[120] | 1 | #!/bin/sh |
---|
| 2 | |
---|
| 3 | # Licensed to the Apache Software Foundation (ASF) under one or more |
---|
| 4 | # contributor license agreements. See the NOTICE file distributed with |
---|
| 5 | # this work for additional information regarding copyright ownership. |
---|
| 6 | # The ASF licenses this file to You under the Apache License, Version 2.0 |
---|
| 7 | # (the "License"); you may not use this file except in compliance with |
---|
| 8 | # the License. You may obtain a copy of the License at |
---|
| 9 | # |
---|
| 10 | # http://www.apache.org/licenses/LICENSE-2.0 |
---|
| 11 | # |
---|
| 12 | # Unless required by applicable law or agreed to in writing, software |
---|
| 13 | # distributed under the License is distributed on an "AS IS" BASIS, |
---|
| 14 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
---|
| 15 | # See the License for the specific language governing permissions and |
---|
| 16 | # limitations under the License. |
---|
| 17 | |
---|
| 18 | |
---|
| 19 | """:" |
---|
| 20 | work_dir=$(dirname $0) |
---|
| 21 | base_name=$(basename $0) |
---|
| 22 | cd $work_dir |
---|
| 23 | |
---|
| 24 | if [ $HOD_PYTHON_HOME ]; then |
---|
| 25 | exec $HOD_PYTHON_HOME -OO $base_name ${1+"$@"} |
---|
| 26 | elif [ -e /usr/bin/python ]; then |
---|
| 27 | exec /usr/bin/python -OO $base_name ${1+"$@"} |
---|
| 28 | elif [ -e /usr/local/bin/python ]; then |
---|
| 29 | exec /usr/local/bin/python -OO $base_name ${1+"$@"} |
---|
| 30 | else |
---|
| 31 | exec python -OO $base_name ${1+"$@"} |
---|
| 32 | fi |
---|
| 33 | ":""" |
---|
| 34 | |
---|
| 35 | """The executable to be used by the user""" |
---|
| 36 | import sys, os, re |
---|
| 37 | |
---|
| 38 | |
---|
| 39 | myName = os.path.basename(sys.argv[0]) |
---|
| 40 | myName = re.sub(".*/", "", myName) |
---|
| 41 | binDirectory = os.path.realpath(sys.argv[0]) |
---|
| 42 | rootDirectory = re.sub("/bin/.*", "", binDirectory) |
---|
| 43 | libDirectory = rootDirectory |
---|
| 44 | |
---|
| 45 | sys.path.append(libDirectory) |
---|
| 46 | |
---|
| 47 | from hodlib.HodRing.hodRing import HodRing |
---|
| 48 | from hodlib.Common.setup import * |
---|
| 49 | from hodlib.Common.util import filter_warnings, get_exception_string, \ |
---|
| 50 | get_exception_error_string, getMapredSystemDirectory, \ |
---|
| 51 | to_http_url, local_fqdn |
---|
| 52 | from hodlib.Common.logger import getLogger, ensureLogDir |
---|
| 53 | from hodlib.Common.xmlrpc import hodXRClient |
---|
| 54 | |
---|
| 55 | filter_warnings() |
---|
| 56 | |
---|
| 57 | reVersion = re.compile(".*(\d+_\d+).*") |
---|
| 58 | |
---|
| 59 | VERSION = '$HeadURL$' |
---|
| 60 | |
---|
| 61 | reMatch = reVersion.match(VERSION) |
---|
| 62 | if reMatch: |
---|
| 63 | VERSION = reMatch.group(1) |
---|
| 64 | VERSION = re.sub("_", ".", VERSION) |
---|
| 65 | else: |
---|
| 66 | VERSION = 'DEV' |
---|
| 67 | |
---|
| 68 | # Definition tuple is of the form: |
---|
| 69 | # (name, type, description, default value, required?, validate?) |
---|
| 70 | # |
---|
| 71 | defList = { 'hodring' : ( |
---|
