1 | #!/usr/bin/env python |
---|
2 | # |
---|
3 | # This file is part of Jobmonarch |
---|
4 | # |
---|
5 | # Copyright (C) 2006-2007 Ramon Bastiaans |
---|
6 | # Copyright (C) 2007 Dave Love (SGE code) |
---|
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 | # |
---|
22 | # SVN $Id: jobmond.py 514 2008-03-08 20:16:32Z bastiaans $ |
---|
23 | # |
---|
24 | |
---|
25 | import sys, getopt, ConfigParser, time, os, socket, string, re |
---|
26 | import xdrlib, socket, syslog, xml, xml.sax |
---|
27 | from xml.sax.handler import feature_namespaces |
---|
28 | |
---|
29 | VERSION='0.3.1' |
---|
30 | |
---|
31 | def usage( ver ): |
---|
32 | |
---|
33 | print 'jobmond %s' %VERSION |
---|
34 | |
---|
35 | if ver: |
---|
36 | return 0 |
---|
37 | |
---|
38 | print |
---|
39 | print 'Purpose:' |
---|
40 | print ' The Job Monitoring Daemon (jobmond) reports batch jobs information and statistics' |
---|
41 | print ' to Ganglia, which can be viewed with Job Monarch web frontend' |
---|
42 | print |
---|
43 | print 'Usage: jobmond [OPTIONS]' |
---|
44 | print |
---|
45 | print ' -c, --config=FILE The configuration file to use (default: /etc/jobmond.conf)' |
---|
46 | print ' -p, --pidfile=FILE Use pid file to store the process id' |
---|
47 | print ' -h, --help Print help and exit' |
---|
48 | print ' -v, --version Print version and exit' |
---|
49 | print |
---|
50 | |
---|
51 | def processArgs( args ): |
---|
52 | |
---|
53 | SHORT_L = 'p:hvc:' |
---|
54 | LONG_L = [ 'help', 'config=', 'pidfile=', 'version' ] |
---|
55 | |
---|
56 | global PIDFILE |
---|
57 | PIDFILE = None |
---|
58 | |
---|
59 | config_filename = '/etc/jobmond.conf' |
---|
60 | |
---|
61 | try: |
---|
62 | |
---|
63 | opts, args = getopt.getopt( args, SHORT_L, LONG_L ) |
---|
64 | |
---|
65 | except getopt.GetoptError, detail: |
---|
66 | |
---|
67 | print detail |
---|
68 | usage() |
---|
69 | sys.exit( 1 ) |
---|
70 | |
---|
71 | for opt, value in opts: |
---|
72 | |
---|
73 | if opt in [ '--config', '-c' ]: |
---|
74 | |
---|
75 | config_filename = value |
---|
76 | |
---|
77 | if opt in [ '--pidfile', '-p' ]: |
---|
78 | |
---|
79 | PIDFILE = value |
---|
80 | |
---|
81 | if opt in [ '--help', '-h' ]: |
---|
82 | |
---|
83 | usage( False ) |
---|
84 | sys.exit( 0 ) |
---|
85 | |
---|
86 | if opt in [ '--version', '-v' ]: |
---|
87 | |
---|
88 | usage( True ) |
---|
89 | sys.exit( 0 ) |
---|
90 | |
---|
91 | return loadConfig( config_filename ) |
---|
92 | |
---|
93 | def loadConfig( filename ): |
---|
94 | |
---|
95 | def getlist( cfg_string ): |
---|
96 | |
---|
97 | my_list = [ ] |
---|
98 | |
---|
99 | for item_txt in cfg_string.split( ',' ): |
---|
100 | |
---|
101 | sep_char = None |
---|
102 | |
---|
103 | item_txt = item_txt.strip() |
---|
104 | |
---|
105 | for s_char in [ "'", '"' ]: |
---|
106 | |
---|
107 | if item_txt.find( s_char ) != -1: |
---|
108 | |
---|
109 | if item_txt.count( s_char ) != 2: |
---|
110 | |
---|
111 | print 'Missing quote: %s' %item_txt |
---|
112 | sys.exit( 1 ) |
---|
113 | |
---|
114 | else: |
---|
115 | |
---|
116 | sep_char = s_char |
---|
117 | break |
---|
118 | |
---|
119 | if sep_char: |
---|
120 | |
---|
121 | item_txt = item_txt.split( sep_char )[1] |
---|
122 | |
---|
123 | my_list.append( item_txt ) |
---|
124 | |
---|
125 | return my_list |
---|
126 | |
---|
127 | cfg = ConfigParser.ConfigParser() |
---|
128 | |
---|
129 | cfg.read( filename ) |
---|
130 | |
---|
131 | global DEBUG_LEVEL, DAEMONIZE, BATCH_SERVER, BATCH_POLL_INTERVAL |
---|
132 | global GMOND_CONF, DETECT_TIME_DIFFS, BATCH_HOST_TRANSLATE |
---|
133 | global BATCH_API, QUEUE, GMETRIC_TARGET, USE_SYSLOG |
---|
134 | global SYSLOG_LEVEL, SYSLOG_FACILITY, GMETRIC_BINARY |
---|
135 | |
---|
136 | DEBUG_LEVEL = cfg.getint( 'DEFAULT', 'DEBUG_LEVEL' ) |
---|
137 | |
---|
138 | DAEMONIZE = cfg.getboolean( 'DEFAULT', 'DAEMONIZE' ) |
---|
139 | |
---|
140 | SYSLOG_LEVEL = -1 |
---|
141 | SYSLOG_FACILITY = None |
---|
142 | |
---|
143 | try: |
---|
144 | USE_SYSLOG = cfg.getboolean( 'DEFAULT', 'USE_SYSLOG' ) |
---|
145 | |
---|
146 | except ConfigParser.NoOptionError: |
---|
147 | |
---|
148 | USE_SYSLOG = True |
---|
149 | |
---|
150 | debug_msg( 0, 'ERROR: no option USE_SYSLOG found: assuming yes' ) |
---|
151 | |
---|
152 | |
---|
153 | |
---|
154 | if USE_SYSLOG: |
---|
155 | |
---|
156 | try: |
---|
157 | SYSLOG_LEVEL = cfg.getint( 'DEFAULT', 'SYSLOG_LEVEL' ) |
---|
158 | |
---|
159 | except ConfigParser.NoOptionError: |
---|
160 | |
---|
161 | debug_msg( 0, 'ERROR: no option SYSLOG_LEVEL found: assuming level 0' ) |
---|
162 | SYSLOG_LEVEL = 0 |
---|
163 | |
---|
164 | try: |
---|
165 | |
---|
166 | SYSLOG_FACILITY = eval( 'syslog.LOG_' + cfg.get( 'DEFAULT', 'SYSLOG_FACILITY' ) ) |
---|
167 | |
---|
168 | except ConfigParser.NoOptionError: |
---|
169 | |
---|
170 | SYSLOG_FACILITY = syslog.LOG_DAEMON |
---|
171 | |
---|
172 | debug_msg( 0, 'ERROR: no option SYSLOG_FACILITY found: assuming facility DAEMON' ) |
---|
173 | |
---|
174 | try: |
---|
175 | |
---|
176 | BATCH_SERVER = cfg.get( 'DEFAULT', 'BATCH_SERVER' ) |
---|
177 | |
---|
178 | except ConfigParser.NoOptionError: |
---|
179 | |
---|
180 | # Backwards compatibility for old configs |
---|
181 | # |
---|
182 | |
---|
183 | BATCH_SERVER = cfg.get( 'DEFAULT', 'TORQUE_SERVER' ) |
---|
184 | api_guess = 'pbs' |
---|
185 | |
---|
186 | try: |
---|
187 | |
---|
188 | BATCH_POLL_INTERVAL = cfg.getint( 'DEFAULT', 'BATCH_POLL_INTERVAL' ) |
---|
189 | |
---|
190 | except ConfigParser.NoOptionError: |
---|
191 | |
---|
192 | # Backwards compatibility for old configs |
---|
193 | # |
---|
194 | |
---|
195 | BATCH_POLL_INTERVAL = cfg.getint( 'DEFAULT', 'TORQUE_POLL_INTERVAL' ) |
---|
196 | api_guess = 'pbs' |
---|
197 | |
---|
198 | try: |
---|
199 | |
---|
