| 1 | #!/usr/bin/env python |
|---|
| 2 | # -*- coding: ISO-8859-1 -*- |
|---|
| 3 | |
|---|
| 4 | ################################## |
|---|
| 5 | # @program smon |
|---|
| 6 | # @description simulation monitor |
|---|
| 7 | # @copyright Copyright â(c)2009 Centre National de la Recherche Scientifique CNRS. |
|---|
| 8 | # All Rights Reservedâ |
|---|
| 9 | # @svn_file $Id: watcher 2545 2013-02-01 09:58:10Z jripsl $ |
|---|
| 10 | # @version $Rev: 2545 $ |
|---|
| 11 | # @lastrevision $Date: 2013-02-01 10:58:10 +0100 (Fri, 01 Feb 2013) $ |
|---|
| 12 | # @license CeCILL (http://dods.ipsl.jussieu.fr/jripsl/smon/LICENSE) |
|---|
| 13 | ################################## |
|---|
| 14 | |
|---|
| 15 | import pika |
|---|
| 16 | import base64 |
|---|
| 17 | import json |
|---|
| 18 | import sys |
|---|
| 19 | import signal |
|---|
| 20 | import traceback |
|---|
| 21 | import smtplib |
|---|
| 22 | from email.mime.text import MIMEText |
|---|
| 23 | import datetime |
|---|
| 24 | |
|---|
| 25 | # line below is to include "smon" package in the search path |
|---|
| 26 | sys.path.append("/opt/supervisor/Monitoring") |
|---|
| 27 | |
|---|
| 28 | import smon.repo_io as repo_io |
|---|
| 29 | import smon.types |
|---|
| 30 | |
|---|
| 31 | CSTE_BROKER_HOST='cstest-broker.ipsl.jussieu.fr' # cstest |
|---|
| 32 | #CSTE_BROKER_HOST='localhost' # vesg4 |
|---|
| 33 | |
|---|
| 34 | class Mail(): |
|---|
| 35 | |
|---|
| 36 | @classmethod |
|---|
| 37 | def mail_example(cls): |
|---|
| 38 | me="jripsl@ipsl.jussieu.fr" |
|---|
| 39 | you="jripsl@ipsl.jussieu.fr" |
|---|
| 40 | body="Alarm" |
|---|
| 41 | object="Supervisor" |
|---|
| 42 | |
|---|
| 43 | cls.mail(me,you,object,body) |
|---|
| 44 | |
|---|
| 45 | @classmethod |
|---|
| 46 | def send_mail(cls,me,you,object,body): |
|---|
| 47 | msg = MIMEText(body) |
|---|
| 48 | msg['Subject'] = object |
|---|
| 49 | msg['From'] = me |
|---|
| 50 | msg['To'] = you |
|---|
| 51 | |
|---|
| 52 | # Send the message via our own SMTP server, but don't include the # envelope header. |
|---|
| 53 | s = smtplib.SMTP('localhost') |
|---|
| 54 | s.sendmail(me,[you], msg.as_string()) |
|---|
| 55 | s.quit() |
|---|
| 56 | |
|---|
| 57 | class Actions(): |
|---|
| 58 | |
|---|
| 59 | @classmethod |
|---|
| 60 | def store_msg(cls,message): |
|---|
| 61 | |
|---|
| 62 | try: |
|---|
| 63 | |
|---|
| 64 | # the simu exists when we are here (see TAG0001 tag) |
|---|
| 65 | s=repo_io.retrieve_simulation(message.simuid) |
|---|
| 66 | |
|---|
| 67 | if s is None: |
|---|
| 68 | raise Exception("WATCH-ERR102","simulation not found") |
|---|
| 69 | |
|---|
| 70 | repo_io.create_message(message,s) |
|---|
| 71 | |
|---|
| 72 | except: |
|---|
| 73 | traceback.print_exc() |
|---|
| 74 | raise |
|---|
| 75 | |
|---|
| 76 | @classmethod |
|---|
| 77 | def cleanup(cls,message): |
|---|
| 78 | repo_io.cleanup() # truncate/delete everything |
|---|
| 79 | |
|---|
| 80 | @classmethod |
|---|
| 81 | def set_sim_status_to_error(cls,message): |
|---|
| 82 | |
|---|
| 83 | try: |
|---|
| 84 | |
|---|
| 85 | s=repo_io.retrieve_simulation(message.simuid) |
|---|
| 86 | |
|---|
| 87 | s.status="error" |
|---|
| 88 | |
|---|
