Package ganeti :: Module bootstrap
[hide private]
[frames] | no frames]

Source Code for Module ganeti.bootstrap

  1  # 
  2  # 
  3   
  4  # Copyright (C) 2006, 2007, 2008, 2010, 2011, 2012 Google Inc. 
  5  # 
  6  # This program is free software; you can redistribute it and/or modify 
  7  # it under the terms of the GNU General Public License as published by 
  8  # the Free Software Foundation; either version 2 of the License, or 
  9  # (at your option) any later version. 
 10  # 
 11  # This program is distributed in the hope that it will be useful, but 
 12  # WITHOUT ANY WARRANTY; without even the implied warranty of 
 13  # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU 
 14  # General Public License for more details. 
 15  # 
 16  # You should have received a copy of the GNU General Public License 
 17  # along with this program; if not, write to the Free Software 
 18  # Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 
 19  # 02110-1301, USA. 
 20   
 21   
 22  """Functions to bootstrap a new cluster. 
 23   
 24  """ 
 25   
 26  import os 
 27  import os.path 
 28  import re 
 29  import logging 
 30  import time 
 31  import tempfile 
 32   
 33  from ganeti import rpc 
 34  from ganeti import ssh 
 35  from ganeti import utils 
 36  from ganeti import errors 
 37  from ganeti import config 
 38  from ganeti import constants 
 39  from ganeti import objects 
 40  from ganeti import ssconf 
 41  from ganeti import serializer 
 42  from ganeti import hypervisor 
 43  from ganeti import bdev 
 44  from ganeti import netutils 
 45  from ganeti import luxi 
 46  from ganeti import jstore 
 47  from ganeti import pathutils 
 48   
 49   
 50  # ec_id for InitConfig's temporary reservation manager 
 51  _INITCONF_ECID = "initconfig-ecid" 
 52   
 53  #: After how many seconds daemon must be responsive 
 54  _DAEMON_READY_TIMEOUT = 10.0 
 55   
 56   
57 -def _InitSSHSetup():
58 """Setup the SSH configuration for the cluster. 59 60 This generates a dsa keypair for root, adds the pub key to the 61 permitted hosts and adds the hostkey to its own known hosts. 62 63 """ 64 priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER) 65 66 for name in priv_key, pub_key: 67 if os.path.exists(name): 68 utils.CreateBackup(name) 69 utils.RemoveFile(name) 70 71 result = utils.RunCmd(["ssh-keygen", "-t", "dsa", 72 "-f", priv_key, 73 "-q", "-N", ""]) 74 if result.failed: 75 raise errors.OpExecError("Could not generate ssh keypair, error %s" % 76 result.output) 77 78 utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
79 80
81 -def GenerateHmacKey(file_name):
82 """Writes a new HMAC key. 83 84 @type file_name: str 85 @param file_name: Path to output file 86 87 """ 88 utils.WriteFile(file_name, data="%s\n" % utils.GenerateSecret(), mode=0400, 89 backup=True)
90 91
92 -def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_spice_cert, 93 new_confd_hmac_key, new_cds, 94 rapi_cert_pem=None, spice_cert_pem=None, 95 spice_cacert_pem=None, cds=None, 96 nodecert_file=pathutils.NODED_CERT_FILE, 97 rapicert_file=pathutils.RAPI_CERT_FILE, 98 spicecert_file=pathutils.SPICE_CERT_FILE, 99 spicecacert_file=pathutils.SPICE_CACERT_FILE, 100 hmackey_file=pathutils.CONFD_HMAC_KEY, 101 cds_file=pathutils.CLUSTER_DOMAIN_SECRET_FILE):
102 """Updates the cluster certificates, keys and secrets. 103 104 @type new_cluster_cert: bool 105 @param new_cluster_cert: Whether to generate a new cluster certificate 106 @type new_rapi_cert: bool 107 @param new_rapi_cert: Whether to generate a new RAPI certificate 108 @type new_spice_cert: bool 109 @param new_spice_cert: Whether to generate a new SPICE certificate 110 @type new_confd_hmac_key: bool 111 @param new_confd_hmac_key: Whether to generate a new HMAC key 112 @type new_cds: bool 113 @param new_cds: Whether to generate a new cluster domain secret 114 @type rapi_cert_pem: string 115 @param rapi_cert_pem: New RAPI certificate in PEM format 116 @type spice_cert_pem: string 117 @param spice_cert_pem: New SPICE certificate in PEM format 118 @type spice_cacert_pem: string 119 @param spice_cacert_pem: Certificate of the CA that signed the SPICE 120 certificate, in PEM format 121 @type cds: string 122 @param cds: New cluster domain secret 123 @type nodecert_file: string 124 @param nodecert_file: optional override of the node cert file path 125 @type rapicert_file: string 126 @param rapicert_file: optional