ceph/src/pybind/mgr/cephadm/upgrade.py

   1 import json
   2 import logging
   3 import time
   4 import uuid
   5 from typing import TYPE_CHECKING, Optional, Dict, NamedTuple
   6
   7 import orchestrator
   8 from cephadm.utils import name_to_config_section
   9 from orchestrator import OrchestratorError, DaemonDescription
  10
  11 if TYPE_CHECKING:
  12     from .module import CephadmOrchestrator
  13
  14
  15 # ceph daemon types that use the ceph container image.
  16 # NOTE: listed in upgrade order!
  17 CEPH_UPGRADE_ORDER = ['mgr', 'mon', 'crash', 'osd', 'mds', 'rgw', 'rbd-mirror']
  18
  19 logger = logging.getLogger(__name__)
  20
  21
  22 class UpgradeState:
  23     def __init__(self,
  24                  target_name: str,
  25                  progress_id: str,
  26                  target_id: Optional[str] = None,
  27                  repo_digest: Optional[str] = None,
  28                  target_version: Optional[str] = None,
  29                  error: Optional[str] = None,
  30                  paused: Optional[bool] = None,
  31                  ):
  32         self._target_name: str = target_name  # Use CephadmUpgrade.target_image instead.
  33         self.progress_id: str = progress_id
  34         self.target_id: Optional[str] = target_id
  35         self.repo_digest: Optional[str] = repo_digest
  36         self.target_version: Optional[str] = target_version
  37         self.error: Optional[str] = error
  38         self.paused: bool = paused or False
  39
  40     def to_json(self) -> dict:
  41         return {
  42             'target_name': self._target_name,
  43             'progress_id': self.progress_id,
  44             'target_id': self.target_id,
  45             'repo_digest': self.repo_digest,
  46             'target_version': self.target_version,
  47             'error': self.error,
  48             'paused': self.paused,
  49         }
  50
  51     @classmethod
  52     def from_json(cls, data) -> Optional['UpgradeState']:
  53         if data:
  54             return cls(**data)
  55         else:
  56             return None
  57
  58
  59 class CephadmUpgrade:
  60     def __init__(self, mgr: "CephadmOrchestrator"):
  61         self.mgr = mgr
  62
  63         t = self.mgr.get_store('upgrade_state')
  64         if t:
  65             self.upgrade_state: Optional[UpgradeState] = UpgradeState.from_json(json.loads(t))
  66         else:
  67             self.upgrade_state = None
  68
  69     @property
  70     def target_image(self) -> str:
  71         assert self.upgrade_state
  72         if not self.mgr.use_repo_digest:
  73             return self.upgrade_state._target_name
  74         if not self.upgrade_state.repo_digest:
  75             return self.upgrade_state._target_name
  76
  77         return self.upgrade_state.repo_digest
  78
  79     def upgrade_status(self) -> orchestrator.UpgradeStatusSpec:
  80         r = orchestrator.UpgradeStatusSpec()
  81         if self.upgrade_state:
  82             r.target_image = self.target_image
  83             r.in_progress = True
  84             if self.upgrade_state.error:
  85                 r.message = 'Error: ' + self.upgrade_state.error
  86             elif self.upgrade_state.paused:
  87                 r.message = 'Upgrade paused'
  88         return r
  89
  90     def upgrade_start(self, image, version) -> str:
  91         if self.mgr.mode != 'root':
  92             raise OrchestratorError('upgrade is not supported in %s mode' % (
  93                 self.mgr.mode))
  94         if version:
  95             try:
  96                 (major, minor, patch) = version.split('.')
