 1145 | |
---|
| 1146 | elif status == 'Q': |
---|
| 1147 | start_timestamp = '' |
---|
| 1148 | count_mynodes = 0 |
---|
| 1149 | numeric_node = 1 |
---|
| 1150 | nodelist = '' |
---|
| 1151 | |
---|
| 1152 | myAttrs = { } |
---|
| 1153 | if name == "": |
---|
| 1154 | myAttrs['name'] = "none" |
---|
| 1155 | else: |
---|
| 1156 | myAttrs['name'] = name |
---|
| 1157 | |
---|
[659] | 1158 | myAttrs[ 'owner' ] = owner |
---|
| 1159 | myAttrs[ 'requested_time' ] = str(requested_time) |
---|
| 1160 | myAttrs[ 'requested_memory' ] = str(requested_memory) |
---|
| 1161 | myAttrs[ 'requested_cpus' ] = str(requested_cpus) |
---|
| 1162 | myAttrs[ 'ppn' ] = str( ppn ) |
---|
| 1163 | myAttrs[ 'status' ] = status |
---|
| 1164 | myAttrs[ 'start_timestamp' ] = str(start_timestamp) |
---|
| 1165 | myAttrs[ 'queue' ] = str(queue) |
---|
| 1166 | myAttrs[ 'queued_timestamp' ] = str(queued_timestamp) |
---|
| 1167 | myAttrs[ 'reported' ] = str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) |
---|
| 1168 | myAttrs[ 'nodes' ] = do_nodelist( nodelist ) |
---|
[660] | 1169 | myAttrs[ 'domain' ] = fqdn_parts( socket.getfqdn() )[1] |
---|
[659] | 1170 | myAttrs[ 'poll_interval' ] = str(BATCH_POLL_INTERVAL) |
---|
[524] | 1171 | |
---|
| 1172 | if self.jobDataChanged( jobs, job_id, myAttrs ) and myAttrs['status'] in [ 'R', 'Q' ]: |
---|
| 1173 | jobs[ job_id ] = myAttrs |
---|
| 1174 | |
---|
[525] | 1175 | debug_msg( 10, printTime() + ' job %s state changed' %(job_id) ) |
---|
[524] | 1176 | |
---|
| 1177 | for id, attrs in jobs.items(): |
---|
| 1178 | if id not in jobs_processed: |
---|
[525] | 1179 | # This one isn't there anymore |
---|
[524] | 1180 | # |
---|
| 1181 | del jobs[ id ] |
---|
| 1182 | |
---|
[660] | 1183 | self.jobs = jobs |
---|
[524] | 1184 | |
---|
[355] | 1185 | class PbsDataGatherer( DataGatherer ): |
---|
[318] | 1186 | |
---|
[659] | 1187 | """This is the DataGatherer for PBS and Torque""" |
---|
[318] | 1188 | |
---|
[659] | 1189 | global PBSQuery, PBSError |
---|
[256] | 1190 | |
---|
[659] | 1191 | def __init__( self ): |
---|
[354] | 1192 | |
---|
[659] | 1193 | """Setup appropriate variables""" |
---|
[23] | 1194 | |
---|
[659] | 1195 | self.jobs = { } |
---|
| 1196 | self.timeoffset = 0 |
---|
| 1197 | self.dp = DataProcessor() |
---|
[354] | 1198 | |
---|
[659] | 1199 | self.initPbsQuery() |
---|
[23] | 1200 | |
---|
[659] | 1201 | def initPbsQuery( self ): |
---|
[91] | 1202 | |
---|
[659] | 1203 | self.pq = None |
---|
[354] | 1204 | |
---|
[659] | 1205 | if( BATCH_SERVER ): |
---|
[354] | 1206 | |
---|
[659] | 1207 | self.pq = PBSQuery( BATCH_SERVER ) |
---|
| 1208 | else: |
---|
| 1209 | self.pq = PBSQuery() |
---|
[91] | 1210 | |
---|
[659] | 1211 | def getJobData( self ): |