200 | GMOND_CONF = cfg.get( 'DEFAULT', 'GMOND_CONF' ) |
---|
201 | |
---|
202 | except ConfigParser.NoOptionError: |
---|
203 | |
---|
204 | GMOND_CONF = None |
---|
205 | |
---|
206 | try: |
---|
207 | |
---|
208 | GMETRIC_BINARY = cfg.get( 'DEFAULT', 'GMETRIC_BINARY' ) |
---|
209 | |
---|
210 | except ConfigParser.NoOptionError: |
---|
211 | |
---|
212 | GMETRIC_BINARY = '/usr/bin/gmetric' |
---|
213 | |
---|
214 | DETECT_TIME_DIFFS = cfg.getboolean( 'DEFAULT', 'DETECT_TIME_DIFFS' ) |
---|
215 | |
---|
216 | BATCH_HOST_TRANSLATE = getlist( cfg.get( 'DEFAULT', 'BATCH_HOST_TRANSLATE' ) ) |
---|
217 | |
---|
218 | try: |
---|
219 | |
---|
220 | BATCH_API = cfg.get( 'DEFAULT', 'BATCH_API' ) |
---|
221 | |
---|
222 | except ConfigParser.NoOptionError, detail: |
---|
223 | |
---|
224 | if BATCH_SERVER and api_guess: |
---|
225 | |
---|
226 | BATCH_API = api_guess |
---|
227 | else: |
---|
228 | debug_msg( 0, "FATAL ERROR: BATCH_API not set and can't make guess" ) |
---|
229 | sys.exit( 1 ) |
---|
230 | |
---|
231 | try: |
---|
232 | |
---|
233 | QUEUE = getlist( cfg.get( 'DEFAULT', 'QUEUE' ) ) |
---|
234 | |
---|
235 | except ConfigParser.NoOptionError, detail: |
---|
236 | |
---|
237 | QUEUE = None |
---|
238 | |
---|
239 | try: |
---|
240 | |
---|
241 | GMETRIC_TARGET = cfg.get( 'DEFAULT', 'GMETRIC_TARGET' ) |
---|
242 | |
---|
243 | except ConfigParser.NoOptionError: |
---|
244 | |
---|
245 | GMETRIC_TARGET = None |
---|
246 | |
---|
247 | if not GMOND_CONF: |
---|
248 | |
---|
249 | debug_msg( 0, "FATAL ERROR: GMETRIC_TARGET and GMOND_CONF both not set! Set at least one!" ) |
---|
250 | sys.exit( 1 ) |
---|
251 | else: |
---|
252 | |
---|
253 | 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!" ) |
---|
254 | |
---|
255 | return True |
---|
256 | |
---|
257 | def fqdn_parts (fqdn): |
---|
258 | """Return pair of host and domain for fully-qualified domain name arg.""" |
---|
259 | parts = fqdn.split (".") |
---|
260 | return (parts[0], string.join(parts[1:], ".")) |
---|
261 | |
---|
262 | METRIC_MAX_VAL_LEN = 900 |
---|
263 | |
---|
264 | class DataProcessor: |
---|
265 | |
---|
266 | """Class for processing of data""" |
---|
267 | |
---|
268 | binary = None |
---|
269 | |
---|
270 | def __init__( self, binary=None ): |
---|
271 | |
---|
272 | """Remember alternate binary location if supplied""" |
---|
273 | |
---|
274 | global GMETRIC_BINARY |
---|
275 | |
---|
276 | if binary: |
---|
277 | self.binary = binary |
---|
278 | |
---|
279 | if not self.binary: |
---|
280 | self.binary = GMETRIC_BINARY |
---|
281 | |
---|
282 | # Timeout for XML |
---|
283 | # |
---|
284 | # From ganglia's documentation: |
---|
285 | # |
---|
286 | # 'A metric will be deleted DMAX seconds after it is received, and |
---|
287 | # DMAX=0 means eternal life.' |
---|
288 | |
---|
289 | self.dmax = str( int( int( BATCH_POLL_INTERVAL ) * 2 ) ) |
---|
290 | |
---|
291 | if GMOND_CONF: |
---|
292 | |
---|
293 | try: |
---|
294 | gmond_file = GMOND_CONF |
---|
295 | |
---|
296 | except NameError: |
---|
297 | gmond_file = '/etc/gmond.conf' |
---|
298 | |
---|
299 | if not os.path.exists( gmond_file ): |
---|
300 | debug_msg( 0, 'FATAL ERROR: ' + gmond_file + ' does not exist' ) |
---|
301 | sys.exit( 1 ) |
---|
302 | |
---|
303 | incompatible = self.checkGmetricVersion() |
---|
304 | |
---|
305 | if incompatible: |
---|
306 | |
---|
307 | debug_msg( 0, 'Gmetric version not compatible, please upgrade to at least 3.0.1' ) |
---|
308 | sys.exit( 1 ) |
---|
309 | |
---|
310 | def checkGmetricVersion( self ): |
---|
311 | |
---|
312 | """ |
---|
313 | Check version of gmetric is at least 3.0.1 |
---|
314 | for the syntax we use |
---|
315 | """ |
---|
316 | |
---|
317 | global METRIC_MAX_VAL_LEN |
---|
318 | |
---|
319 | incompatible = 0 |
---|
320 | |
---|
321 | gfp = os.popen( self.binary + ' --version' ) |
---|
322 | lines = gfp.readlines() |
---|
323 | |
---|
324 | gfp.close() |
---|
325 | |
---|
326 | for line in lines: |
---|
327 | |
---|
328 | line = line.split( ' ' ) |
---|
329 | |
---|
330 | if len( line ) == 2 and str( line ).find( 'gmetric' ) != -1: |
---|
331 | |
---|
332 | gmetric_version = line[1].split( '\n' )[0] |
---|
333 | |
---|
334 | version_major = int( gmetric_version.split( '.' )[0] ) |
---|
335 | version_minor = int( gmetric_version.split( '.' )[1] ) |
---|
336 | version_patch = int( gmetric_version.split( '.' )[2] ) |
---|
337 | |
---|
338 | incompatible = 0 |
---|
339 | |
---|
340 | if version_major < 3: |
---|
341 | |
---|
342 | incompatible = 1 |
---|
343 | |
---|
344 | elif version_major == 3: |
---|
345 | |
---|
346 | if version_minor == 0: |
---|
347 | |
---|
348 | if version_patch < 1: |
---|
349 | |
---|
350 | incompatible = 1 |
---|
351 | |
---|
352 | # Gmetric 3.0.1 >< 3.0.3 had a bug in the max metric length |
---|
353 | # |
---|
354 | if version_patch < 3: |
---|
355 | |
---|
356 | METRIC_MAX_VAL_LEN = 900 |
---|
357 | |
---|
358 | elif version_patch >= 3: |
---|
359 | |
---|
360 | METRIC_MAX_VAL_LEN = 1400 |
---|
361 | |
---|
362 | return incompatible |
---|
363 | |
---|
364 | def multicastGmetric( self, metricname, metricval, valtype='string', units='' ): |
---|
365 | |
---|
366 | """Call gmetric binary and multicast""" |
---|
367 | |
---|
368 | cmd = self.binary |
---|
369 | |
---|
370 | if GMETRIC_TARGET: |
---|
371 | |
---|
372 | GMETRIC_TARGET_HOST = GMETRIC_TARGET.split( ':' )[0] |
---|
373 | GMETRIC_TARGET_PORT = GMETRIC_TARGET.split( ':' )[1] |
---|
374 | |
---|
375 | metric_debug = "[gmetric] name: %s - val: %s - dmax: %s" %( str( metricname ), str( metricval ), str( self.dmax ) ) |
---|
376 | |
---|
377 | debug_msg( 10, printTime() + ' ' + metric_debug) |
---|
378 | |
---|
379 | gm = Gmetric( GMETRIC_TARGET_HOST, GMETRIC_TARGET_PORT ) |
---|
380 | |
---|