| 89 | repo_io.update_simulation_status(s) |
|---|
| 90 | |
|---|
| 91 | except: |
|---|
| 92 | traceback.print_exc() |
|---|
| 93 | raise |
|---|
| 94 | |
|---|
| 95 | @classmethod |
|---|
| 96 | def set_sim_status_to_complete(cls,message): |
|---|
| 97 | |
|---|
| 98 | s=repo_io.retrieve_simulation(message.simuid) |
|---|
| 99 | |
|---|
| 100 | s.status="complete" |
|---|
| 101 | |
|---|
| 102 | repo_io.update_simulation_status(s) |
|---|
| 103 | |
|---|
| 104 | @classmethod |
|---|
| 105 | def crea_sim(cls,message): |
|---|
| 106 | |
|---|
| 107 | s=repo_io.retrieve_simulation(message.simuid) |
|---|
| 108 | |
|---|
| 109 | if s is not None: |
|---|
| 110 | #repo_io.delete_simulation(name) |
|---|
| 111 | |
|---|
| 112 | s.status="running" |
|---|
| 113 | repo_io.update_simulation_status(s) |
|---|
| 114 | |
|---|
| 115 | else: |
|---|
| 116 | simulation=smon.types.Simulation(name=message.simuid,status="running") |
|---|
| 117 | |
|---|
| 118 | |
|---|
| 119 | repo_io.create_simulation(simulation) |
|---|
| 120 | |
|---|
| 121 | @classmethod |
|---|
| 122 | def mail(cls): |
|---|
| 123 | cls.mail_example() |
|---|
| 124 | |
|---|
| 125 | @classmethod |
|---|
| 126 | def print_stdout(cls,message): |
|---|
| 127 | # used for debug |
|---|
| 128 | |
|---|
| 129 | """ |
|---|
| 130 | if message.file is not None: |
|---|
| 131 | print "%s %s %s %s %s\n"%(message.code,message.jobid,message.command,message.timestamp,message.file) |
|---|
| 132 | else: |
|---|
| 133 | print "%s %s %s %s\n"%(message.code,message.jobid,message.command,message.timestamp) |
|---|
| 134 | """ |
|---|
| 135 | |
|---|
| 136 | print "%s %s %s %s\n"%(message.code,message.jobid,message.command,message.timestamp) |
|---|
| 137 | #pass |
|---|
| 138 | |
|---|
| 139 | @classmethod |
|---|
| 140 | def log(cls,message): |
|---|
| 141 | with open("/opt/supervisor/log/supervisor.log", "a") as log_file: |
|---|
| 142 | log_file.write("%s %s %s %s %s\n"%(datetime.datetime.now().strftime('%Y%m%d_%H%M%S'), message.code,message.jobid,message.timestamp,message.command)) |
|---|
| 143 | |
|---|
| 144 | @classmethod |
|---|
| 145 | def execActions(cls,message): |
|---|
| 146 | |
|---|
| 147 | message_code=message.code |
|---|
| 148 | |
|---|
| 149 | for action in MessageActionsMapping.mapping[message_code]: |
|---|
| 150 | proc_name=action |
|---|
| 151 | |
|---|
| 152 | try: |
|---|
| 153 | getattr(Actions, proc_name)(message) |
|---|
| 154 | except Exception,e: |
|---|
| 155 | traceback.print_exc() |
|---|
| 156 | |
|---|
| 157 | raise Exception("WATCH-ERR002","procedure error (%s,%s)"%(proc_name,str(e))) |
|---|
| 158 | |
|---|
| 159 | class MessageActionsMapping(): |
|---|
| 160 | |
|---|
| 161 | # debug |
|---|
| 162 | # |
|---|
| 163 | # TAG0001: note that crea_sim must be BEFORE store_msg in the list (because when we insert the msg, we need the simu_id) |
|---|
| 164 | # |
|---|
| 165 | mapping = { "0000":["crea_sim", "log", "store_msg", "print_stdout"], |
|---|
| 166 | "0100":["log", "store_msg", "print_stdout", "set_sim_status_to_complete"], |
|---|
| 167 | "1000":["log", "store_msg", "print_stdout"], |
|---|
| 168 | "1100":["log", "store_msg", "print_stdout"], |