---|
[354] | 1212 | |
---|
[659] | 1213 | """Gather all data on current jobs in Torque""" |
---|
[26] | 1214 | |
---|
[659] | 1215 | joblist = {} |
---|
| 1216 | self.cur_time = 0 |
---|
[349] | 1217 | |
---|
[659] | 1218 | try: |
---|
| 1219 | joblist = self.pq.getjobs() |
---|
| 1220 | self.cur_time = time.time() |
---|
[354] | 1221 | |
---|
[659] | 1222 | except PBSError, detail: |
---|
[354] | 1223 | |
---|
[659] | 1224 | debug_msg( 10, "Caught PBS unavailable, skipping until next polling interval: " + str( detail ) ) |
---|
| 1225 | return None |
---|
[354] | 1226 | |
---|
[659] | 1227 | jobs_processed = [ ] |
---|
[26] | 1228 | |
---|
[659] | 1229 | for name, attrs in joblist.items(): |
---|
| 1230 | display_queue = 1 |
---|
| 1231 | job_id = name.split( '.' )[0] |
---|
[26] | 1232 | |
---|
[661] | 1233 | owner = self.getAttr( attrs, 'Job_Owner' ) |
---|
[659] | 1234 | name = self.getAttr( attrs, 'Job_Name' ) |
---|
| 1235 | queue = self.getAttr( attrs, 'queue' ) |
---|
[661] | 1236 | nodect = self.getAttr( attrs['nodes'], 'nodect' ) |
---|
| 1237 | exec_group = self.getAttr( attrs, 'egroup' ) |
---|
[317] | 1238 | |
---|
[660] | 1239 | requested_time = self.getAttr( attrs['Resource_List'], 'walltime' ) |
---|
| 1240 | requested_memory = self.getAttr( attrs['Resource_List'], 'mem' ) |
---|
[95] | 1241 | |
---|
[661] | 1242 | |
---|
| 1243 | requested_nodes = '' |
---|
[660] | 1244 | mynoderequest = self.getAttr( attrs['Resource_List'], 'nodes' ) |
---|
[95] | 1245 | |
---|
[659] | 1246 | ppn = '' |
---|
[661] | 1247 | attributes = '' |
---|
[281] | 1248 | |
---|
[661] | 1249 | if mynoderequest.find( ':' ) != -1: |
---|
[95] | 1250 | |
---|
[659] | 1251 | mynoderequest_fields = mynoderequest.split( ':' ) |
---|
[281] | 1252 | |
---|
[659] | 1253 | for mynoderequest_field in mynoderequest_fields: |
---|
[281] | 1254 | |
---|
[661] | 1255 | if mynoderequest_field.isdigit(): |
---|
| 1256 | |
---|
| 1257 | continue #TODO add requested_nodes if is hostname(s) |
---|
| 1258 | |
---|
[659] | 1259 | if mynoderequest_field.find( 'ppn' ) != -1: |
---|
[281] | 1260 | |
---|
[659] | 1261 | ppn = mynoderequest_field.split( 'ppn=' )[1] |
---|
[281] | 1262 | |
---|
[661] | 1263 | else: |
---|
| 1264 | |
---|
| 1265 | if attributes == '': |
---|
| 1266 | |
---|
| 1267 | attributes = '%s' %mynoderequest_field |
---|
| 1268 | else: |
---|
| 1269 | attributes = '%s:%s' %( attributes, mynoderequest_field ) |
---|
| 1270 | |
---|
[659] | 1271 | status = self.getAttr( attrs, 'job_state' ) |
---|
[25] | 1272 | |
---|
[661] | 1273 | if status in [ 'Q', 'R', 'W' ]: |
---|
[450] | 1274 | |
---|
[659] | 1275 | jobs_processed.append( job_id ) |
---|
[450] | 1276 | |
---|
[661] | 1277 | create_timestamp = self.getAttr( attrs, 'ctime' ) |