381 | gm.send( str( metricname ), str( metricval ), str( self.dmax ), valtype, units ) |
---|
382 | |
---|
383 | else: |
---|
384 | try: |
---|
385 | cmd = cmd + ' -c' + GMOND_CONF |
---|
386 | |
---|
387 | except NameError: |
---|
388 | |
---|
389 | debug_msg( 10, 'Assuming /etc/gmond.conf for gmetric cmd (omitting)' ) |
---|
390 | |
---|
391 | cmd = cmd + ' -n' + str( metricname )+ ' -v"' + str( metricval )+ '" -t' + str( valtype ) + ' -d' + str( self.dmax ) |
---|
392 | |
---|
393 | if len( units ) > 0: |
---|
394 | |
---|
395 | cmd = cmd + ' -u"' + units + '"' |
---|
396 | |
---|
397 | debug_msg( 10, printTime() + ' ' + cmd ) |
---|
398 | |
---|
399 | os.system( cmd ) |
---|
400 | |
---|
401 | class DataGatherer: |
---|
402 | |
---|
403 | """Skeleton class for batch system DataGatherer""" |
---|
404 | |
---|
405 | def printJobs( self, jobs ): |
---|
406 | |
---|
407 | """Print a jobinfo overview""" |
---|
408 | |
---|
409 | for name, attrs in self.jobs.items(): |
---|
410 | |
---|
411 | print 'job %s' %(name) |
---|
412 | |
---|
413 | for name, val in attrs.items(): |
---|
414 | |
---|
415 | print '\t%s = %s' %( name, val ) |
---|
416 | |
---|
417 | def printJob( self, jobs, job_id ): |
---|
418 | |
---|
419 | """Print job with job_id from jobs""" |
---|
420 | |
---|
421 | print 'job %s' %(job_id) |
---|
422 | |
---|
423 | for name, val in jobs[ job_id ].items(): |
---|
424 | |
---|
425 | print '\t%s = %s' %( name, val ) |
---|
426 | |
---|
427 | def getAttr( self, attrs, name ): |
---|
428 | |
---|
429 | """Return certain attribute from dictionary, if exists""" |
---|
430 | |
---|
431 | if attrs.has_key( name ): |
---|
432 | |
---|
433 | return attrs[ name ] |
---|
434 | else: |
---|
435 | return '' |
---|
436 | |
---|
437 | def jobDataChanged( self, jobs, job_id, attrs ): |
---|
438 | |
---|
439 | """Check if job with attrs and job_id in jobs has changed""" |
---|
440 | |
---|
441 | if jobs.has_key( job_id ): |
---|
442 | |
---|
443 | oldData = jobs[ job_id ] |
---|
444 | else: |
---|
445 | return 1 |
---|
446 | |
---|
447 | for name, val in attrs.items(): |
---|
448 | |
---|
449 | if oldData.has_key( name ): |
---|
450 | |
---|
451 | if oldData[ name ] != attrs[ name ]: |
---|
452 | |
---|
453 | return 1 |
---|
454 | |
---|
455 | else: |
---|
456 | return 1 |
---|
457 | |
---|
458 | return 0 |
---|
459 | |
---|
460 | def submitJobData( self ): |
---|
461 | |
---|
462 | """Submit job info list""" |
---|
463 | |
---|
464 | global BATCH_API |
---|
465 | |
---|
466 | self.dp.multicastGmetric( 'MONARCH-HEARTBEAT', str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
467 | |
---|
468 | running_jobs = 0 |
---|
469 | queued_jobs = 0 |
---|
470 | |
---|
471 | # Count how many running/queued jobs we found |
---|
472 | # |
---|
473 | for jobid, jobattrs in self.jobs.items(): |
---|
474 | |
---|
475 | if jobattrs[ 'status' ] == 'Q': |
---|
476 | |
---|
477 | queued_jobs += 1 |
---|
478 | |
---|
479 | elif jobattrs[ 'status' ] == 'R': |
---|
480 | |
---|
481 | running_jobs += 1 |
---|
482 | |
---|
483 | # Report running/queued jobs as seperate metric for a nice RRD graph |
---|
484 | # |
---|
485 | self.dp.multicastGmetric( 'MONARCH-RJ', str( running_jobs ), 'uint32', 'jobs' ) |
---|
486 | self.dp.multicastGmetric( 'MONARCH-QJ', str( queued_jobs ), 'uint32', 'jobs' ) |
---|
487 | |
---|
488 | # Report down/offline nodes in batch (PBS only ATM) |
---|
489 | # |
---|
490 | if BATCH_API == 'pbs': |
---|
491 | |
---|
492 | domain = fqdn_parts( socket.getfqdn() )[1] |
---|
493 | |
---|
494 | downed_nodes = list() |
---|
495 | offline_nodes = list() |
---|
496 | |
---|
497 | l = ['state'] |
---|
498 | |
---|
499 | for name, node in self.pq.getnodes().items(): |
---|
500 | |
---|
501 | if ( node[ 'state' ].find( "down" ) != -1 ): |
---|
502 | |
---|
503 | downed_nodes.append( name ) |
---|
504 | |
---|
505 | if ( node[ 'state' ].find( "offline" ) != -1 ): |
---|
506 | |
---|
507 | offline_nodes.append( name ) |
---|
508 | |
---|
509 | downnodeslist = do_nodelist( downed_nodes ) |
---|
510 | offlinenodeslist = do_nodelist( offline_nodes ) |
---|
511 | |
---|
512 | down_str = 'nodes=%s domain=%s reported=%s' %( string.join( downnodeslist, ';' ), domain, str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
513 | offl_str = 'nodes=%s domain=%s reported=%s' %( string.join( offlinenodeslist, ';' ), domain, str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) ) |
---|
514 | self.dp.multicastGmetric( 'MONARCH-DOWN' , down_str ) |
---|
515 | self.dp.multicastGmetric( 'MONARCH-OFFLINE', offl_str ) |
---|
516 | |
---|
517 | # Now let's spread the knowledge |
---|
518 | # |
---|
519 | for jobid, jobattrs in self.jobs.items(): |
---|
520 | |
---|
521 | # Make gmetric values for each job: respect max gmetric value length |
---|
522 | # |
---|
523 | gmetric_val = self.compileGmetricVal( jobid, jobattrs ) |
---|
524 | metric_increment = 0 |
---|
525 | |
---|
526 | # If we have more job info than max gmetric value length allows, split it up |
---|
527 | # amongst multiple metrics |
---|
528 | # |
---|
529 | for val in gmetric_val: |
---|
530 | |
---|
531 | self.dp.multicastGmetric( 'MONARCH-JOB-' + jobid + '-' + str(metric_increment), val ) |
---|
532 | |
---|
533 | # Increase follow number if this jobinfo is split up amongst more than 1 gmetric |
---|
534 | # |
---|
535 | metric_increment = metric_increment + 1 |
---|
536 | |
---|
537 | def compileGmetricVal( self, jobid, jobattrs ): |
---|
538 | |
---|
539 | """Create a val string for gmetric of jobinfo""" |
---|
540 | |
---|
541 | gval_lists = [ ] |
---|
542 | val_list = { } |
---|
543 | |
---|
544 | for val_name, val_value in jobattrs.items(): |
---|
545 | |
---|
546 | # These are our own metric names, i.e.: status, start_timestamp, etc |
---|