| 72 | ('temp-dir', 'directory', 'hod work directories', |
---|
| 73 | False, None, True, False), |
---|
| 74 | |
---|
| 75 | ('log-dir', 'directory', 'hod logging directory.', |
---|
| 76 | False, os.path.join(rootDirectory, 'logs'), False, True), |
---|
| 77 | |
---|
| 78 | ('log-destination-uri', 'string', |
---|
| 79 | 'URI to store logs to, local://some_path or ' |
---|
| 80 | + 'hdfs://host:port/some_path', |
---|
| 81 | False, None, False, True), |
---|
| 82 | |
---|
| 83 | ('pkgs', 'directory', 'Path to Hadoop to use in case of uploading to HDFS', |
---|
| 84 | False, None, False, True), |
---|
| 85 | |
---|
| 86 | ('syslog-address', 'address', 'Syslog address.', |
---|
| 87 | False, None, False, True), |
---|
| 88 | |
---|
| 89 | ('java-home', 'directory', 'Java home directory.', |
---|
| 90 | False, None, True, True), |
---|
| 91 | |
---|
| 92 | ('debug', 'pos_int', 'Debugging level, 0-4.', |
---|
| 93 | False, 3, True, True), |
---|
| 94 | |
---|
| 95 | ('register', 'bool', 'Register with service registry?', |
---|
| 96 | False, True, True, True), |
---|
| 97 | |
---|
| 98 | ('stream', 'bool', 'Output to stderr.', |
---|
| 99 | False, False, False, True), |
---|
| 100 | |
---|
| 101 | ('userid', 'user_account', |
---|
| 102 | 'User ID the hod shell is running under.', |
---|
| 103 | False, None, True, False), |
---|
| 104 | |
---|
| 105 | ('xrs-port-range', 'range', 'XML-RPC port range n-m.', |
---|
| 106 | False, None, True, True), |
---|
| 107 | |
---|
| 108 | ('http-port-range', 'range', 'HTTP port range n-m.', |
---|
| 109 | False, None, True, True), |
---|
| 110 | |
---|
| 111 | ('command', 'string', 'Command for hodring to run.', |
---|
| 112 | False, None, False, True), |
---|
| 113 | |
---|
| 114 | ('service-id', 'string', 'Service ID.', |
---|
| 115 | False, None, False, True), |
---|
| 116 | |
---|
| 117 | ('download-addr', 'string', 'Download HTTP address.', |
---|
| 118 | False, None, False, True), |
---|
| 119 | |
---|
| 120 | ('svcrgy-addr', 'address', 'Service registry XMLRPC address.', |
---|
| 121 | False, None, True, True), |
---|
| 122 | |
---|
| 123 | ('ringmaster-xrs-addr', 'address', 'Ringmaster XML-RPC address.', |
---|
| 124 | False, None, False, True), |
---|
| 125 | |
---|
| 126 | ('tarball-retry-initial-time', 'pos_float','initial retry time for tarball download', |
---|
| 127 | False, 1, False, True), |
---|
| 128 | |
---|
| 129 | ('tarball-retry-interval', 'pos_float','interval to spread retries for tarball download', |
---|
| 130 | False, 3, False, True), |
---|
| 131 | |
---|
| 132 | ('cmd-retry-initial-time', 'pos_float','initial retry time for getting commands', |
---|
| 133 | False, 2, False, True), |
---|
| 134 | |
---|
| 135 | ('cmd-retry-interval', 'pos_float','interval to spread retries for getting commands', |
---|
| 136 | False, 2, False, True), |
---|
| 137 | |
---|
| 138 | ('mapred-system-dir-root', 'string', 'Root under which mapreduce system directory names are generated by HOD.', |
---|
| 139 | False, '/mapredsystem', False, False)) |
---|
| 140 | } |
---|
| 141 | |
---|
| 142 | if __name__ == '__main__': |
---|