---|
| 1278 | running_nodes = '' |
---|
| 1279 | exec_nodestr = '' |
---|
[243] | 1280 | |
---|
[659] | 1281 | if status == 'R': |
---|
[133] | 1282 | |
---|
[661] | 1283 | #start_timestamp = self.getAttr( attrs, 'etime' ) |
---|
| 1284 | start_timestamp = self.getAttr( attrs, 'start_time' ) |
---|
| 1285 | exec_nodestr = self.getAttr( attrs, 'exec_host' ) |
---|
[133] | 1286 | |
---|
[661] | 1287 | nodes = exec_nodestr.split( '+' ) |
---|
[659] | 1288 | nodeslist = do_nodelist( nodes ) |
---|
[661] | 1289 | running_nodes = string.join( nodeslist, ' ' ) |
---|
[354] | 1290 | |
---|
[659] | 1291 | if DETECT_TIME_DIFFS: |
---|
[185] | 1292 | |
---|
[659] | 1293 | # If a job start if later than our current date, |
---|
| 1294 | # that must mean the Torque server's time is later |
---|
| 1295 | # than our local time. |
---|
| 1296 | |
---|
| 1297 | if int( start_timestamp ) > int( int( self.cur_time ) + int( self.timeoffset ) ): |
---|
[185] | 1298 | |
---|
[659] | 1299 | self.timeoffset = int( int(start_timestamp) - int(self.cur_time) ) |
---|
[185] | 1300 | |
---|
[659] | 1301 | elif status == 'Q': |
---|
[95] | 1302 | |
---|
[659] | 1303 | # 'mynodequest' can be a string in the following syntax according to the |
---|
| 1304 | # Torque Administator's manual: |
---|
| 1305 | # |
---|
| 1306 | # {<node_count> | <hostname>}[:ppn=<ppn>][:<property>[:<property>]...][+ ...] |
---|
| 1307 | # {<node_count> | <hostname>}[:ppn=<ppn>][:<property>[:<property>]...][+ ...] |
---|
| 1308 | # etc |
---|
| 1309 | # |
---|
[451] | 1310 | |
---|
[659] | 1311 | # |
---|
| 1312 | # For now we only count the amount of nodes request and ignore properties |
---|
| 1313 | # |
---|
[451] | 1314 | |
---|
[659] | 1315 | start_timestamp = '' |
---|
| 1316 | count_mynodes = 0 |
---|
[354] | 1317 | |
---|
[661] | 1318 | queued_timestamp = self.getAttr( attrs, 'qtime' ) |
---|
| 1319 | |
---|
[659] | 1320 | for node in mynoderequest.split( '+' ): |
---|
[67] | 1321 | |
---|
[659] | 1322 | # Just grab the {node_count|hostname} part and ignore properties |
---|
| 1323 | # |
---|
| 1324 | nodepart = node.split( ':' )[0] |
---|
[67] | 1325 | |
---|
[659] | 1326 | # Let's assume a node_count value |
---|
| 1327 | # |
---|
| 1328 | numeric_node = 1 |
---|
[451] | 1329 | |
---|
[659] | 1330 | # Chop the value up into characters |
---|
| 1331 | # |
---|
| 1332 | for letter in nodepart: |
---|
[67] | 1333 | |
---|
[659] | 1334 | # If this char is not a digit (0-9), this must be a hostname |
---|
| 1335 | # |
---|
| 1336 | if letter not in string.digits: |
---|
[133] | 1337 | |
---|
[659] | 1338 | numeric_node = 0 |
---|
[133] | 1339 | |
---|
[659] | 1340 | # If this is a hostname, just count this as one (1) node |