547 | # |
---|
548 | val_list_names_len = len( string.join( val_list.keys() ) ) + len(val_list.keys()) |
---|
549 | |
---|
550 | # These are their corresponding values |
---|
551 | # |
---|
552 | val_list_vals_len = len( string.join( val_list.values() ) ) + len(val_list.values()) |
---|
553 | |
---|
554 | if val_name == 'nodes' and jobattrs['status'] == 'R': |
---|
555 | |
---|
556 | node_str = None |
---|
557 | |
---|
558 | for node in val_value: |
---|
559 | |
---|
560 | if node_str: |
---|
561 | |
---|
562 | node_str = node_str + ';' + node |
---|
563 | else: |
---|
564 | node_str = node |
---|
565 | |
---|
566 | # Make sure if we add this new info, that the total metric's value length does not exceed METRIC_MAX_VAL_LEN |
---|
567 | # |
---|
568 | if (val_list_names_len + len(val_name) ) + (val_list_vals_len + len(node_str) ) > METRIC_MAX_VAL_LEN: |
---|
569 | |
---|
570 | # It's too big, we need to make a new gmetric for the additional info |
---|
571 | # |
---|
572 | val_list[ val_name ] = node_str |
---|
573 | |
---|
574 | gval_lists.append( val_list ) |
---|
575 | |
---|
576 | val_list = { } |
---|
577 | node_str = None |
---|
578 | |
---|
579 | val_list[ val_name ] = node_str |
---|
580 | |
---|
581 | gval_lists.append( val_list ) |
---|
582 | |
---|
583 | val_list = { } |
---|
584 | |
---|
585 | elif val_value != '': |
---|
586 | |
---|
587 | # Make sure if we add this new info, that the total metric's value length does not exceed METRIC_MAX_VAL_LEN |
---|
588 | # |
---|
589 | if (val_list_names_len + len(val_name) ) + (val_list_vals_len + len(str(val_value)) ) > METRIC_MAX_VAL_LEN: |
---|
590 | |
---|
591 | # It's too big, we need to make a new gmetric for the additional info |
---|
592 | # |
---|
593 | gval_lists.append( val_list ) |
---|
594 | |
---|
595 | val_list = { } |
---|
596 | |
---|
597 | val_list[ val_name ] = val_value |
---|
598 | |
---|
599 | if len( val_list ) > 0: |
---|
600 | |
---|
601 | gval_lists.append( val_list ) |
---|
602 | |
---|
603 | str_list = [ ] |
---|
604 | |
---|
605 | # Now append the value names and values together, i.e.: stop_timestamp=value, etc |
---|
606 | # |
---|
607 | for val_list in gval_lists: |
---|
608 | |
---|
609 | my_val_str = None |
---|
610 | |
---|
611 | for val_name, val_value in val_list.items(): |
---|
612 | |
---|
613 | if my_val_str: |
---|
614 | |
---|
615 | my_val_str = my_val_str + ' ' + val_name + '=' + val_value |
---|
616 | else: |
---|
617 | my_val_str = val_name + '=' + val_value |
---|
618 | |
---|
619 | str_list.append( my_val_str ) |
---|
620 | |
---|
621 | return str_list |
---|
622 | |
---|
623 | def daemon( self ): |
---|
624 | |
---|
625 | """Run as daemon forever""" |
---|
626 | |
---|
627 | # Fork the first child |
---|
628 | # |
---|
629 | pid = os.fork() |
---|
630 | if pid > 0: |
---|
631 | sys.exit(0) # end parent |
---|
632 | |
---|
633 | # creates a session and sets the process group ID |
---|
634 | # |
---|
635 | os.setsid() |
---|
636 | |
---|
637 | # Fork the second child |
---|
638 | # |
---|
639 | pid = os.fork() |
---|
640 | if pid > 0: |
---|
641 | sys.exit(0) # end parent |
---|
642 | |
---|
643 | write_pidfile() |
---|
644 | |
---|
645 | # Go to the root directory and set the umask |
---|
646 | # |
---|
647 | os.chdir('/') |
---|
648 | os.umask(0) |
---|
649 | |
---|
650 | sys.stdin.close() |
---|
651 | sys.stdout.close() |
---|
652 | sys.stderr.close() |
---|
653 | |
---|
654 | os.open('/dev/null', os.O_RDWR) |
---|
655 | os.dup2(0, 1) |
---|
656 | os.dup2(0, 2) |
---|
657 | |
---|
658 | self.run() |
---|
659 | |
---|
660 | def run( self ): |
---|
661 | |
---|
662 | """Main thread""" |
---|
663 | |
---|
664 | while ( 1 ): |
---|
665 | |
---|
666 | self.getJobData() |
---|
667 | self.submitJobData() |
---|
668 | time.sleep( BATCH_POLL_INTERVAL ) |
---|
669 | |
---|
670 | # SGE code by Dave Love <fx@gnu.org>. Tested with SGE 6.0u8 and 6.0u11. |
---|
671 | # Probably needs modification for SGE 6.1. See also the fixmes. |
---|
672 | |
---|
673 | class NoJobs (Exception): |
---|
674 | """Exception raised by empty job list in qstat output.""" |
---|
675 | pass |
---|
676 | |
---|
677 | class SgeQstatXMLParser(xml.sax.handler.ContentHandler): |
---|
678 | """SAX handler for XML output from Sun Grid Engine's `qstat'.""" |
---|
679 | |
---|
680 | def __init__(self): |
---|
681 | self.value = "" |
---|
682 | self.joblist = [] |
---|
683 | self.job = {} |
---|
684 | self.queue = "" |
---|
685 | self.in_joblist = False |
---|
686 | self.lrequest = False |
---|
687 | xml.sax.handler.ContentHandler.__init__(self) |
---|
688 | |
---|
689 | # The structure of the output is as follows. Unfortunately |
---|
690 | # it's voluminous, and probably doesn't scale to large |
---|
691 | # clusters/queues. |
---|
692 | |
---|
693 | # <detailed_job_info xmlns:xsd="http://www.w3.org/2001/XMLSchema"> |
---|
694 | # <djob_info> |
---|
695 | # <qmaster_response> <!-- job --> |
---|
696 | # ... |
---|
697 | # <JB_ja_template> |
---|
698 | # <ulong_sublist> |
---|
699 | # ... <!-- start_time, state ... --> |
---|
700 | # </ulong_sublist> |
---|
701 | # </JB_ja_template> |
---|
702 | # <JB_ja_tasks> |
---|
703 | # <ulong_sublist> |
---|
704 | # ... <!-- task info |
---|
705 | # </ulong_sublist> |
---|
706 | # ... |
---|
707 | # </JB_ja_tasks> |
---|
708 | # ... |
---|
709 | # </qmaster_response> |
---|
710 | # </djob_info> |
---|
711 | # <messages> |
---|
712 | # ... |
---|
713 | |
---|
714 | # NB. We might treat each task as a separate job, like |
---|
715 | # straight qstat output, but the web interface expects jobs to |
---|
716 | # be identified by integers, not, say, <job number>.<task>. |
---|
717 | |
---|
718 | # So, I lied. If the job list is empty, we get invalid XML |
---|
719 | # like this, which we need to defend against: |