| 143 | |
---|
| 144 | confDef = definition() |
---|
| 145 | confDef.add_defs(defList) |
---|
| 146 | hodRingOptions = options(confDef, "./%s [OPTIONS]" % myName, VERSION) |
---|
| 147 | ensureLogDir(hodRingOptions['hodring']['log-dir']) |
---|
| 148 | service = None |
---|
| 149 | try: |
---|
| 150 | (status, statusMsgs) = hodRingOptions.verify() |
---|
| 151 | if not status: |
---|
| 152 | raise Exception("%s" % statusMsgs) |
---|
| 153 | hodRingOptions['hodring']['base-dir'] = rootDirectory |
---|
| 154 | service = HodRing(hodRingOptions) |
---|
| 155 | service.start() |
---|
| 156 | service.wait() |
---|
| 157 | |
---|
| 158 | if service.log: |
---|
| 159 | log = service.log |
---|
| 160 | else: |
---|
| 161 | log = getLogger(hodRingOptions['hodring'],'hodring') |
---|
| 162 | |
---|
| 163 | list = [] |
---|
| 164 | |
---|
| 165 | runningHadoops = service.getRunningValues() |
---|
| 166 | |
---|
| 167 | mrSysDirManager = None |
---|
| 168 | for cmd in runningHadoops: |
---|
| 169 | if cmd.name == 'jobtracker': |
---|
| 170 | mrSysDirManager = cmd.getMRSystemDirectoryManager() |
---|
| 171 | log.debug("addding %s to cleanup list..." % cmd) |
---|
| 172 | cmd.addCleanup(list) |
---|
| 173 | |
---|
| 174 | list.append(service.getTempDir()) |
---|
| 175 | log.debug(list) |
---|
| 176 | |
---|
| 177 | # archive_logs now |
---|
| 178 | cmdString = os.path.join(rootDirectory, "bin", "hodcleanup") # same python |
---|
| 179 | |
---|
| 180 | if (len(runningHadoops) == 0): |
---|
| 181 | log.info("len(runningHadoops) == 0, No running cluster?") |
---|
| 182 | log.info("Skipping __copy_archive_to_dfs") |
---|
| 183 | hadoopString = "" |
---|
| 184 | else: hadoopString=runningHadoops[0].path |
---|
| 185 | |
---|
| 186 | #construct the arguments |
---|
| 187 | if hodRingOptions['hodring'].has_key('log-destination-uri'): |
---|
| 188 | cmdString = cmdString + " --log-destination-uri " \ |
---|
| 189 | + hodRingOptions['hodring']['log-destination-uri'] |
---|
| 190 | |
---|
| 191 | hadoopLogDirs = service.getHadoopLogDirs() |
---|
| 192 | if hadoopLogDirs: |
---|
| 193 | cmdString = cmdString \ |
---|
| 194 | + " --hadoop-log-dirs " \ |
---|
| 195 | + ",".join(hadoopLogDirs) |
---|
| 196 | |
---|
| 197 | cmdString = cmdString \ |
---|
| 198 | + " --temp-dir " \ |
---|
| 199 | + service._cfg['temp-dir'] \ |
---|
| 200 | + " --hadoop-command-string " \ |
---|
| 201 | + hadoopString \ |
---|
| 202 | + " --user-id " \ |
---|
| 203 | + service._cfg['userid'] \ |
---|
| 204 | + " --service-id " \ |
---|
| 205 | + service._cfg['service-id'] \ |
---|
| 206 | + " --hodring-debug " \ |
---|
| 207 | + str(hodRingOptions['hodring']['debug']) \ |
---|
| 208 | + " --hodring-log-dir " \ |
---|
| 209 | + hodRingOptions['hodring']['log-dir'] \ |
---|
| 210 | + " --hodring-cleanup-list " \ |
---|
| 211 | + ",".join(list) |
---|
| 212 | |
---|
| 213 | if hodRingOptions['hodring'].has_key('syslog-address'): |
---|
| 214 | syslogAddr = hodRingOptions['hodring']['syslog-address'][0] + \ |
---|
| 215 | ':' + str(hodRingOptions['hodring']['syslog-address'][1]) |
---|
