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, List, Tuple
   6
   7 import orchestrator
   8 from cephadm.serve import CephadmServe
   9 from cephadm.services.cephadmservice import CephadmDaemonDeploySpec
  10 from cephadm.utils import ceph_release_to_major, name_to_config_section, CEPH_UPGRADE_ORDER, MONITORING_STACK_TYPES
  11 from orchestrator import OrchestratorError, DaemonDescription, daemon_type_to_service
  12
  13 if TYPE_CHECKING:
  14     from .module import CephadmOrchestrator
  15
  16
  17 logger = logging.getLogger(__name__)
  18
  19
  20 def normalize_image_digest(digest: str, default_registry: str) -> str:
  21     # normal case:
  22     #   ceph/ceph -> docker.io/ceph/ceph
  23     # edge cases that shouldn't ever come up:
  24     #   ubuntu -> docker.io/ubuntu    (ubuntu alias for library/ubuntu)
  25     # no change:
  26     #   quay.ceph.io/ceph/ceph -> ceph
  27     #   docker.io/ubuntu -> no change
  28     bits = digest.split('/')
  29     if '.' not in bits[0] or len(bits) < 3:
  30         digest = 'docker.io/' + digest
  31     return digest
  32
  33
  34 class UpgradeState:
  35     def __init__(self,
  36                  target_name: str,
  37                  progress_id: str,
  38                  target_id: Optional[str] = None,
  39                  target_digests: Optional[List[str]] = None,
  40                  target_version: Optional[str] = None,
  41                  error: Optional[str] = None,
  42                  paused: Optional[bool] = None,
  43                  fs_original_max_mds: Optional[Dict[str, int]] = None,
  44                  ):
  45         self._target_name: str = target_name  # Use CephadmUpgrade.target_image instead.
  46         self.progress_id: str = progress_id
  47         self.target_id: Optional[str] = target_id
  48         self.target_digests: Optional[List[str]] = target_digests
  49         self.target_version: Optional[str] = target_version
  50         self.error: Optional[str] = error
  51         self.paused: bool = paused or False
  52         self.fs_original_max_mds: Optional[Dict[str, int]] = fs_original_max_mds
  53
  54     def to_json(self) -> dict:
  55         return {
  56             'target_name': self._target_name,
  57             'progress_id': self.progress_id,
  58             'target_id': self.target_id,
  59             'target_digests': self.target_digests,
  60             'target_version': self.target_version,
  61             'fs_original_max_mds': self.fs_original_max_mds,
  62             'error': self.error,
  63             'paused': self.paused,
  64         }
  65
  66     @classmethod
  67     def from_json(cls, data: dict) -> Optional['UpgradeState']:
  68         if data:
  69             c = {k: v for k, v in data.items()}
  70             if 'repo_digest' in c:
  71                 c['target_digests'] = [c.pop('repo_digest')]
  72             return cls(**c)
  73         else:
  74             return None
  75
  76
  77 class CephadmUpgrade:
  78     UPGRADE_ERRORS = [
  79         'UPGRADE_NO_STANDBY_MGR',
  80         'UPGRADE_FAILED_PULL',
  81         'UPGRADE_REDEPLOY_DAEMON',
  82         'UPGRADE_BAD_TARGET_VERSION',
  83         'UPGRADE_EXCEPTION'
  84     ]
  85
  86     def __init__(self, mgr: "CephadmOrchestrator"):
  87         self.mgr = mgr
  88
  89         t = self.mgr.get_store('upgrade_state')
  90         if t:
  91             self.upgrade_state: Optional[UpgradeState] = UpgradeState.from_json(json.loads(t))
  92         else:
  93             self.upgrade_state = None
  94
  95     @property
  96     def target_image(self) -> str:
  97         assert self.upgrade_state
  98         if not self.mgr.use_repo_digest:
  99             return self.upgrade_state._target_name
 100         if not self.upgrade_state.target_digests:
 101             return self.upgrade_state._target_name
 102
 103         # FIXME: we assume the first digest is the best one to use
 104         return self.upgrade_state.target_digests[0]
 105
 106     def upgrade_status(self) -> orchestrator.UpgradeStatusSpec:
 107         r = orchestrator.UpgradeStatusSpec()
 108         if self.upgrade_state:
 109             r.target_image = self.target_image
 110             r.in_progress = True
 111             r.progress, r.services_complete = self._get_upgrade_info()