override of the rapi cert file path 127 @type spicecert_file: string 128 @param spicecert_file: optional override of the spice cert file path 129 @type spicecacert_file: string 130 @param spicecacert_file: optional override of the spice CA cert file path 131 @type hmackey_file: string 132 @param hmackey_file: optional override of the hmac key file path 133 134 """ 135 # noded SSL certificate 136 cluster_cert_exists = os.path.exists(nodecert_file) 137 if new_cluster_cert or not cluster_cert_exists: 138 if cluster_cert_exists: 139 utils.CreateBackup(nodecert_file) 140 141 logging.debug("Generating new cluster certificate at %s", nodecert_file) 142 utils.GenerateSelfSignedSslCert(nodecert_file) 143 144 # confd HMAC key 145 if new_confd_hmac_key or not os.path.exists(hmackey_file): 146 logging.debug("Writing new confd HMAC key to %s", hmackey_file) 147 GenerateHmacKey(hmackey_file) 148 149 # RAPI 150 rapi_cert_exists = os.path.exists(rapicert_file) 151 152 if rapi_cert_pem: 153 # Assume rapi_pem contains a valid PEM-formatted certificate and key 154 logging.debug("Writing RAPI certificate at %s", rapicert_file) 155 utils.WriteFile(rapicert_file, data=rapi_cert_pem, backup=True) 156 157 elif new_rapi_cert or not rapi_cert_exists: 158 if rapi_cert_exists: 159 utils.CreateBackup(rapicert_file) 160 161 logging.debug("Generating new RAPI certificate at %s", rapicert_file) 162 utils.GenerateSelfSignedSslCert(rapicert_file) 163 164 # SPICE 165 spice_cert_exists = os.path.exists(spicecert_file) 166 spice_cacert_exists = os.path.exists(spicecacert_file) 167 if spice_cert_pem: 168 # spice_cert_pem implies also spice_cacert_pem 169 logging.debug("Writing SPICE certificate at %s", spicecert_file) 170 utils.WriteFile(spicecert_file, data=spice_cert_pem, backup=True) 171 logging.debug("Writing SPICE CA certificate at %s", spicecacert_file) 172 utils.WriteFile(spicecacert_file, data=spice_cacert_pem, backup=True) 173 elif new_spice_cert or not spice_cert_exists: 174 if spice_cert_exists: 175 utils.CreateBackup(spicecert_file) 176 if spice_cacert_exists: 177 utils.CreateBackup(spicecacert_file) 178 179 logging.debug("Generating new self-signed SPICE certificate at %s", 180 spicecert_file) 181 (_, cert_pem) = utils.GenerateSelfSignedSslCert(spicecert_file) 182 183 # Self-signed certificate -> the public certificate is also the CA public 184 # certificate 185 logging.debug("Writing the public certificate to %s", 186 spicecert_file) 187 utils.io.WriteFile(spicecacert_file, mode=0400, data=cert_pem) 188 189 # Cluster domain secret 190 if cds: 191 logging.debug("Writing cluster domain secret to %s", cds_file) 192 utils.WriteFile(cds_file, data=cds, backup=True) 193 194 elif new_cds or not os.path.exists(cds_file): 195 logging.debug("Generating new cluster domain secret at %s", cds_file) 196 GenerateHmacKey(cds_file)
197 198
199 -def _InitGanetiServerSetup(master_name):
200 """Setup the necessary configuration for the initial node daemon. 201 202 This creates the nodepass file containing the shared password for 203 the cluster, generates the SSL certificate and starts the node daemon. 204 205 @type master_name: str 206 @param master_name: Name of the master node 207 208 """ 209 # Generate cluster secrets 210 GenerateClusterCrypto(True, False, False, False, False) 211 212 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start", constants.NODED]) 213 if result.failed: 214 raise errors.OpExecError("Could not start the node daemon, command %s" 215 " had exitcode %s and error %s" % 216 (result.cmd, result.exit_code, result.output)) 217 218 _WaitForNodeDaemon(master_name)
219 220
221 -def _WaitForNodeDaemon(node_name):
222 """Wait for node daemon to become responsive. 223 224 """ 225 def _CheckNodeDaemon(): 226 # Pylint bug <http://www.logilab.org/ticket/35642> 227 # pylint: disable=E1101 228 result = rpc.BootstrapRunner().call_version([node_name])[node_name] 229 if result.fail_msg: 230 raise utils.RetryAgain()
231 232 try: 233 utils.Retry(_CheckNodeDaemon, 1.0, _DAEMON_READY_TIMEOUT) 234 except utils.RetryTimeout: 235 raise errors.OpExecError("Node daemon on %s didn't answer queries within" 236 " %s seconds" % (node_name, _DAEMON_READY_TIMEOUT)) 237 238
239 -def _WaitForMasterDaemon():
240 """Wait for master daemon to become responsive. 241 242 """ 243 def _CheckMasterDaemon(): 244 try: 245 cl = luxi.Client() 246 (cluster_name, ) = cl.QueryConfigValues(["cluster_name"]) 247 except Exception: 248 raise utils.RetryAgain() 249 250 logging.debug("Received cluster name %s from master", cluster_name)