| 216 | cmdString = cmdString + " --hodring-syslog-address " + syslogAddr |
---|
| 217 | if service._cfg.has_key('pkgs'): |
---|
| 218 | cmdString = cmdString + " --pkgs " + service._cfg['pkgs'] |
---|
| 219 | |
---|
| 220 | if mrSysDirManager is not None: |
---|
| 221 | cmdString = "%s %s" % (cmdString, mrSysDirManager.toCleanupArgs()) |
---|
| 222 | |
---|
| 223 | log.info("cleanup commandstring : ") |
---|
| 224 | log.info(cmdString) |
---|
| 225 | |
---|
| 226 | # clean up |
---|
| 227 | cmd = ['/bin/sh', '-c', cmdString] |
---|
| 228 | |
---|
| 229 | mswindows = (sys.platform == "win32") |
---|
| 230 | originalcwd = os.getcwd() |
---|
| 231 | |
---|
| 232 | if not mswindows: |
---|
| 233 | try: |
---|
| 234 | pid = os.fork() |
---|
| 235 | if pid > 0: |
---|
| 236 | # exit first parent |
---|
| 237 | log.info("child(pid: %s) is now doing cleanup" % pid) |
---|
| 238 | sys.exit(0) |
---|
| 239 | except OSError, e: |
---|
| 240 | log.error("fork failed: %d (%s)" % (e.errno, e.strerror)) |
---|
| 241 | sys.exit(1) |
---|
| 242 | |
---|
| 243 | # decouple from parent environment |
---|
| 244 | os.chdir("/") |
---|
| 245 | os.setsid() |
---|
| 246 | os.umask(0) |
---|
| 247 | |
---|
| 248 | MAXFD = 128 # more than enough file descriptors to close. Just in case. |
---|
| 249 | for i in xrange(0, MAXFD): |
---|
| 250 | try: |
---|
| 251 | os.close(i) |
---|
| 252 | except OSError: |
---|
| 253 | pass |
---|
| 254 | |
---|
| 255 | try: |
---|
| 256 | os.execvp(cmd[0], cmd) |
---|
| 257 | finally: |
---|
| 258 | log.critical("exec failed") |
---|
| 259 | os._exit(1) |
---|
| 260 | |
---|
| 261 | except Exception, e: |
---|
| 262 | if service: |
---|
| 263 | if service.log: |
---|
| 264 | log = service.log |
---|
| 265 | else: |
---|
| 266 | log = getLogger(hodRingOptions['hodring'], 'hodring') |
---|
| 267 | log.error("Error in bin/hodring %s. \nStack trace:\n%s" %(get_exception_error_string(),get_exception_string())) |
---|
| 268 | |
---|
| 269 | log.info("now trying informing to ringmaster") |
---|
| 270 | log.info(hodRingOptions['hodring']['ringmaster-xrs-addr']) |
---|
| 271 | log.info(hodRingOptions.normalizeValue('hodring', 'ringmaster-xrs-addr')) |
---|
| 272 | log.info(to_http_url(hodRingOptions.normalizeValue( \ |
---|
| 273 | 'hodring', 'ringmaster-xrs-addr'))) |
---|
| 274 | # Report errors to the Ringmaster if possible |
---|
| 275 | try: |
---|
| 276 | ringXRAddress = to_http_url(hodRingOptions.normalizeValue( \ |
---|
| 277 | 'hodring', 'ringmaster-xrs-addr')) |
---|
| 278 | log.debug("Creating ringmaster XML-RPC client.") |
---|
| 279 | ringClient = hodXRClient(ringXRAddress) |
---|
| 280 | if ringClient is not None: |
---|
| 281 | addr = local_fqdn() + "_" + str(os.getpid()) |
---|
| 282 | ringClient.setHodRingErrors(addr, str(e)) |
---|
| 283 | log.info("Reported errors to ringmaster at %s" % ringXRAddress) |
---|
| 284 | except Exception, e: |
---|
| 285 | log.error("Failed to report errors to ringmaster at %s" % ringXRAddress) |
---|
| 286 | log.error("Reason : %s" % get_exception_string()) |
---|
| 287 | # End of reporting errors to the client |
---|