---|
| 1341 | # |
---|
| 1342 | if not numeric_node: |
---|
[354] | 1343 | |
---|
[659] | 1344 | count_mynodes = count_mynodes + 1 |
---|
| 1345 | else: |
---|
[451] | 1346 | |
---|
[659] | 1347 | # If this a number, it must be the node_count |
---|
| 1348 | # and increase our count with it's value |
---|
| 1349 | # |
---|
| 1350 | try: |
---|
| 1351 | count_mynodes = count_mynodes + int( nodepart ) |
---|
[354] | 1352 | |
---|
[659] | 1353 | except ValueError, detail: |
---|
[354] | 1354 | |
---|
[659] | 1355 | # When we arrive here I must be bugged or very confused |
---|
| 1356 | # THIS SHOULD NOT HAPPEN! |
---|
| 1357 | # |
---|
| 1358 | debug_msg( 10, str( detail ) ) |
---|
| 1359 | debug_msg( 10, "Encountered weird node in Resources_List?!" ) |
---|
| 1360 | debug_msg( 10, 'nodepart = ' + str( nodepart ) ) |
---|
| 1361 | debug_msg( 10, 'job = ' + str( name ) ) |
---|
| 1362 | debug_msg( 10, 'attrs = ' + str( attrs ) ) |
---|
| 1363 | |
---|
| 1364 | nodeslist = str( count_mynodes ) |
---|
| 1365 | else: |
---|
| 1366 | start_timestamp = '' |
---|
| 1367 | nodeslist = '' |
---|
[133] | 1368 | |
---|
[659] | 1369 | myAttrs = { } |
---|
[26] | 1370 | |
---|
[661] | 1371 | myAttrs[ 'name' ] = str( name ) |
---|
| 1372 | myAttrs[ 'status' ] = str( status ) |
---|
| 1373 | myAttrs[ 'queue' ] = str( queue ) |
---|
| 1374 | myAttrs[ 'owner' ] = str( owner ) |
---|
| 1375 | myAttrs[ 'nodect' ] = str( nodect ) |
---|
| 1376 | myAttrs[ 'exec.group' ] = str( exec_group ) |
---|
| 1377 | myAttrs[ 'exec.hostnames' ] = str( running_nodes ) |
---|
| 1378 | myAttrs[ 'exec.nodestr' ] = str( exec_nodestr ) |
---|
| 1379 | myAttrs[ 'req.walltime' ] = str( requested_time ) |
---|
| 1380 | myAttrs[ 'req.memory' ] = str( requested_memory ) |
---|
| 1381 | myAttrs[ 'req.nodes' ] = str( requested_nodes ) |
---|
| 1382 | myAttrs[ 'req.ppn' ] = str( ppn ) |
---|
| 1383 | myAttrs[ 'req.attributes' ] = str( attributes ) |
---|
| 1384 | myAttrs[ 'timestamp.running' ] = str( start_timestamp ) |
---|
| 1385 | myAttrs[ 'timestamp.created' ] = str( create_timestamp ) |
---|
| 1386 | myAttrs[ 'timestamp.queued' ] = str( queued_timestamp ) |
---|
[354] | 1387 | |
---|
[661] | 1388 | if self.jobDataChanged( self.jobs, job_id, myAttrs ) and myAttrs['status'] in [ 'R', 'Q', 'W' ]: |
---|
[61] | 1389 | |
---|
[659] | 1390 | self.jobs[ job_id ] = myAttrs |
---|
[26] | 1391 | |
---|
[659] | 1392 | for id, attrs in self.jobs.items(): |
---|
[76] | 1393 | |
---|
[659] | 1394 | if id not in jobs_processed: |
---|
[76] | 1395 | |
---|
[659] | 1396 | # This one isn't there anymore; toedeledoki! |
---|
| 1397 | # |
---|
| 1398 | del self.jobs[ id ] |
---|
[76] | 1399 | |
---|
[363] | 1400 | # |
---|
| 1401 | # Gmetric by Nick Galbreath - nickg(a.t)modp(d.o.t)com |