251 252 try: 253 utils.Retry(_CheckMasterDaemon, 1.0, _DAEMON_READY_TIMEOUT) 254 except utils.RetryTimeout: 255 raise errors.OpExecError("Master daemon didn't answer queries within" 256 " %s seconds" % _DAEMON_READY_TIMEOUT) 257 258
259 -def _WaitForSshDaemon(hostname, port, family):
260 """Wait for SSH daemon to become responsive. 261 262 """ 263 hostip = netutils.GetHostname(name=hostname, family=family).ip 264 265 def _CheckSshDaemon(): 266 if netutils.TcpPing(hostip, port, timeout=1.0, live_port_needed=True): 267 logging.debug("SSH daemon on %s:%s (IP address %s) has become" 268 " responsive", hostname, port, hostip) 269 else: 270 raise utils.RetryAgain()
271 272 try: 273 utils.Retry(_CheckSshDaemon, 1.0, _DAEMON_READY_TIMEOUT) 274 except utils.RetryTimeout: 275 raise errors.OpExecError("SSH daemon on %s:%s (IP address %s) didn't" 276 " become responsive within %s seconds" % 277 (hostname, port, hostip, _DAEMON_READY_TIMEOUT)) 278 279
280 -def RunNodeSetupCmd(cluster_name, node, basecmd, debug, verbose, 281 use_cluster_key, ask_key, strict_host_check, data):
282 """Runs a command to configure something on a remote machine. 283 284 @type cluster_name: string 285 @param cluster_name: Cluster name 286 @type node: string 287 @param node: Node name 288 @type basecmd: string 289 @param basecmd: Base command (path on the remote machine) 290 @type debug: bool 291 @param debug: Enable debug output 292 @type verbose: bool 293 @param verbose: Enable verbose output 294 @type use_cluster_key: bool 295 @param use_cluster_key: See L{ssh.SshRunner.BuildCmd} 296 @type ask_key: bool 297 @param ask_key: See L{ssh.SshRunner.BuildCmd} 298 @type strict_host_check: bool 299 @param strict_host_check: See L{ssh.SshRunner.BuildCmd} 300 @param data: JSON-serializable input data for script (passed to stdin) 301 302 """ 303 cmd = [basecmd] 304 305 # Pass --debug/--verbose to the external script if set on our invocation 306 if debug: 307 cmd.append("--debug") 308 309 if verbose: 310 cmd.append("--verbose") 311 312 family = ssconf.SimpleStore().GetPrimaryIPFamily() 313 srun = ssh.SshRunner(cluster_name, 314 ipv6=(family == netutils.IP6Address.family)) 315 scmd = srun.BuildCmd(node, constants.SSH_LOGIN_USER, 316 utils.ShellQuoteArgs(cmd), 317 batch=False, ask_key=ask_key, quiet=False, 318 strict_host_check=strict_host_check, 319 use_cluster_key=use_cluster_key) 320 321 tempfh = tempfile.TemporaryFile() 322 try: 323 tempfh.write(serializer.DumpJson(data)) 324 tempfh.seek(0) 325 326 result = utils.RunCmd(scmd, interactive=True, input_fd=tempfh) 327 finally: 328 tempfh.close() 329 330 if result.failed: 331 raise errors.OpExecError("Command '%s' failed: %s" % 332 (result.cmd, result.fail_reason)) 333 334 _WaitForSshDaemon(node, netutils.GetDaemonPort(constants.SSH), family)
335 336
337 -def _InitFileStorage(file_storage_dir):
338 """Initialize if needed the file storage. 339 340 @param file_storage_dir: the user-supplied value 341 @return: either empty string (if file storage was disabled at build 342 time) or the normalized path to the storage directory 343 344 """ 345 file_storage_dir = os.path.normpath(file_storage_dir) 346 347 if not os.path.isabs(file_storage_dir): 348 raise errors.OpPrereqError("File storage directory '%s' is not an absolute" 349 " path" % file_storage_dir, errors.ECODE_INVAL) 350 351 if not os.path.exists(file_storage_dir): 352 try: 353 os.makedirs(file_storage_dir, 0750) 354 except OSError, err: 355 raise errors.OpPrereqError("Cannot create file storage directory" 356 " '%s': %s" % (file_storage_dir, err), 357 errors.ECODE_ENVIRON) 358 359 if not os.path.isdir(file_storage_dir): 360 raise errors.OpPrereqError("The file storage directory '%s' is not" 361 " a directory." % file_storage_dir, 362 errors.ECODE_ENVIRON) 363 return file_storage_dir
364 365
366 -def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913, R0914 367 master_netmask, master_netdev, file_storage_dir, 368 shared_file_storage_dir, candidate_pool_size, secondary_ip=None, 369 vg_name=None, beparams=None, nicparams=None, ndparams=None, 370 hvparams=None, diskparams=None, enabled_hypervisors=None, 371 modify_etc_hosts=True, modify_ssh_setup=True, 372 maintain_node_health=False, drbd_helper=None, uid_pool=None, 373 default_iallocator=None, primary_ip_version=None, ipolicy=None, 374 prealloc_wipe_disks=False, use_external_mip_script=False, 375 hv_state=None, disk_state=None, enabled_disk_templates=None):