  97                 assert int(minor) >= 0
  98                 assert int(patch) >= 0
  99             except:
 100                 raise OrchestratorError('version must be in the form X.Y.Z (e.g., 15.2.3)')
 101             if int(major) < 15 or (int(major) == 15 and int(minor) < 2):
 102                 raise OrchestratorError('cephadm only supports octopus (15.2.0) or later')
 103             target_name = self.mgr.container_image_base + ':v' + version
 104         elif image:
 105             target_name = image
 106         else:
 107             raise OrchestratorError('must specify either image or version')
 108         if self.upgrade_state:
 109             if self.upgrade_state._target_name != target_name:
 110                 raise OrchestratorError(
 111                     'Upgrade to %s (not %s) already in progress' %
 112                     (self.upgrade_state._target_name, target_name))
 113             if self.upgrade_state.paused:
 114                 self.upgrade_state.paused = False
 115                 self._save_upgrade_state()
 116                 return 'Resumed upgrade to %s' % self.target_image
 117             return 'Upgrade to %s in progress' % self.target_image
 118         self.upgrade_state = UpgradeState(
 119             target_name=target_name,
 120             progress_id=str(uuid.uuid4())
 121         )
 122         self._update_upgrade_progress(0.0)
 123         self._save_upgrade_state()
 124         self._clear_upgrade_health_checks()
 125         self.mgr.event.set()
 126         return 'Initiating upgrade to %s' % (target_name)
 127
 128     def upgrade_pause(self) -> str:
 129         if not self.upgrade_state:
 130             raise OrchestratorError('No upgrade in progress')
 131         if self.upgrade_state.paused:
 132             return 'Upgrade to %s already paused' % self.target_image
 133         self.upgrade_state.paused = True
 134         self._save_upgrade_state()
 135         return 'Paused upgrade to %s' % self.target_image
 136
 137     def upgrade_resume(self) -> str:
 138         if not self.upgrade_state:
 139             raise OrchestratorError('No upgrade in progress')
 140         if not self.upgrade_state.paused:
 141             return 'Upgrade to %s not paused' % self.target_image
 142         self.upgrade_state.paused = False
 143         self._save_upgrade_state()
 144         self.mgr.event.set()
 145         return 'Resumed upgrade to %s' % self.target_image
 146
 147     def upgrade_stop(self) -> str:
 148         if not self.upgrade_state:
 149             return 'No upgrade in progress'
 150         if self.upgrade_state.progress_id:
 151             self.mgr.remote('progress', 'complete',
 152                             self.upgrade_state.progress_id)
 153         target_image = self.target_image
 154         self.upgrade_state = None
 155         self._save_upgrade_state()
 156         self._clear_upgrade_health_checks()
 157         self.mgr.event.set()
 158         return 'Stopped upgrade to %s' % target_image
 159
 160     def continue_upgrade(self) -> bool:
 161         """
 162         Returns false, if nothing was done.
 163         :return:
 164         """
 165         if self.upgrade_state and not self.upgrade_state.paused:
 166             self._do_upgrade()
 167             return True
 168         return False
 169
 170     def _wait_for_ok_to_stop(self, s: DaemonDescription) -> bool:
 171         # only wait a little bit; the service might go away for something
 172         tries = 4
 173         while tries > 0:
 174             if not self.upgrade_state or self.upgrade_state.paused:
 175                 return False
 176
 177             r = self.mgr.cephadm_services[s.daemon_type].ok_to_stop([s.daemon_id])
 178
 179             if not r.retval:
 180                 logger.info(f'Upgrade: {r.stdout}')
 181                 return True
 182             logger.error(f'Upgrade: {r.stderr}')
 183
 184             time.sleep(15)
 185             tries -= 1
 186         return False
 187
 188     def _clear_upgrade_health_checks(self) -> None:
 189         for k in ['UPGRADE_NO_STANDBY_MGR',
 190                   'UPGRADE_FAILED_PULL']:
 191             if k in self.mgr.health_checks:
 192                 del self.mgr.health_checks[k]
 193         self.mgr.set_health_checks(self.mgr.health_checks)
 194
 195     def _fail_upgrade(self, alert_id, alert) -> None:
 196         logger.error('Upgrade: Paused due to %s: %s' % (alert_id,
 197                                                         alert['summary']))
 198         if not self.upgrade_state:
 199             assert False, 'No upgrade in progress'
 200
 201         self.upgrade_state.error = alert_id + ': ' + alert['summary']
 202         self.upgrade_state.paused = True
 203         self._save_upgrade_state()
 204         self.mgr.health_checks[alert_id] = alert
 205         self.mgr.set_health_checks(self.mgr.health_checks)
 206
 207     def _update_upgrade_progress(self, progress) -> None:
 208         if not self.upgrade_state:
 209             assert False, 'No upgrade in progress'