---|
720 | |
---|
721 | # <unknown_jobs xmlns:xsd="http://www.w3.org/2001/XMLSchema"> |
---|
722 | # <> |
---|
723 | # <ST_name>*</ST_name> |
---|
724 | # </> |
---|
725 | # </unknown_jobs> |
---|
726 | |
---|
727 | def startElement(self, name, attrs): |
---|
728 | self.value = "" |
---|
729 | if name == "djob_info": # job list |
---|
730 | self.in_joblist = True |
---|
731 | elif name == "qmaster_response" and self.in_joblist: # job |
---|
732 | self.job = {"job_state": "U", "slots": 0, |
---|
733 | "nodes": [], "queued_timestamp": "", |
---|
734 | "queued_timestamp": "", "queue": "", |
---|
735 | "ppn": "0", "RN_max": 0, |
---|
736 | # fixme in endElement |
---|
737 | "requested_memory": 0, "requested_time": 0 |
---|
738 | } |
---|
739 | self.joblist.append(self.job) |
---|
740 | elif name == "qstat_l_requests": # resource request |
---|
741 | self.lrequest = True |
---|
742 | elif name == "unknown_jobs": |
---|
743 | raise NoJobs |
---|
744 | |
---|
745 | def characters(self, ch): |
---|
746 | self.value += ch |
---|
747 | |
---|
748 | def endElement(self, name): |
---|
749 | """Snarf job elements contents into job dictionary. |
---|
750 | Translate keys if appropriate.""" |
---|
751 | |
---|
752 | name_trans = { |
---|
753 | "JB_job_number": "number", |
---|
754 | "JB_job_name": "name", "JB_owner": "owner", |
---|
755 | "queue_name": "queue", "JAT_start_time": "start_timestamp", |
---|
756 | "JB_submission_time": "queued_timestamp" |
---|
757 | } |
---|
758 | value = self.value |
---|
759 | |
---|
760 | if name == "djob_info": |
---|
761 | self.in_joblist = False |
---|
762 | self.job = {} |
---|
763 | elif name == "JAT_master_queue": |
---|
764 | self.job["queue"] = value.split("@")[0] |
---|
765 | elif name == "JG_qhostname": |
---|
766 | if not (value in self.job["nodes"]): |
---|
767 | self.job["nodes"].append(value) |
---|
768 | elif name == "JG_slots": # slots in use |
---|
769 | self.job["slots"] += int(value) |
---|
770 | elif name == "RN_max": # requested slots (tasks or parallel) |
---|
771 | self.job["RN_max"] = max (self.job["RN_max"], |
---|
772 | int(value)) |
---|
773 | elif name == "JAT_state": # job state (bitwise or) |
---|
774 | value = int (value) |
---|
775 | # Status values from sge_jobL.h |
---|
776 | #define JIDLE 0x00000000 |
---|
777 | #define JHELD 0x00000010 |
---|
778 | #define JMIGRATING 0x00000020 |
---|
779 | #define JQUEUED 0x00000040 |
---|
780 | #define JRUNNING 0x00000080 |
---|
781 | #define JSUSPENDED 0x00000100 |
---|
782 | #define JTRANSFERING 0x00000200 |
---|
783 | #define JDELETED 0x00000400 |
---|
784 | #define JWAITING 0x00000800 |
---|
785 | #define JEXITING 0x00001000 |
---|
786 | #define JWRITTEN 0x00002000 |
---|
787 | #define JSUSPENDED_ON_THRESHOLD 0x00010000 |
---|
788 | #define JFINISHED 0x00010000 |
---|
789 | if value & 0x80: |
---|
790 | self.job["status"] = "R" |
---|
791 | elif value & 0x40: |
---|
792 | self.job["status"] = "Q" |
---|
793 | else: |
---|
794 | self.job["status"] = "O" # `other' |
---|
795 | elif name == "CE_name" and self.lrequest and self.value in \ |
---|
796 | ("h_cpu", "s_cpu", "cpu", "h_core", "s_core"): |
---|
797 | # We're in a container for an interesting resource |
---|
798 | # request; record which type. |
---|
799 | self.lrequest = self.value |
---|
800 | elif name == "CE_doubleval" and self.lrequest: |
---|
801 | # if we're in a container for an interesting |
---|
802 | # resource request, use the maxmimum of the hard |
---|
803 | # and soft requests to record the requested CPU |
---|
804 | # or core. Fixme: I'm not sure if this logic is |
---|
805 | # right. |
---|
806 | if self.lrequest in ("h_core", "s_core"): |
---|
807 | self.job["requested_memory"] = \ |
---|
808 | max (float (value), |
---|
809 | self.job["requested_memory"]) |
---|
810 | # Fixme: Check what cpu means, c.f [hs]_cpu. |
---|
811 | elif self.lrequest in ("h_cpu", "s_cpu", "cpu"): |
---|
812 | self.job["requested_time"] = \ |
---|
813 | max (float (value), |
---|
814 | self.job["requested_time"]) |
---|
815 | elif name == "qstat_l_requests": |
---|
816 | self.lrequest = False |
---|
817 | elif self.job and self.in_joblist: |
---|
818 | if name in name_trans: |
---|
819 | name = name_trans[name] |
---|
820 | self.job[name] = value |
---|
821 | |
---|
822 | # Abstracted from PBS original. |
---|
823 | # Fixme: Is it worth (or appropriate for PBS) sorting the result? |
---|
824 | def do_nodelist (nodes): |
---|
825 | """Translate node list as appropriate.""" |
---|
826 | nodeslist = [ ] |
---|
827 | my_domain = fqdn_parts(socket.getfqdn())[1] |
---|
828 | for node in nodes: |
---|
829 | host = node.split( '/' )[0] # not relevant for SGE |
---|
830 | h, host_domain = fqdn_parts(host) |
---|
831 | if host_domain == my_domain: |
---|
832 | host = h |
---|
833 | if nodeslist.count( host ) == 0: |
---|
834 | for translate_pattern in BATCH_HOST_TRANSLATE: |
---|
835 | if translate_pattern.find( '/' ) != -1: |
---|
836 | translate_orig = \ |
---|
837 | translate_pattern.split( '/' )[1] |
---|
838 | translate_new = \ |
---|
839 | translate_pattern.split( '/' )[2] |
---|
840 | host = re.sub( translate_orig, |
---|
841 | translate_new, host ) |
---|
842 | if not host in nodeslist: |
---|
843 | nodeslist.append( host ) |
---|
844 | return nodeslist |
---|
845 | |
---|
846 | class SgeDataGatherer(DataGatherer): |
---|
847 | |
---|
848 | jobs = {} |
---|
849 | |
---|
850 | def __init__( self ): |
---|
851 | self.jobs = {} |
---|
852 | self.timeoffset = 0 |
---|
853 | self.dp = DataProcessor() |
---|
854 | |
---|
855 | def getJobData( self ): |
---|
856 | """Gather all data on current jobs in SGE""" |
---|
857 | |
---|
858 | import popen2 |
---|
859 | |
---|
860 | self.cur_time = 0 |
---|
861 | queues = "" |