376 """Initialise the cluster. 377 378 @type candidate_pool_size: int 379 @param candidate_pool_size: master candidate pool size 380 @type enabled_disk_templates: list of string 381 @param enabled_disk_templates: list of disk_templates to be used in this 382 cluster 383 384 """ 385 # TODO: complete the docstring 386 if config.ConfigWriter.IsCluster(): 387 raise errors.OpPrereqError("Cluster is already initialised", 388 errors.ECODE_STATE) 389 390 if not enabled_hypervisors: 391 raise errors.OpPrereqError("Enabled hypervisors list must contain at" 392 " least one member", errors.ECODE_INVAL) 393 invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES 394 if invalid_hvs: 395 raise errors.OpPrereqError("Enabled hypervisors contains invalid" 396 " entries: %s" % invalid_hvs, 397 errors.ECODE_INVAL) 398 399 if not enabled_disk_templates: 400 raise errors.OpPrereqError("Enabled disk templates list must contain at" 401 " least one member", errors.ECODE_INVAL) 402 invalid_disk_templates = \ 403 set(enabled_disk_templates) - constants.DISK_TEMPLATES 404 if invalid_disk_templates: 405 raise errors.OpPrereqError("Enabled disk templates list contains invalid" 406 " entries: %s" % invalid_disk_templates, 407 errors.ECODE_INVAL) 408 409 try: 410 ipcls = netutils.IPAddress.GetClassFromIpVersion(primary_ip_version) 411 except errors.ProgrammerError: 412 raise errors.OpPrereqError("Invalid primary ip version: %d." % 413 primary_ip_version, errors.ECODE_INVAL) 414 415 hostname = netutils.GetHostname(family=ipcls.family) 416 if not ipcls.IsValid(hostname.ip): 417 raise errors.OpPrereqError("This host's IP (%s) is not a valid IPv%d" 418 " address." % (hostname.ip, primary_ip_version), 419 errors.ECODE_INVAL) 420 421 if ipcls.IsLoopback(hostname.ip): 422 raise errors.OpPrereqError("This host's IP (%s) resolves to a loopback" 423 " address. Please fix DNS or %s." % 424 (hostname.ip, pathutils.ETC_HOSTS), 425 errors.ECODE_ENVIRON) 426 427 if not ipcls.Own(hostname.ip): 428 raise errors.OpPrereqError("Inconsistency: this host's name resolves" 429 " to %s,\nbut this ip address does not" 430 " belong to this host" % 431 hostname.ip, errors.ECODE_ENVIRON) 432 433 clustername = netutils.GetHostname(name=cluster_name, family=ipcls.family) 434 435 if netutils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, timeout=5): 436 raise errors.OpPrereqError("Cluster IP already active", 437 errors.ECODE_NOTUNIQUE) 438 439 if not secondary_ip: 440 if primary_ip_version == constants.IP6_VERSION: 441 raise errors.OpPrereqError("When using a IPv6 primary address, a valid" 442 " IPv4 address must be given as secondary", 443 errors.ECODE_INVAL) 444 secondary_ip = hostname.ip 445 446 if not netutils.IP4Address.IsValid(secondary_ip): 447 raise errors.OpPrereqError("Secondary IP address (%s) has to be a valid" 448 " IPv4 address." % secondary_ip, 449 errors.ECODE_INVAL) 450 451 if not netutils.IP4Address.Own(secondary_ip): 452 raise errors.OpPrereqError("You gave %s as secondary IP," 453 " but it does not belong to this host." % 454 secondary_ip, errors.ECODE_ENVIRON) 455 456 if master_netmask is not None: 457 if not ipcls.ValidateNetmask(master_netmask): 458 raise errors.OpPrereqError("CIDR netmask (%s) not valid for IPv%s " % 459 (master_netmask, primary_ip_version), 460 errors.ECODE_INVAL) 461 else: 462 master_netmask = ipcls.iplen 463 464 if vg_name is not None: 465 # Check if volume group is valid 466 vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name, 467 constants.MIN_VG_SIZE) 468 if vgstatus: 469 raise errors.OpPrereqError("Error: %s\nspecify --no-lvm-storage if" 470 " you are not using lvm" % vgstatus, 471 errors.ECODE_INVAL) 472 473 if drbd_helper is not None: 474 try: 475 curr_helper = bdev.BaseDRBD.GetUsermodeHelper() 476 except errors.BlockDeviceError, err: 477 raise errors.OpPrereqError("Error while checking drbd helper" 478 " (specify --no-drbd-storage if you are not" 479 " using drbd): %s" % str(err), 480 errors.ECODE_ENVIRON) 481 if drbd_helper != curr_helper: 482 raise errors.OpPrereqError("Error: requiring %s as drbd helper but %s" 483 " is the current helper" % (drbd_helper, 484 curr_helper), 485 errors.ECODE_INVAL) 486 487 logging.debug("Stopping daemons (if any are running)") 488 