 210
 211         if not self.upgrade_state.progress_id:
 212             self.upgrade_state.progress_id = str(uuid.uuid4())
 213             self._save_upgrade_state()
 214         self.mgr.remote('progress', 'update', self.upgrade_state.progress_id,
 215                         ev_msg='Upgrade to %s' % self.target_image,
 216                         ev_progress=progress)
 217
 218     def _save_upgrade_state(self) -> None:
 219         if not self.upgrade_state:
 220             self.mgr.set_store('upgrade_state', None)
 221             return
 222         self.mgr.set_store('upgrade_state', json.dumps(self.upgrade_state.to_json()))
 223
 224     def get_distinct_container_image_settings(self) -> Dict[str, str]:
 225         # get all distinct container_image settings
 226         image_settings = {}
 227         ret, out, err = self.mgr.check_mon_command({
 228             'prefix': 'config dump',
 229             'format': 'json',
 230         })
 231         config = json.loads(out)
 232         for opt in config:
 233             if opt['name'] == 'container_image':
 234                 image_settings[opt['section']] = opt['value']
 235         return image_settings
 236
 237     def _do_upgrade(self):
 238         # type: () -> None
 239         if not self.upgrade_state:
 240             logger.debug('_do_upgrade no state, exiting')
 241             return
 242
 243         target_image = self.target_image
 244         target_id = self.upgrade_state.target_id
 245         if not target_id or (self.mgr.use_repo_digest and not self.upgrade_state.repo_digest):
 246             # need to learn the container hash
 247             logger.info('Upgrade: First pull of %s' % target_image)
 248             try:
 249                 target_id, target_version, repo_digest = self.mgr._get_container_image_info(
 250                     target_image)
 251             except OrchestratorError as e:
 252                 self._fail_upgrade('UPGRADE_FAILED_PULL', {
 253                     'severity': 'warning',
 254                     'summary': 'Upgrade: failed to pull target image',
 255                     'count': 1,
 256                     'detail': [str(e)],
 257                 })
 258                 return
 259             self.upgrade_state.target_id = target_id
 260             self.upgrade_state.target_version = target_version
 261             self.upgrade_state.repo_digest = repo_digest
 262             self._save_upgrade_state()
 263             target_image = self.target_image
 264         target_version = self.upgrade_state.target_version
 265         logger.info('Upgrade: Target is %s with id %s' % (target_image,
 266                                                           target_id))
 267
 268         image_settings = self.get_distinct_container_image_settings()
 269
 270         daemons = self.mgr.cache.get_daemons()
 271         done = 0
 272         for daemon_type in CEPH_UPGRADE_ORDER:
 273             logger.info('Upgrade: Checking %s daemons...' % daemon_type)
 274             need_upgrade_self = False
 275             for d in daemons:
 276                 if d.daemon_type != daemon_type:
 277                     continue
 278                 if d.container_image_id == target_id:
 279                     logger.debug('daemon %s.%s version correct' % (
 280                         daemon_type, d.daemon_id))
 281                     done += 1
 282                     continue
 283                 logger.debug('daemon %s.%s not correct (%s, %s, %s)' % (
 284                     daemon_type, d.daemon_id,
 285                     d.container_image_name, d.container_image_id, d.version))
 286
 287                 if self.mgr.daemon_is_self(d.daemon_type, d.daemon_id):
 288                     logger.info('Upgrade: Need to upgrade myself (mgr.%s)' %
 289                                 self.mgr.get_mgr_id())
 290                     need_upgrade_self = True
 291                     continue
 292
 293                 # make sure host has latest container image
 294                 out, err, code = self.mgr._run_cephadm(
 295                     d.hostname, '', 'inspect-image', [],
 296                     image=target_image, no_fsid=True, error_ok=True)
 297                 if code or json.loads(''.join(out)).get('image_id') != target_id:
 298                     logger.info('Upgrade: Pulling %s on %s' % (target_image,
 299                                                                d.hostname))
 300                     out, err, code = self.mgr._run_cephadm(
 301                         d.hostname, '', 'pull', [],
 302                         image=target_image, no_fsid=True, error_ok=True)
 303                     if code:
 304                         self._fail_upgrade('UPGRADE_FAILED_PULL', {
 305                             'severity': 'warning',
 306                             'summary': 'Upgrade: failed to pull target image',
 307                             'count': 1,
 308                             'detail': [
 309                                 'failed to pull %s on host %s' % (target_image,