---|
| 1402 | # Version 1.0 - 21-April2-2007 |
---|
| 1403 | # http://code.google.com/p/embeddedgmetric/ |
---|
| 1404 | # |
---|
| 1405 | # Modified by: Ramon Bastiaans |
---|
| 1406 | # For the Job Monarch Project, see: https://subtrac.sara.nl/oss/jobmonarch/ |
---|
| 1407 | # |
---|
| 1408 | # added: DEFAULT_TYPE for Gmetric's |
---|
| 1409 | # added: checkHostProtocol to determine if target is multicast or not |
---|
| 1410 | # changed: allow default for Gmetric constructor |
---|
| 1411 | # changed: allow defaults for all send() values except dmax |
---|
| 1412 | # |
---|
| 1413 | |
---|
[362] | 1414 | GMETRIC_DEFAULT_TYPE = 'string' |
---|
| 1415 | GMETRIC_DEFAULT_HOST = '127.0.0.1' |
---|
| 1416 | GMETRIC_DEFAULT_PORT = '8649' |
---|
[659] | 1417 | GMETRIC_DEFAULT_UNITS = '' |
---|
[362] | 1418 | |
---|
| 1419 | class Gmetric: |
---|
| 1420 | |
---|
[659] | 1421 | global GMETRIC_DEFAULT_HOST, GMETRIC_DEFAULT_PORT |
---|
[362] | 1422 | |
---|
[659] | 1423 | slope = { 'zero' : 0, 'positive' : 1, 'negative' : 2, 'both' : 3, 'unspecified' : 4 } |
---|
| 1424 | type = ( '', 'string', 'uint16', 'int16', 'uint32', 'int32', 'float', 'double', 'timestamp' ) |
---|
| 1425 | protocol = ( 'udp', 'multicast' ) |
---|
[362] | 1426 | |
---|
[659] | 1427 | def __init__( self, host=GMETRIC_DEFAULT_HOST, port=GMETRIC_DEFAULT_PORT ): |
---|
[362] | 1428 | |
---|
[659] | 1429 | global GMETRIC_DEFAULT_TYPE |
---|
[362] | 1430 | |
---|
[659] | 1431 | self.prot = self.checkHostProtocol( host ) |
---|
| 1432 | self.msg = xdrlib.Packer() |
---|
| 1433 | self.socket = socket.socket( socket.AF_INET, socket.SOCK_DGRAM ) |
---|
[362] | 1434 | |
---|
[659] | 1435 | if self.prot not in self.protocol: |
---|
[362] | 1436 | |
---|
[659] | 1437 | raise ValueError( "Protocol must be one of: " + str( self.protocol ) ) |
---|
[362] | 1438 | |
---|
[659] | 1439 | if self.prot == 'multicast': |
---|
[362] | 1440 | |
---|
[659] | 1441 | # Set multicast options |
---|
| 1442 | # |
---|
| 1443 | self.socket.setsockopt( socket.IPPROTO_IP, socket.IP_MULTICAST_TTL, 20 ) |
---|
[362] | 1444 | |
---|
[659] | 1445 | self.hostport = ( host, int( port ) ) |
---|
| 1446 | self.slopestr = 'both' |
---|
| 1447 | self.tmax = 60 |
---|
[362] | 1448 | |
---|
[659] | 1449 | def checkHostProtocol( self, ip ): |
---|
[362] | 1450 | |
---|
[659] | 1451 | """Detect if a ip adress is a multicast address""" |
---|
[471] | 1452 | |
---|
[659] | 1453 | MULTICAST_ADDRESS_MIN = ( "224", "0", "0", "0" ) |
---|
| 1454 | MULTICAST_ADDRESS_MAX = ( "239", "255", "255", "255" ) |
---|
[362] | 1455 | |
---|
[659] | 1456 | ip_fields = ip.split( '.' ) |
---|
[362] | 1457 | |
---|
[659] | 1458 | if ip_fields >= MULTICAST_ADDRESS_MIN and ip_fields <= MULTICAST_ADDRESS_MAX: |