result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-all"]) 489 if result.failed: 490 raise errors.OpExecError("Could not stop daemons, command %s" 491 " had exitcode %s and error '%s'" % 492 (result.cmd, result.exit_code, result.output)) 493 494 if constants.ENABLE_FILE_STORAGE: 495 file_storage_dir = _InitFileStorage(file_storage_dir) 496 else: 497 file_storage_dir = "" 498 499 if constants.ENABLE_SHARED_FILE_STORAGE: 500 shared_file_storage_dir = _InitFileStorage(shared_file_storage_dir) 501 else: 502 shared_file_storage_dir = "" 503 504 if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix): 505 raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix, 506 errors.ECODE_INVAL) 507 508 result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev]) 509 if result.failed: 510 raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" % 511 (master_netdev, 512 result.output.strip()), errors.ECODE_INVAL) 513 514 dirs = [(pathutils.RUN_DIR, constants.RUN_DIRS_MODE)] 515 utils.EnsureDirs(dirs) 516 517 objects.UpgradeBeParams(beparams) 518 utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES) 519 utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES) 520 521 objects.NIC.CheckParameterSyntax(nicparams) 522 523 full_ipolicy = objects.FillIPolicy(constants.IPOLICY_DEFAULTS, ipolicy) 524 525 if ndparams is not None: 526 utils.ForceDictType(ndparams, constants.NDS_PARAMETER_TYPES) 527 else: 528 ndparams = dict(constants.NDC_DEFAULTS) 529 530 # This is ugly, as we modify the dict itself 531 # FIXME: Make utils.ForceDictType pure functional or write a wrapper 532 # around it 533 if hv_state: 534 for hvname, hvs_data in hv_state.items(): 535 utils.ForceDictType(hvs_data, constants.HVSTS_PARAMETER_TYPES) 536 hv_state[hvname] = objects.Cluster.SimpleFillHvState(hvs_data) 537 else: 538 hv_state = dict((hvname, constants.HVST_DEFAULTS) 539 for hvname in enabled_hypervisors) 540 541 # FIXME: disk_state has no default values yet 542 if disk_state: 543 for storage, ds_data in disk_state.items(): 544 if storage not in constants.DS_VALID_TYPES: 545 raise errors.OpPrereqError("Invalid storage type in disk state: %s" % 546 storage, errors.ECODE_INVAL) 547 for ds_name, state in ds_data.items(): 548 utils.ForceDictType(state, constants.DSS_PARAMETER_TYPES) 549 ds_data[ds_name] = objects.Cluster.SimpleFillDiskState(state) 550 551 # hvparams is a mapping of hypervisor->hvparams dict 552 for hv_name, hv_params in hvparams.iteritems(): 553 utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES) 554 hv_class = hypervisor.GetHypervisor(hv_name) 555 hv_class.CheckParameterSyntax(hv_params) 556 557 # diskparams is a mapping of disk-template->diskparams dict 558 for template, dt_params in diskparams.items(): 559 param_keys = set(dt_params.keys()) 560 default_param_keys = set(constants.DISK_DT_DEFAULTS[template].keys()) 561 if not (param_keys <= default_param_keys): 562 unknown_params = param_keys - default_param_keys 563 raise errors.OpPrereqError("Invalid parameters for disk template %s:" 564 " %s" % (template, 565 utils.CommaJoin(unknown_params)), 566 errors.ECODE_INVAL) 567 utils.ForceDictType(dt_params, constants.DISK_DT_TYPES) 568 if template == constants.DT_DRBD8 and vg_name is not None: 569 # The default METAVG value is equal to the VG name set at init time, 570 # if provided 571 dt_params[constants.DRBD_DEFAULT_METAVG] = vg_name 572 573 try: 574 utils.VerifyDictOptions(diskparams, constants.DISK_DT_DEFAULTS) 575 except errors.OpPrereqError, err: 576 raise errors.OpPrereqError("While verify diskparam options: %s" % err, 577 errors.ECODE_INVAL) 578 579 # set up ssh config and /etc/hosts 580 rsa_sshkey = "" 581 dsa_sshkey = "" 582 if os.path.isfile(pathutils.SSH_HOST_RSA_PUB): 583 sshline = utils.ReadFile(pathutils.SSH_HOST_RSA_PUB) 584 rsa_sshkey = sshline.split(" ")[1] 585 if os.path.isfile(pathutils.SSH_HOST_DSA_PUB): 586 sshline = utils.ReadFile(pathutils.SSH_HOST_DSA_PUB) 587 dsa_sshkey = sshline.split(" ")[1] 588 if not rsa_sshkey and not dsa_sshkey: 589 raise errors.OpPrereqError("Failed to find SSH public keys", 590 errors.ECODE_ENVIRON) 591 592 if modify_etc_hosts: 593 utils.AddHostToEtcHosts(hostname.name, hostname.ip) 594 595 if modify_ssh_setup: 596 _InitSSHSetup() 597 598 if default_iallocator is not None: 599 alloc_script = utils.FindFile(default_iallocator, 600 constants.IALLOCATOR_SEARCH_PATH, 601 os.path.isfile) 602 if alloc_script is None: 