---|
862 | if QUEUE: # only for specific queues |
---|
863 | # Fixme: assumes queue names don't contain single |
---|
864 | # quote or comma. Don't know what the SGE rules are. |
---|
865 | queues = " -q '" + string.join (QUEUE, ",") + "'" |
---|
866 | # Note the comment in SgeQstatXMLParser about scaling with |
---|
867 | # this method of getting data. I haven't found better one. |
---|
868 | # Output with args `-xml -ext -f -r' is easier to parse |
---|
869 | # in some ways, harder in others, but it doesn't provide |
---|
870 | # the submission time, at least. |
---|
871 | piping = popen2.Popen3("qstat -u '*' -j '*' -xml" + queues, |
---|
872 | True) |
---|
873 | qstatparser = SgeQstatXMLParser() |
---|
874 | parse_err = 0 |
---|
875 | try: |
---|
876 | xml.sax.parse(piping.fromchild, qstatparser) |
---|
877 | except NoJobs: |
---|
878 | pass |
---|
879 | except: |
---|
880 | parse_err = 1 |
---|
881 | if piping.wait(): |
---|
882 | debug_msg(10, |
---|
883 | "qstat error, skipping until next polling interval: " |
---|
884 | + piping.childerr.readline()) |
---|
885 | return None |
---|
886 | elif parse_err: |
---|
887 | debug_msg(10, "Bad XML output from qstat"()) |
---|
888 | exit (1) |
---|
889 | for f in piping.fromchild, piping.tochild, piping.childerr: |
---|
890 | f.close() |
---|
891 | self.cur_time = time.time() |
---|
892 | jobs_processed = [] |
---|
893 | for job in qstatparser.joblist: |
---|
894 | job_id = job["number"] |
---|
895 | if job["status"] in [ 'Q', 'R' ]: |
---|
896 | jobs_processed.append(job_id) |
---|
897 | if job["status"] == "R": |
---|
898 | job["nodes"] = do_nodelist (job["nodes"]) |
---|
899 | # Fixme: Is this right? |
---|
900 | job["ppn"] = float(job["slots"]) / \ |
---|
901 | len(job["nodes"]) |
---|
902 | if DETECT_TIME_DIFFS: |
---|
903 | # If a job start is later than our |
---|
904 | # current date, that must mean |
---|
905 | # the SGE server's time is later |
---|
906 | # than our local time. |
---|
907 | start_timestamp = \ |
---|
908 | int (job["start_timestamp"]) |
---|
909 | if start_timestamp > \ |
---|
910 | int(self.cur_time) + \ |
---|
911 | int(self.timeoffset): |
---|
912 | |
---|
913 | self.timeoffset = \ |
---|
914 | start_timestamp - \ |
---|
915 | int(self.cur_time) |
---|
916 | else: |
---|
917 | # fixme: Note sure what this should be: |
---|
918 | job["ppn"] = job["RN_max"] |
---|
919 | job["nodes"] = "1" |
---|
920 | |
---|
921 | myAttrs = {} |
---|
922 | for attr in ["name", "queue", "owner", |
---|
923 | "requested_time", "status", |
---|
924 | "requested_memory", "ppn", |
---|
925 | "start_timestamp", "queued_timestamp"]: |
---|
926 | myAttrs[attr] = str(job[attr]) |
---|
927 | myAttrs["nodes"] = job["nodes"] |
---|
928 | myAttrs["reported"] = str(int(self.cur_time) + \ |
---|
929 | int(self.timeoffset)) |
---|
930 | myAttrs["domain"] = fqdn_parts(socket.getfqdn())[1] |
---|
931 | myAttrs["poll_interval"] = str(BATCH_POLL_INTERVAL) |
---|
932 | |
---|
933 | if self.jobDataChanged(self.jobs, job_id, myAttrs) \ |
---|
934 | and myAttrs["status"] in ["R", "Q"]: |
---|
935 | self.jobs[job_id] = myAttrs |
---|
936 | for id, attrs in self.jobs.items(): |
---|
937 | if id not in jobs_processed: |
---|
938 | del self.jobs[id] |
---|
939 | |
---|
940 | class PbsDataGatherer( DataGatherer ): |
---|
941 | |
---|
942 | """This is the DataGatherer for PBS and Torque""" |
---|
943 | |
---|
944 | global PBSQuery |
---|
945 | |
---|
946 | def __init__( self ): |
---|
947 | |
---|
948 | """Setup appropriate variables""" |
---|
949 | |
---|
950 | self.jobs = { } |
---|
951 | self.timeoffset = 0 |
---|
952 | self.dp = DataProcessor() |
---|
953 | |
---|
954 | self.initPbsQuery() |
---|
955 | |
---|
956 | def initPbsQuery( self ): |
---|
957 | |
---|
958 | self.pq = None |
---|
959 | |
---|
960 | if( BATCH_SERVER ): |
---|
961 | |
---|
962 | self.pq = PBSQuery( BATCH_SERVER ) |
---|
963 | else: |
---|
964 | self.pq = PBSQuery() |
---|
965 | |
---|
966 | def getJobData( self ): |
---|
967 | |
---|
968 | """Gather all data on current jobs in Torque""" |
---|
969 | |
---|
970 | joblist = {} |
---|
971 | self.cur_time = 0 |
---|
972 | |
---|
973 | try: |
---|
974 | joblist = self.pq.getjobs() |
---|
975 | self.cur_time = time.time() |
---|
976 | |
---|
977 | except PBSError, detail: |
---|
978 | |
---|
979 | debug_msg( 10, "Caught PBS unavailable, skipping until next polling interval: " + str( detail ) ) |
---|
980 | return None |
---|
981 | |
---|
982 | jobs_processed = [ ] |
---|
983 | |
---|
984 | for name, attrs in joblist.items(): |
---|
985 | display_queue = 1 |
---|
986 | job_id = name.split( '.' )[0] |
---|
987 | |
---|
988 | name = self.getAttr( attrs, 'Job_Name' ) |
---|
989 | queue = self.getAttr( attrs, 'queue' ) |
---|
990 | |
---|
991 | if QUEUE: |
---|
992 | for q in QUEUE: |
---|
993 | if q == queue: |
---|
994 | display_queue = 1 |
---|
995 | break |
---|
996 | else: |
---|
997 | display_queue = 0 |
---|
998 | continue |
---|
999 | if display_queue == 0: |
---|
1000 | continue |
---|
1001 | |
---|
1002 | |
---|
1003 | owner = self.getAttr( attrs, 'Job_Owner' ).split( '@' )[0] |
---|
1004 | requested_time = self.getAttr( attrs, 'Resource_List.walltime' ) |
---|
1005 | requested_memory = self.getAttr( attrs, 'Resource_List.mem' ) |
---|
1006 | |
---|
1007 | mynoderequest = self.getAttr( attrs, 'Resource_List.nodes' ) |
---|
1008 | |
---|
1009 | ppn = '' |
---|
1010 | |
---|
1011 | if mynoderequest.find( ':' ) != -1 and mynoderequest.find( 'ppn' ) != -1: |
---|
1012 | |
---|
1013 | mynoderequest_fields = mynoderequest.split( ':' ) |
---|
1014 | |
---|
1015 | for mynoderequest_field in mynoderequest_fields: |
---|
1016 | |
---|
1017 | if mynoderequest_field.find( 'ppn' ) != -1: |
---|
1018 | |
---|
1019 | ppn = mynoderequest_field.split( 'ppn=' )[1] |