|---|
| 169 | "2000":["log", "store_msg", "print_stdout"], |
|---|
| 170 | "3000":["log", "store_msg", "print_stdout"], |
|---|
| 171 | "8888":["cleanup"], |
|---|
| 172 | "9000":["log", "store_msg", "print_stdout"], |
|---|
| 173 | "9999":["log", "store_msg", "print_stdout", "set_sim_status_to_error"] } |
|---|
| 174 | |
|---|
| 175 | # prod |
|---|
| 176 | # |
|---|
| 177 | # TAG0001: note that crea_sim must be BEFORE store_msg in the list (because when we insert the msg, we need the simu_id) |
|---|
| 178 | # |
|---|
| 179 | """ |
|---|
| 180 | mapping = { "0000":["crea_sim", "log", "store_msg"], |
|---|
| 181 | "0100":["log", "store_msg", "set_sim_status_to_complete"], |
|---|
| 182 | "1000":["log", "store_msg"], |
|---|
| 183 | "1100":["log", "store_msg"], |
|---|
| 184 | "2000":["log", "store_msg"], |
|---|
| 185 | "3000":["log", "store_msg"], |
|---|
| 186 | "8888":["cleanup"], |
|---|
| 187 | "9000":["log", "store_msg", "mail"], |
|---|
| 188 | "9999":["log", "store_msg", "set_sim_status_to_error", "mail"] } |
|---|
| 189 | """ |
|---|
| 190 | |
|---|
| 191 | class Watcher(): |
|---|
| 192 | |
|---|
| 193 | @classmethod |
|---|
| 194 | def start(cls): |
|---|
| 195 | repo_io.init() # open DB connection |
|---|
| 196 | |
|---|
| 197 | @classmethod |
|---|
| 198 | def stop(cls): |
|---|
| 199 | repo_io.free() # close DB connection |
|---|
| 200 | |
|---|
| 201 | @classmethod |
|---|
| 202 | def main(self): |
|---|
| 203 | |
|---|
| 204 | """ |
|---|
| 205 | # parse args |
|---|
| 206 | parser = argparse.ArgumentParser(prog='watcher') |
|---|
| 207 | parser.add_argument('-v', dest='verbose',required=False,action='store_true') |
|---|
| 208 | args = parser.parse_args() |
|---|
| 209 | |
|---|
| 210 | # check |
|---|
| 211 | if not os.path.exists(SMON.smon_home): |
|---|
| 212 | sys.exit(1) |
|---|
| 213 | |
|---|
| 214 | SMON.init_singleton() |
|---|
| 215 | """ |
|---|
| 216 | |
|---|
| 217 | connection = pika.BlockingConnection(pika.ConnectionParameters(host=CSTE_BROKER_HOST)) |
|---|
| 218 | self.channel = connection.channel() |
|---|
| 219 | |
|---|
| 220 | |
|---|
| 221 | print ' [*] Waiting for messages. To exit press CTRL+C' |
|---|
| 222 | |
|---|
| 223 | def callback(ch, method, properties, raw_msg): |
|---|
| 224 | |
|---|
| 225 | # msg fmt: body:base64,file:base64 (no JSON here !!!) |
|---|
| 226 | |
|---|
| 227 | |
|---|
| 228 | |
|---|
| 229 | |
|---|
| 230 | # first deserialization (no JSON here !!!) |
|---|
| 231 | |
|---|
| 232 | fields=raw_msg.split(",") |
|---|
| 233 | |
|---|
| 234 | l__tmp_dic={} |
|---|
| 235 | |
|---|
| 236 | for field in fields: |
|---|
| 237 | |
|---|
| 238 | # debug |
|---|
| 239 | #print " [x] Received %s" % field |
|---|
| 240 | |
|---|
| 241 | splitted_field=field.split(":") |
|---|
| 242 | |
|---|
| 243 | key=splitted_field[0] |
|---|
| 244 | val=splitted_field[1] |
|---|
| 245 | |
|---|
| 246 | l__tmp_dic[key]=val |
|---|
| 247 | |
|---|
| 248 | |
|---|
| 249 | # debug |
|---|
| 250 | #print " [x] Received %s (encoded)" % l__tmp_dic["body"] |
|---|
| 251 | |
|---|
| 252 | |
|---|
| 253 | # base64 decode body |
|---|
| 254 | base64_decoded_msg=base64.b64decode(l__tmp_dic["body"]) |