603 raise errors.OpPrereqError("Invalid default iallocator script '%s'" 604 " specified" % default_iallocator, 605 errors.ECODE_INVAL) 606 elif constants.HTOOLS: 607 # htools was enabled at build-time, we default to it 608 if utils.FindFile(constants.IALLOC_HAIL, 609 constants.IALLOCATOR_SEARCH_PATH, 610 os.path.isfile): 611 default_iallocator = constants.IALLOC_HAIL 612 613 now = time.time() 614 615 # init of cluster config file 616 cluster_config = objects.Cluster( 617 serial_no=1, 618 rsahostkeypub=rsa_sshkey, 619 dsahostkeypub=dsa_sshkey, 620 highest_used_port=(constants.FIRST_DRBD_PORT - 1), 621 mac_prefix=mac_prefix, 622 volume_group_name=vg_name, 623 tcpudp_port_pool=set(), 624 master_node=hostname.name, 625 master_ip=clustername.ip, 626 master_netmask=master_netmask, 627 master_netdev=master_netdev, 628 cluster_name=clustername.name, 629 file_storage_dir=file_storage_dir, 630 shared_file_storage_dir=shared_file_storage_dir, 631 enabled_hypervisors=enabled_hypervisors, 632 beparams={constants.PP_DEFAULT: beparams}, 633 nicparams={constants.PP_DEFAULT: nicparams}, 634 ndparams=ndparams, 635 hvparams=hvparams, 636 diskparams=diskparams, 637 candidate_pool_size=candidate_pool_size, 638 modify_etc_hosts=modify_etc_hosts, 639 modify_ssh_setup=modify_ssh_setup, 640 uid_pool=uid_pool, 641 ctime=now, 642 mtime=now, 643 maintain_node_health=maintain_node_health, 644 drbd_usermode_helper=drbd_helper, 645 default_iallocator=default_iallocator, 646 primary_ip_family=ipcls.family, 647 prealloc_wipe_disks=prealloc_wipe_disks, 648 use_external_mip_script=use_external_mip_script, 649 ipolicy=full_ipolicy, 650 hv_state_static=hv_state, 651 disk_state_static=disk_state, 652 enabled_disk_templates=enabled_disk_templates, 653 ) 654 master_node_config = objects.Node(name=hostname.name, 655 primary_ip=hostname.ip, 656 secondary_ip=secondary_ip, 657 serial_no=1, 658 master_candidate=True, 659 offline=False, drained=False, 660 ctime=now, mtime=now, 661 ) 662 InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config) 663 cfg = config.ConfigWriter(offline=True) 664 ssh.WriteKnownHostsFile(cfg, pathutils.SSH_KNOWN_HOSTS_FILE) 665 cfg.Update(cfg.GetClusterInfo(), logging.error) 666 ssconf.WriteSsconfFiles(cfg.GetSsconfValues()) 667 668 # set up the inter-node password and certificate 669 _InitGanetiServerSetup(hostname.name) 670 671 logging.debug("Starting daemons") 672 result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-all"]) 673 if result.failed: 674 raise errors.OpExecError("Could not start daemons, command %s" 675 " had exitcode %s and error %s" % 676 (result.cmd, result.exit_code, result.output)) 677 678 _WaitForMasterDaemon()
679 680
681 -def InitConfig(version, cluster_config, master_node_config, 682 cfg_file=pathutils.CLUSTER_CONF_FILE):
683 """Create the initial cluster configuration. 684 685 It will contain the current node, which will also be the master 686 node, and no instances. 687 688 @type version: int 689 @param version: configuration version 690 @type cluster_config: L{objects.Cluster} 691 @param cluster_config: cluster configuration 692 @type master_node_config: L{objects.Node} 693 @param master_node_config: master node configuration 694 @type cfg_file: string 695 @param cfg_file: configuration file path 696 697 """ 698 uuid_generator = config.TemporaryReservationManager() 699 cluster_config.uuid = uuid_generator.Generate([], utils.NewUUID, 700 _INITCONF_ECID) 701 master_node_config.uuid = uuid_generator.Generate([], utils.NewUUID, 702 _INITCONF_ECID) 703 nodes = { 704 master_node_config.name: master_node_config, 705 } 706 default_nodegroup = objects.NodeGroup( 707 uuid=uuid_generator.Generate([], utils.NewUUID, _INITCONF_ECID), 708 name=constants.INITIAL_NODE_GROUP_NAME, 709 members=[master_node_config.name], 710 diskparams={}, 711 ) 712 nodegroups = { 713 default_nodegroup.uuid: default_nodegroup, 714 } 715 now = time.time() 716 config_data = objects.ConfigData(version=version, 717 cluster=cluster_config, 718 nodegroups=nodegroups, 719 nodes=nodes, 720 instances={}, 721 networks={}, 722 serial_no=1, 723 ctime=now, mtime=now) 724 utils.WriteFile(cfg_file, 725 data=serializer.Dump(config_data.ToDict()), 726 mode=0600)
727 728
729 -def FinalizeClusterDestroy(master):