---|
1020 | |
---|
1021 | status = self.getAttr( attrs, 'job_state' ) |
---|
1022 | |
---|
1023 | if status in [ 'Q', 'R' ]: |
---|
1024 | |
---|
1025 | jobs_processed.append( job_id ) |
---|
1026 | |
---|
1027 | queued_timestamp = self.getAttr( attrs, 'ctime' ) |
---|
1028 | |
---|
1029 | if status == 'R': |
---|
1030 | |
---|
1031 | start_timestamp = self.getAttr( attrs, 'mtime' ) |
---|
1032 | nodes = self.getAttr( attrs, 'exec_host' ).split( '+' ) |
---|
1033 | |
---|
1034 | nodeslist = do_nodelist( nodes ) |
---|
1035 | |
---|
1036 | if DETECT_TIME_DIFFS: |
---|
1037 | |
---|
1038 | # If a job start if later than our current date, |
---|
1039 | # that must mean the Torque server's time is later |
---|
1040 | # than our local time. |
---|
1041 | |
---|
1042 | if int( start_timestamp ) > int( int( self.cur_time ) + int( self.timeoffset ) ): |
---|
1043 | |
---|
1044 | self.timeoffset = int( int(start_timestamp) - int(self.cur_time) ) |
---|
1045 | |
---|
1046 | elif status == 'Q': |
---|
1047 | |
---|
1048 | # 'mynodequest' can be a string in the following syntax according to the |
---|
1049 | # Torque Administator's manual: |
---|
1050 | # |
---|
1051 | # {<node_count> | <hostname>}[:ppn=<ppn>][:<property>[:<property>]...][+ ...] |
---|
1052 | # {<node_count> | <hostname>}[:ppn=<ppn>][:<property>[:<property>]...][+ ...] |
---|
1053 | # etc |
---|
1054 | # |
---|
1055 | |
---|
1056 | # |
---|
1057 | # For now we only count the amount of nodes request and ignore properties |
---|
1058 | # |
---|
1059 | |
---|
1060 | start_timestamp = '' |
---|
1061 | count_mynodes = 0 |
---|
1062 | |
---|
1063 | for node in mynoderequest.split( '+' ): |
---|
1064 | |
---|
1065 | # Just grab the {node_count|hostname} part and ignore properties |
---|
1066 | # |
---|
1067 | nodepart = node.split( ':' )[0] |
---|
1068 | |
---|
1069 | # Let's assume a node_count value |
---|
1070 | # |
---|
1071 | numeric_node = 1 |
---|
1072 | |
---|
1073 | # Chop the value up into characters |
---|
1074 | # |
---|
1075 | for letter in nodepart: |
---|
1076 | |
---|
1077 | # If this char is not a digit (0-9), this must be a hostname |
---|
1078 | # |
---|
1079 | if letter not in string.digits: |
---|
1080 | |
---|
1081 | numeric_node = 0 |
---|
1082 | |
---|
1083 | # If this is a hostname, just count this as one (1) node |
---|
1084 | # |
---|
1085 | if not numeric_node: |
---|
1086 | |
---|
1087 | count_mynodes = count_mynodes + 1 |
---|
1088 | else: |
---|
1089 | |
---|
1090 | # If this a number, it must be the node_count |
---|
1091 | # and increase our count with it's value |
---|
1092 | # |
---|
1093 | try: |
---|
1094 | count_mynodes = count_mynodes + int( nodepart ) |
---|
1095 | |
---|
1096 | except ValueError, detail: |
---|
1097 | |
---|
1098 | # When we arrive here I must be bugged or very confused |
---|
1099 | # THIS SHOULD NOT HAPPEN! |
---|
1100 | # |
---|
1101 | debug_msg( 10, str( detail ) ) |
---|
1102 | debug_msg( 10, "Encountered weird node in Resources_List?!" ) |
---|
1103 | debug_msg( 10, 'nodepart = ' + str( nodepart ) ) |
---|
1104 | debug_msg( 10, 'job = ' + str( name ) ) |
---|
1105 | debug_msg( 10, 'attrs = ' + str( attrs ) ) |
---|
1106 | |
---|
1107 | nodeslist = str( count_mynodes ) |
---|
1108 | else: |
---|
1109 | start_timestamp = '' |
---|
1110 | nodeslist = '' |
---|
1111 | |
---|
1112 | myAttrs = { } |
---|
1113 | |
---|
1114 | myAttrs[ 'name' ] = str( name ) |
---|
1115 | myAttrs[ 'queue' ] = str( queue ) |
---|
1116 | myAttrs[ 'owner' ] = str( owner ) |
---|
1117 | myAttrs[ 'requested_time' ] = str( requested_time ) |
---|
1118 | myAttrs[ 'requested_memory' ] = str( requested_memory ) |
---|
1119 | myAttrs[ 'ppn' ] = str( ppn ) |
---|
1120 | myAttrs[ 'status' ] = str( status ) |
---|
1121 | myAttrs[ 'start_timestamp' ] = str( start_timestamp ) |
---|
1122 | myAttrs[ 'queued_timestamp' ] = str( queued_timestamp ) |
---|
1123 | myAttrs[ 'reported' ] = str( int( int( self.cur_time ) + int( self.timeoffset ) ) ) |
---|
1124 | myAttrs[ 'nodes' ] = nodeslist |
---|
1125 | myAttrs[ 'domain' ] = fqdn_parts( socket.getfqdn() )[1] |
---|
1126 | myAttrs[ 'poll_interval' ] = str( BATCH_POLL_INTERVAL ) |
---|
1127 | |
---|
1128 | if self.jobDataChanged( self.jobs, job_id, myAttrs ) and myAttrs['status'] in [ 'R', 'Q' ]: |
---|
1129 | |
---|
1130 | self.jobs[ job_id ] = myAttrs |
---|
1131 | |
---|
1132 | for id, attrs in self.jobs.items(): |
---|
1133 | |
---|
1134 | if id not in jobs_processed: |
---|
1135 | |
---|
1136 | # This one isn't there anymore; toedeledoki! |
---|
1137 | # |
---|
1138 | del self.jobs[ id ] |
---|
1139 | |
---|
1140 | # |
---|
1141 | # Gmetric by Nick Galbreath - nickg(a.t)modp(d.o.t)com |
---|
1142 | # Version 1.0 - 21-April2-2007 |
---|
1143 | # http://code.google.com/p/embeddedgmetric/ |
---|
1144 | # |
---|
1145 | # Modified by: Ramon Bastiaans |
---|
1146 | # For the Job Monarch Project, see: https://subtrac.sara.nl/oss/jobmonarch/ |
---|
1147 | # |
---|
1148 | # added: DEFAULT_TYPE for Gmetric's |
---|
1149 | # added: checkHostProtocol to determine if target is multicast or not |
---|
1150 | # changed: allow default for Gmetric constructor |
---|
1151 | # changed: allow defaults for all send() values except dmax |
---|
1152 | # |
---|
1153 | |
---|
1154 | GMETRIC_DEFAULT_TYPE = 'string' |
---|
1155 | GMETRIC_DEFAULT_HOST = '127.0.0.1' |
---|
1156 | GMETRIC_DEFAULT_PORT = '8649' |
---|
1157 | GMETRIC_DEFAULT_UNITS = '' |
---|
1158 | |
---|
1159 | class Gmetric: |
---|
1160 | |
---|
1161 | global GMETRIC_DEFAULT_HOST, GMETRIC_DEFAULT_PORT |
---|
1162 | |
---|
1163 | slope = { 'zero' : 0, 'positive' : 1, 'negative' : 2, 'both' : 3, 'unspecified' : 4 } |
---|
1164 | type = ( '', 'string', 'uint16', 'int16', 'uint32', 'int32', 'float', 'double', 'timestamp' ) |