---|
[362] | 1459 | |
---|
[659] | 1460 | return 'multicast' |
---|
| 1461 | else: |
---|
| 1462 | return 'udp' |
---|
[362] | 1463 | |
---|
[659] | 1464 | def send( self, name, value, dmax, typestr = '', units = '' ): |
---|
[362] | 1465 | |
---|
[659] | 1466 | if len( units ) == 0: |
---|
| 1467 | units = GMETRIC_DEFAULT_UNITS |
---|
[471] | 1468 | |
---|
[659] | 1469 | if len( typestr ) == 0: |
---|
| 1470 | typestr = GMETRIC_DEFAULT_TYPE |
---|
[362] | 1471 | |
---|
[659] | 1472 | msg = self.makexdr( name, value, typestr, units, self.slopestr, self.tmax, dmax ) |
---|
[409] | 1473 | |
---|
[659] | 1474 | return self.socket.sendto( msg, self.hostport ) |
---|
[362] | 1475 | |
---|
[659] | 1476 | def makexdr( self, name, value, typestr, unitstr, slopestr, tmax, dmax ): |
---|
[362] | 1477 | |
---|
[659] | 1478 | if slopestr not in self.slope: |
---|
[362] | 1479 | |
---|
[659] | 1480 | raise ValueError( "Slope must be one of: " + str( self.slope.keys() ) ) |
---|
[362] | 1481 | |
---|
[659] | 1482 | if typestr not in self.type: |
---|
[362] | 1483 | |
---|
[659] | 1484 | raise ValueError( "Type must be one of: " + str( self.type ) ) |
---|
[362] | 1485 | |
---|
[659] | 1486 | if len( name ) == 0: |
---|
[362] | 1487 | |
---|
[659] | 1488 | raise ValueError( "Name must be non-empty" ) |
---|
[362] | 1489 | |
---|
[659] | 1490 | self.msg.reset() |
---|
| 1491 | self.msg.pack_int( 0 ) |
---|
| 1492 | self.msg.pack_string( typestr ) |
---|
| 1493 | self.msg.pack_string( name ) |
---|
| 1494 | self.msg.pack_string( str( value ) ) |
---|
| 1495 | self.msg.pack_string( unitstr ) |
---|
| 1496 | self.msg.pack_int( self.slope[ slopestr ] ) |
---|
| 1497 | self.msg.pack_uint( int( tmax ) ) |
---|
| 1498 | self.msg.pack_uint( int( dmax ) ) |
---|
[362] | 1499 | |
---|
[659] | 1500 | return self.msg.get_buffer() |
---|
[362] | 1501 | |
---|
[26] | 1502 | def printTime( ): |
---|
[354] | 1503 | |
---|
[659] | 1504 | """Print current time/date in human readable format for log/debug""" |
---|
[26] | 1505 | |
---|
[659] | 1506 | return time.strftime("%a, %d %b %Y %H:%M:%S") |
---|
[26] | 1507 | |
---|
| 1508 | def debug_msg( level, msg ): |
---|
[354] | 1509 | |
---|
[659] | 1510 | """Print msg if at or above current debug level""" |
---|
[26] | 1511 | |
---|
[659] | 1512 | global DAEMONIZE, DEBUG_LEVEL, SYSLOG_LEVEL |
---|
[377] | 1513 | |
---|
[660] | 1514 | if (not DAEMONIZE and DEBUG_LEVEL >= level): |
---|
[659] | 1515 | sys.stderr.write( msg + '\n' ) |
---|
[26] | 1516 | |
---|
[659] | 1517 | if (DAEMONIZE and USE_SYSLOG and SYSLOG_LEVEL >= level): |
---|
| 1518 | syslog.syslog( msg ) |
---|
[373] | 1519 | |
---|
[307] | 1520 | def write_pidfile(): |
---|
| 1521 | |