730 """Execute the last steps of cluster destroy 731 732 This function shuts down all the daemons, completing the destroy 733 begun in cmdlib.LUDestroyOpcode. 734 735 """ 736 cfg = config.ConfigWriter() 737 modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup 738 runner = rpc.BootstrapRunner() 739 740 master_params = cfg.GetMasterNetworkParameters() 741 master_params.name = master 742 ems = cfg.GetUseExternalMipScript() 743 result = runner.call_node_deactivate_master_ip(master_params.name, 744 master_params, ems) 745 746 msg = result.fail_msg 747 if msg: 748 logging.warning("Could not disable the master IP: %s", msg) 749 750 result = runner.call_node_stop_master(master) 751 msg = result.fail_msg 752 if msg: 753 logging.warning("Could not disable the master role: %s", msg) 754 755 result = runner.call_node_leave_cluster(master, modify_ssh_setup) 756 msg = result.fail_msg 757 if msg: 758 logging.warning("Could not shutdown the node daemon and cleanup" 759 " the node: %s", msg)
760 761
762 -def SetupNodeDaemon(opts, cluster_name, node):
763 """Add a node to the cluster. 764 765 This function must be called before the actual opcode, and will ssh 766 to the remote node, copy the needed files, and start ganeti-noded, 767 allowing the master to do the rest via normal rpc calls. 768 769 @param cluster_name: the cluster name 770 @param node: the name of the new node 771 772 """ 773 data = { 774 constants.NDS_CLUSTER_NAME: cluster_name, 775 constants.NDS_NODE_DAEMON_CERTIFICATE: 776 utils.ReadFile(pathutils.NODED_CERT_FILE), 777 constants.NDS_SSCONF: ssconf.SimpleStore().ReadAll(), 778 constants.NDS_START_NODE_DAEMON: True, 779 } 780 781 RunNodeSetupCmd(cluster_name, node, pathutils.NODE_DAEMON_SETUP, 782 opts.debug, opts.verbose, 783 True, opts.ssh_key_check, opts.ssh_key_check, data) 784 785 _WaitForNodeDaemon(node)
786 787
788 -def MasterFailover(no_voting=False):
789 """Failover the master node. 790 791 This checks that we are not already the master, and will cause the 792 current master to cease being master, and the non-master to become 793 new master. 794 795 @type no_voting: boolean 796 @param no_voting: force the operation without remote nodes agreement 797 (dangerous) 798 799 """ 800 sstore = ssconf.SimpleStore() 801 802 old_master, new_master = ssconf.GetMasterAndMyself(sstore) 803 node_list = sstore.GetNodeList() 804 mc_list = sstore.GetMasterCandidates() 805 806 if old_master == new_master: 807 raise errors.OpPrereqError("This commands must be run on the node" 808 " where you want the new master to be." 809 " %s is already the master" % 810 old_master, errors.ECODE_INVAL) 811 812 if new_master not in mc_list: 813 mc_no_master = [name for name in mc_list if name != old_master] 814 raise errors.OpPrereqError("This node is not among the nodes marked" 815 " as master candidates. Only these nodes" 816 " can become masters. Current list of" 817 " master candidates is:\n" 818 "%s" % ("\n".join(mc_no_master)), 819 errors.ECODE_STATE) 820 821 if not no_voting: 822 vote_list = GatherMasterVotes(node_list) 823 824 if vote_list: 825 voted_master = vote_list[0][0] 826 if voted_master is None: 827 raise errors.OpPrereqError("Cluster is inconsistent, most nodes did" 828 " not respond.", errors.ECODE_ENVIRON) 829 elif voted_master != old_master: 830 raise errors.OpPrereqError("I have a wrong configuration, I believe" 831 " the master is %s but the other nodes" 832 " voted %s. Please resync the configuration" 833 " of this node." % 834 (old_master, voted_master), 835 errors.ECODE_STATE) 836 # end checks 837 838 rcode = 0 839 840 logging.info("Setting master to %s, old master: %s", new_master, old_master) 841 842 try: 843 # instantiate a real config writer, as we now know we have the 844 # configuration data 845 cfg = config.ConfigWriter(accept_foreign=True) 846 847 cluster_info = cfg.GetClusterInfo() 848 cluster_info.master_node = new_master 849 # this will also regenerate the ssconf files, since we updated the 850 # cluster info 851 cfg.Update(cluster_info, logging.error) 852 except errors.ConfigurationError, err: 853 logging.error("Error while trying to set the new master: %s", 854 str(err)) 855 return 1 856 857 # if cfg.Update worked, then it means the old master daemon won't be 858 # able now to write its own config file (we rely on locking in both 859 # backend.UploadFile() and ConfigWriter._Write(); hence the next 860 # step is to kill the old master 861 862 logging.info("Stopping the master daemon on node %s", old_master) 863 864 runner = rpc.BootstrapRunner() 865 master_params = cfg.GetMasterNetworkParameters() 866 master_params.name = old_master 867 ems = cfg.GetUseExternalMipScript() 868 result = runner.call_node_deactivate_master_ip(master_params.name, 869 master_params, ems) 870 871 msg = result.fail_msg 872 if msg: 873 logging.warning("Could not disable the master IP: %s", msg) 874 875 result = runner.call_node_stop_master(old_master) 876 msg = result.fail_msg 877 if msg: 878 logging.error("Could not disable the master role on the old master" 879 " %s, please disable manually: %s", old_master, msg) 880 881 logging.info("Checking master IP non-reachability...") 