---|
1165 | protocol = ( 'udp', 'multicast' ) |
---|
1166 | |
---|
1167 | def __init__( self, host=GMETRIC_DEFAULT_HOST, port=GMETRIC_DEFAULT_PORT ): |
---|
1168 | |
---|
1169 | global GMETRIC_DEFAULT_TYPE |
---|
1170 | |
---|
1171 | self.prot = self.checkHostProtocol( host ) |
---|
1172 | self.msg = xdrlib.Packer() |
---|
1173 | self.socket = socket.socket( socket.AF_INET, socket.SOCK_DGRAM ) |
---|
1174 | |
---|
1175 | if self.prot not in self.protocol: |
---|
1176 | |
---|
1177 | raise ValueError( "Protocol must be one of: " + str( self.protocol ) ) |
---|
1178 | |
---|
1179 | if self.prot == 'multicast': |
---|
1180 | |
---|
1181 | # Set multicast options |
---|
1182 | # |
---|
1183 | self.socket.setsockopt( socket.IPPROTO_IP, socket.IP_MULTICAST_TTL, 20 ) |
---|
1184 | |
---|
1185 | self.hostport = ( host, int( port ) ) |
---|
1186 | self.slopestr = 'both' |
---|
1187 | self.tmax = 60 |
---|
1188 | |
---|
1189 | def checkHostProtocol( self, ip ): |
---|
1190 | |
---|
1191 | """Detect if a ip adress is a multicast address""" |
---|
1192 | |
---|
1193 | MULTICAST_ADDRESS_MIN = ( "224", "0", "0", "0" ) |
---|
1194 | MULTICAST_ADDRESS_MAX = ( "239", "255", "255", "255" ) |
---|
1195 | |
---|
1196 | ip_fields = ip.split( '.' ) |
---|
1197 | |
---|
1198 | if ip_fields >= MULTICAST_ADDRESS_MIN and ip_fields <= MULTICAST_ADDRESS_MAX: |
---|
1199 | |
---|
1200 | return 'multicast' |
---|
1201 | else: |
---|
1202 | return 'udp' |
---|
1203 | |
---|
1204 | def send( self, name, value, dmax, typestr = '', units = '' ): |
---|
1205 | |
---|
1206 | if len( units ) == 0: |
---|
1207 | units = GMETRIC_DEFAULT_UNITS |
---|
1208 | |
---|
1209 | if len( typestr ) == 0: |
---|
1210 | typestr = GMETRIC_DEFAULT_TYPE |
---|
1211 | |
---|
1212 | msg = self.makexdr( name, value, typestr, units, self.slopestr, self.tmax, dmax ) |
---|
1213 | |
---|
1214 | return self.socket.sendto( msg, self.hostport ) |
---|
1215 | |
---|
1216 | def makexdr( self, name, value, typestr, unitstr, slopestr, tmax, dmax ): |
---|
1217 | |
---|
1218 | if slopestr not in self.slope: |
---|
1219 | |
---|
1220 | raise ValueError( "Slope must be one of: " + str( self.slope.keys() ) ) |
---|
1221 | |
---|
1222 | if typestr not in self.type: |
---|
1223 | |
---|
1224 | raise ValueError( "Type must be one of: " + str( self.type ) ) |
---|
1225 | |
---|
1226 | if len( name ) == 0: |
---|
1227 | |
---|
1228 | raise ValueError( "Name must be non-empty" ) |
---|
1229 | |
---|
1230 | self.msg.reset() |
---|
1231 | self.msg.pack_int( 0 ) |
---|
1232 | self.msg.pack_string( typestr ) |
---|
1233 | self.msg.pack_string( name ) |
---|
1234 | self.msg.pack_string( str( value ) ) |
---|
1235 | self.msg.pack_string( unitstr ) |
---|
1236 | self.msg.pack_int( self.slope[ slopestr ] ) |
---|
1237 | self.msg.pack_uint( int( tmax ) ) |
---|
1238 | self.msg.pack_uint( int( dmax ) ) |
---|
1239 | |
---|
1240 | return self.msg.get_buffer() |
---|
1241 | |
---|
1242 | def printTime( ): |
---|
1243 | |
---|
1244 | """Print current time/date in human readable format for log/debug""" |
---|
1245 | |
---|
1246 | return time.strftime("%a, %d %b %Y %H:%M:%S") |
---|
1247 | |
---|
1248 | def debug_msg( level, msg ): |
---|
1249 | |
---|
1250 | """Print msg if at or above current debug level""" |
---|
1251 | |
---|
1252 | global DAEMONIZE, DEBUG_LEVEL, SYSLOG_LEVEL |
---|
1253 | |
---|
1254 | if (not DAEMONIZE and DEBUG_LEVEL >= level): |
---|
1255 | sys.stderr.write( msg + '\n' ) |
---|
1256 | |
---|
1257 | if (DAEMONIZE and USE_SYSLOG and SYSLOG_LEVEL >= level): |
---|
1258 | syslog.syslog( msg ) |
---|
1259 | |
---|
1260 | def write_pidfile(): |
---|
1261 | |
---|
1262 | # Write pidfile if PIDFILE is set |
---|
1263 | # |
---|
1264 | if PIDFILE: |
---|
1265 | |
---|
1266 | pid = os.getpid() |
---|
1267 | |
---|
1268 | pidfile = open( PIDFILE, 'w' ) |
---|
1269 | |
---|
1270 | pidfile.write( str( pid ) ) |
---|
1271 | pidfile.close() |
---|
1272 | |
---|
1273 | def main(): |
---|
1274 | |
---|
1275 | """Application start""" |
---|
1276 | |
---|
1277 | global PBSQuery, PBSError |
---|
1278 | global SYSLOG_FACILITY, USE_SYSLOG, BATCH_API, DAEMONIZE |
---|
1279 | |
---|
1280 | if not processArgs( sys.argv[1:] ): |
---|
1281 | |
---|
1282 | sys.exit( 1 ) |
---|
1283 | |
---|
1284 | # Load appropriate DataGatherer depending on which BATCH_API is set |
---|
1285 | # and any required modules for the Gatherer |
---|
1286 | # |
---|
1287 | if BATCH_API == 'pbs': |
---|
1288 | |
---|
1289 | try: |
---|
1290 | from PBSQuery import PBSQuery, PBSError |
---|
1291 | |
---|
1292 | except ImportError: |
---|
1293 | |
---|
1294 | debug_msg( 0, "FATAL ERROR: BATCH_API set to 'pbs' but python module 'pbs_python' is not installed" ) |
---|
1295 | sys.exit( 1 ) |
---|
1296 | |
---|
1297 | gather = PbsDataGatherer() |
---|
1298 | |
---|
1299 | elif BATCH_API == 'sge': |
---|
1300 | |
---|
1301 | # Tested with SGE 6.0u11. |
---|
1302 | # debug_msg( 0, "FATAL ERROR: BATCH_API 'sge' implementation is currently broken, check future releases" ) |
---|
1303 | |
---|
1304 | # sys.exit( 1 ) |
---|
1305 | |
---|
1306 | gather = SgeDataGatherer() |
---|
1307 | |
---|
1308 | else: |
---|
1309 | debug_msg( 0, "FATAL ERROR: unknown BATCH_API '" + BATCH_API + "' is not supported" ) |
---|
1310 | |
---|
1311 | sys.exit( 1 ) |
---|
1312 | |
---|
1313 | if( DAEMONIZE and USE_SYSLOG ): |
---|
1314 | |
---|
1315 | syslog.openlog( 'jobmond', syslog.LOG_NOWAIT, SYSLOG_FACILITY ) |
---|
1316 | |
---|
1317 | |
---|
1318 | if DAEMONIZE: |
---|
1319 | |
---|
1320 | gather.daemon() |
---|
1321 | else: |
---|
1322 | gather.run() |
---|
1323 | |
---|
1324 | # wh00t? someone started me! :) |
---|
1325 | # |
---|
1326 | if __name__ == '__main__': |
---|
1327 | main() |
---|