---|
[659] | 1522 | # Write pidfile if PIDFILE is set |
---|
| 1523 | # |
---|
| 1524 | if PIDFILE: |
---|
[307] | 1525 | |
---|
[659] | 1526 | pid = os.getpid() |
---|
[354] | 1527 | |
---|
[659] | 1528 | pidfile = open( PIDFILE, 'w' ) |
---|
[354] | 1529 | |
---|
[659] | 1530 | pidfile.write( str( pid ) ) |
---|
| 1531 | pidfile.close() |
---|
[307] | 1532 | |
---|
[23] | 1533 | def main(): |
---|
[354] | 1534 | |
---|
[659] | 1535 | """Application start""" |
---|
[23] | 1536 | |
---|
[659] | 1537 | global PBSQuery, PBSError, lsfObject |
---|
| 1538 | global SYSLOG_FACILITY, USE_SYSLOG, BATCH_API, DAEMONIZE |
---|
[256] | 1539 | |
---|
[659] | 1540 | if not processArgs( sys.argv[1:] ): |
---|
[354] | 1541 | |
---|
[659] | 1542 | sys.exit( 1 ) |
---|
[212] | 1543 | |
---|
[659] | 1544 | # Load appropriate DataGatherer depending on which BATCH_API is set |
---|
| 1545 | # and any required modules for the Gatherer |
---|
| 1546 | # |
---|
| 1547 | if BATCH_API == 'pbs': |
---|
[256] | 1548 | |
---|
[659] | 1549 | try: |
---|
| 1550 | from PBSQuery import PBSQuery, PBSError |
---|
[256] | 1551 | |
---|
[659] | 1552 | except ImportError: |
---|
[256] | 1553 | |
---|
[659] | 1554 | debug_msg( 0, "FATAL ERROR: BATCH_API set to 'pbs' but python module 'pbs_python' is not installed" ) |
---|
| 1555 | sys.exit( 1 ) |
---|
[256] | 1556 | |
---|
[659] | 1557 | gather = PbsDataGatherer() |
---|
[256] | 1558 | |
---|
[659] | 1559 | elif BATCH_API == 'sge': |
---|
[256] | 1560 | |
---|
[659] | 1561 | # Tested with SGE 6.0u11. |
---|
| 1562 | # |
---|
| 1563 | gather = SgeDataGatherer() |
---|
[368] | 1564 | |
---|
[659] | 1565 | elif BATCH_API == 'lsf': |
---|
[368] | 1566 | |
---|
[659] | 1567 | try: |
---|
| 1568 | from lsfObject import lsfObject |
---|
| 1569 | except: |
---|
| 1570 | debug_msg(0, "fatal error: BATCH_API set to 'lsf' but python module is not found or installed") |
---|
| 1571 | sys.exit( 1) |
---|
[256] | 1572 | |
---|
[659] | 1573 | gather = LsfDataGatherer() |
---|
[524] | 1574 | |
---|
[659] | 1575 | else: |
---|
| 1576 | debug_msg( 0, "FATAL ERROR: unknown BATCH_API '" + BATCH_API + "' is not supported" ) |
---|
[354] | 1577 | |
---|
[659] | 1578 | sys.exit( 1 ) |
---|
[256] | 1579 | |
---|
[659] | 1580 | if( DAEMONIZE and USE_SYSLOG ): |
---|
[373] | 1581 | |
---|
[659] | 1582 | syslog.openlog( 'jobmond', syslog.LOG_NOWAIT, SYSLOG_FACILITY ) |
---|
[373] | 1583 | |
---|
[659] | 1584 | if DAEMONIZE: |
---|
[354] | 1585 | |
---|
[659] | 1586 | gather.daemon() |
---|
| 1587 | else: |
---|
| 1588 | gather.run() |
---|
[23] | 1589 | |
---|
[256] | 1590 | # wh00t? someone started me! :) |
---|
[65] | 1591 | # |
---|
[23] | 1592 | if __name__ == '__main__': |
---|
[659] | 1593 | main() |
---|