882 883 master_ip = sstore.GetMasterIP() 884 total_timeout = 30 885 886 # Here we have a phase where no master should be running 887 def _check_ip(): 888 if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT): 889 raise utils.RetryAgain()
890 891 try: 892 utils.Retry(_check_ip, (1, 1.5, 5), total_timeout) 893 except utils.RetryTimeout: 894 logging.warning("The master IP is still reachable after %s seconds," 895 " continuing but activating the master on the current" 896 " node will probably fail", total_timeout) 897 898 if jstore.CheckDrainFlag(): 899 logging.info("Undraining job queue") 900 jstore.SetDrainFlag(False) 901 902 logging.info("Starting the master daemons on the new master") 903 904 result = rpc.BootstrapRunner().call_node_start_master_daemons(new_master, 905 no_voting) 906 msg = result.fail_msg 907 if msg: 908 logging.error("Could not start the master role on the new master" 909 " %s, please check: %s", new_master, msg) 910 rcode = 1 911 912 logging.info("Master failed over from %s to %s", old_master, new_master) 913 return rcode 914 915
916 -def GetMaster():
917 """Returns the current master node. 918 919 This is a separate function in bootstrap since it's needed by 920 gnt-cluster, and instead of importing directly ssconf, it's better 921 to abstract it in bootstrap, where we do use ssconf in other 922 functions too. 923 924 """ 925 sstore = ssconf.SimpleStore() 926 927 old_master, _ = ssconf.GetMasterAndMyself(sstore) 928 929 return old_master
930 931
932 -def GatherMasterVotes(node_list):
933 """Check the agreement on who is the master. 934 935 This function will return a list of (node, number of votes), ordered 936 by the number of votes. Errors will be denoted by the key 'None'. 937 938 Note that the sum of votes is the number of nodes this machine 939 knows, whereas the number of entries in the list could be different 940 (if some nodes vote for another master). 941 942 We remove ourselves from the list since we know that (bugs aside) 943 since we use the same source for configuration information for both 944 backend and boostrap, we'll always vote for ourselves. 945 946 @type node_list: list 947 @param node_list: the list of nodes to query for master info; the current 948 node will be removed if it is in the list 949 @rtype: list 950 @return: list of (node, votes) 951 952 """ 953 myself = netutils.Hostname.GetSysName() 954 try: 955 node_list.remove(myself) 956 except ValueError: 957 pass 958 if not node_list: 959 # no nodes left (eventually after removing myself) 960 return [] 961 results = rpc.BootstrapRunner().call_master_info(node_list) 962 if not isinstance(results, dict): 963 # this should not happen (unless internal error in rpc) 964 logging.critical("Can't complete rpc call, aborting master startup") 965 return [(None, len(node_list))] 966 votes = {} 967 for node in results: 968 nres = results[node] 969 data = nres.payload 970 msg = nres.fail_msg 971 fail = False 972 if msg: 973 logging.warning("Error contacting node %s: %s", node, msg) 974 fail = True 975 # for now we accept both length 3, 4 and 5 (data[3] is primary ip version 976 # and data[4] is the master netmask) 977 elif not isinstance(data, (tuple, list)) or len(data) < 3: 978 logging.warning("Invalid data received from node %s: %s", node, data) 979 fail = True 980 if fail: 981 if None not in votes: 982 votes[None] = 0 983 votes[None] += 1 984 continue 985 master_node = data[2] 986 if master_node not in votes: 987 votes[master_node] = 0 988 votes[master_node] += 1 989 990 vote_list = [v for v in votes.items()] 991 # sort first on number of votes then on name, since we want None 992 # sorted later if we have the half of the nodes not responding, and 993 # half voting all for the same master 994 vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True) 995 996 return vote_list
997