 310                                                                   d.hostname)],
 311                         })
 312                         return
 313                     r = json.loads(''.join(out))
 314                     if r.get('image_id') != target_id:
 315                         logger.info('Upgrade: image %s pull on %s got new image %s (not %s), restarting' % (
 316                             target_image, d.hostname, r['image_id'], target_id))
 317                         self.upgrade_state.target_id = r['image_id']
 318                         self._save_upgrade_state()
 319                         return
 320
 321                 self._update_upgrade_progress(done / len(daemons))
 322
 323                 if not d.container_image_id:
 324                     if d.container_image_name == target_image:
 325                         logger.debug(
 326                             'daemon %s has unknown container_image_id but has correct image name' % (d.name()))
 327                         continue
 328                 if not self._wait_for_ok_to_stop(d):
 329                     return
 330                 logger.info('Upgrade: Redeploying %s.%s' %
 331                             (d.daemon_type, d.daemon_id))
 332                 self.mgr._daemon_action(
 333                     d.daemon_type,
 334                     d.daemon_id,
 335                     d.hostname,
 336                     'redeploy',
 337                     image=target_image
 338                 )
 339                 return
 340
 341             if need_upgrade_self:
 342                 try:
 343                     self.mgr.mgr_service.fail_over()
 344                 except OrchestratorError as e:
 345                     self._fail_upgrade('UPGRADE_NO_STANDBY_MGR', {
 346                         'severity': 'warning',
 347                         'summary': f'Upgrade: {e}',
 348                         'count': 1,
 349                         'detail': [
 350                             'The upgrade process needs to upgrade the mgr, '
 351                             'but it needs at least one standby to proceed.',
 352                         ],
 353                     })
 354                     return
 355
 356                 return  # unreachable code, as fail_over never returns
 357             elif daemon_type == 'mgr':
 358                 if 'UPGRADE_NO_STANDBY_MGR' in self.mgr.health_checks:
 359                     del self.mgr.health_checks['UPGRADE_NO_STANDBY_MGR']
 360                     self.mgr.set_health_checks(self.mgr.health_checks)
 361
 362             # make sure 'ceph versions' agrees
 363             ret, out_ver, err = self.mgr.check_mon_command({
 364                 'prefix': 'versions',
 365             })
 366             j = json.loads(out_ver)
 367             for version, count in j.get(daemon_type, {}).items():
 368                 if version != target_version:
 369                     logger.warning(
 370                         'Upgrade: %d %s daemon(s) are %s != target %s' %
 371                         (count, daemon_type, version, target_version))
 372
 373             # push down configs
 374             if image_settings.get(daemon_type) != target_image:
 375                 logger.info('Upgrade: Setting container_image for all %s...' %
 376                             daemon_type)
 377                 self.mgr.set_container_image(name_to_config_section(daemon_type), target_image)
 378             to_clean = []
 379             for section in image_settings.keys():
 380                 if section.startswith(name_to_config_section(daemon_type) + '.'):
 381                     to_clean.append(section)
 382             if to_clean:
 383                 logger.debug('Upgrade: Cleaning up container_image for %s...' %
 384                              to_clean)
 385                 for section in to_clean:
 386                     ret, image, err = self.mgr.check_mon_command({
 387                         'prefix': 'config rm',
 388                         'name': 'container_image',
 389                         'who': section,
 390                     })
 391
 392             logger.info('Upgrade: All %s daemons are up to date.' %
 393                         daemon_type)
 394
 395         # clean up
 396         logger.info('Upgrade: Finalizing container_image settings')
 397         self.mgr.set_container_image('global', target_image)
 398
 399         for daemon_type in CEPH_UPGRADE_ORDER:
 400             ret, image, err = self.mgr.check_mon_command({
 401                 'prefix': 'config rm',
 402                 'name': 'container_image',
 403                 'who': name_to_config_section(daemon_type),
 404             })
 405
 406         logger.info('Upgrade: Complete!')
 407         if self.upgrade_state.progress_id:
 408             self.mgr.remote('progress', 'complete',
 409                             self.upgrade_state.progress_id)
 410         self.upgrade_state = None
 411         self._save_upgrade_state()
 412         return