|---|
| 255 | |
|---|
| 256 | |
|---|
| 257 | # debug |
|---|
| 258 | #print " [x] Received %s" % raw_msg |
|---|
| 259 | #print " [x] Received %s (uudecoded)" % base64_decoded_msg |
|---|
| 260 | #print " [x] Received %s (uudecoded)" % base64_decoded_msg |
|---|
| 261 | |
|---|
| 262 | |
|---|
| 263 | # message deserialization |
|---|
| 264 | message=None |
|---|
| 265 | try: |
|---|
| 266 | JSON_msg=json.loads(base64_decoded_msg) |
|---|
| 267 | message=smon.types.Message(JSON_msg) # all JSON object members will be available in smon.types.Message object |
|---|
| 268 | |
|---|
| 269 | # non working |
|---|
| 270 | #print message.type |
|---|
| 271 | |
|---|
| 272 | # working |
|---|
| 273 | #print message.code |
|---|
| 274 | |
|---|
| 275 | |
|---|
| 276 | |
|---|
| 277 | except Exception,e: |
|---|
| 278 | print "ERR009 - exception occurs (exception=%s,msg=%s)"%(str(e),base64_decoded_msg) |
|---|
| 279 | |
|---|
| 280 | traceback.print_exc() |
|---|
| 281 | raise |
|---|
| 282 | |
|---|
| 283 | |
|---|
| 284 | |
|---|
| 285 | # manage config-card file which is attached to the "0000" code message (this file is base64 encoded and need to be unencoded) |
|---|
| 286 | # |
|---|
| 287 | if "file" in l__tmp_dic: |
|---|
| 288 | |
|---|
| 289 | # base64 decode file |
|---|
| 290 | base64_decoded_file=base64.b64decode(l__tmp_dic["file"]) |
|---|
| 291 | |
|---|
| 292 | # add as msg attribute |
|---|
| 293 | message.file=base64_decoded_file |
|---|
| 294 | |
|---|
| 295 | |
|---|
| 296 | |
|---|
| 297 | # execute actions |
|---|
| 298 | try: |
|---|
| 299 | # message code based action |
|---|
| 300 | Actions.execActions(message) |
|---|
| 301 | |
|---|
| 302 | except Exception,e: |
|---|
| 303 | print "ERR019 - exception occurs (exception=%s)"%(str(e)) |
|---|
| 304 | #print "ERR019 - exception occurs (exception=%s,msg=%s)"%(str(e),base64_decoded_msg) |
|---|
| 305 | |
|---|
| 306 | traceback.print_exc() |
|---|
| 307 | |
|---|
| 308 | raise |
|---|
| 309 | |
|---|
| 310 | |
|---|
| 311 | # slow down consumer |
|---|
| 312 | #time.sleep(0.5) |
|---|
| 313 | |
|---|
| 314 | self.channel.queue_declare(queue='myqueue') |
|---|
| 315 | |
|---|
| 316 | self.channel.basic_consume(callback, queue='myqueue', no_ack=True) |
|---|
| 317 | |
|---|
| 318 | self.channel.start_consuming() |
|---|
| 319 | |
|---|
| 320 | |
|---|
| 321 | """ |
|---|
| 322 | SMON.free_singleton() |
|---|
| 323 | """ |
|---|
| 324 | |
|---|
| 325 | def signal_handler(signal, frame): |
|---|
| 326 | print 'You pressed Ctrl+C!' |
|---|
| 327 | Watcher.channel.stop_consuming() |
|---|
| 328 | Watcher.stop() |
|---|
| 329 | sys.exit(0) |
|---|
| 330 | |
|---|
| 331 | if __name__ == '__main__': |
|---|
| 332 | |
|---|
| 333 | signal.signal(signal.SIGINT, signal_handler) |
|---|
| 334 | |
|---|
| 335 | try: |
|---|
| 336 | |
|---|
| 337 | Watcher.start() |
|---|
| 338 | |
|---|
| 339 | Watcher.main() |
|---|
| 340 | |
|---|
| 341 | Watcher.stop() |
|---|
| 342 | |
|---|
| 343 | sys.exit(0) |
|---|
| 344 | |
|---|
| 345 | except Exception, e: |
|---|
| 346 | |
|---|
| 347 | traceback.print_exc() |
|---|
| 348 | |
|---|
| 349 | sys.exit(1) |
|---|