 112             # accessing self.upgrade_info_str will throw an exception if it
 113             # has not been set in _do_upgrade yet
 114             try:
 115                 r.message = self.upgrade_info_str
 116             except AttributeError:
 117                 pass
 118             if self.upgrade_state.error:
 119                 r.message = 'Error: ' + self.upgrade_state.error
 120             elif self.upgrade_state.paused:
 121                 r.message = 'Upgrade paused'
 122         return r
 123
 124     def _get_upgrade_info(self) -> Tuple[str, List[str]]:
 125         if not self.upgrade_state or not self.upgrade_state.target_digests:
 126             return '', []
 127
 128         daemons = [d for d in self.mgr.cache.get_daemons() if d.daemon_type in CEPH_UPGRADE_ORDER]
 129
 130         if any(not d.container_image_digests for d in daemons if d.daemon_type == 'mgr'):
 131             return '', []
 132
 133         completed_daemons = [(d.daemon_type, any(d in self.upgrade_state.target_digests for d in (
 134             d.container_image_digests or []))) for d in daemons if d.daemon_type]
 135
 136         done = len([True for completion in completed_daemons if completion[1]])
 137
 138         completed_types = list(set([completion[0] for completion in completed_daemons if all(
 139             c[1] for c in completed_daemons if c[0] == completion[0])]))
 140
 141         return '%s/%s ceph daemons upgraded' % (done, len(daemons)), completed_types
 142
 143     def _check_target_version(self, version: str) -> Optional[str]:
 144         try:
 145             (major, minor, _) = version.split('.', 2)
 146             assert int(minor) >= 0
 147             # patch might be a number or {number}-g{sha1}
 148         except ValueError:
 149             return 'version must be in the form X.Y.Z (e.g., 15.2.3)'
 150         if int(major) < 15 or (int(major) == 15 and int(minor) < 2):
 151             return 'cephadm only supports octopus (15.2.0) or later'
 152
 153         # to far a jump?
 154         current_version = self.mgr.version.split('ceph version ')[1]
 155         (current_major, current_minor, _) = current_version.split('-')[0].split('.', 2)
 156         if int(current_major) < int(major) - 2:
 157             return f'ceph can only upgrade 1 or 2 major versions at a time; {current_version} -> {version} is too big a jump'
 158         if int(current_major) > int(major):
 159             return f'ceph cannot downgrade major versions (from {current_version} to {version})'
 160         if int(current_major) == int(major):
 161             if int(current_minor) > int(minor):
 162                 return f'ceph cannot downgrade to a {"rc" if minor == "1" else "dev"} release'
 163
 164         # check mon min
 165         monmap = self.mgr.get("mon_map")
 166         mon_min = monmap.get("min_mon_release", 0)
 167         if mon_min < int(major) - 2:
 168             return f'min_mon_release ({mon_min}) < target {major} - 2; first complete an upgrade to an earlier release'
 169
 170         # check osd min
 171         osdmap = self.mgr.get("osd_map")
 172         osd_min_name = osdmap.get("require_osd_release", "argonaut")
 173         osd_min = ceph_release_to_major(osd_min_name)
 174         if osd_min < int(major) - 2:
 175             return f'require_osd_release ({osd_min_name} or {osd_min}) < target {major} - 2; first complete an upgrade to an earlier release'
 176
 177         return None
 178
 179     def upgrade_start(self, image: str, version: str) -> str:
 180         if self.mgr.mode != 'root':
 181             raise OrchestratorError('upgrade is not supported in %s mode' % (
 182                 self.mgr.mode))
 183         if version:
 184             version_error = self._check_target_version(version)
 185             if version_error:
 186                 raise OrchestratorError(version_error)
 187             target_name = self.mgr.container_image_base + ':v' + version
 188         elif image:
 189             target_name = normalize_image_digest(image, self.mgr.default_registry)
 190         else:
 191             raise OrchestratorError('must specify either image or version')
 192         if self.upgrade_state:
 193             if self.upgrade_state._target_name != target_name:
 194                 raise OrchestratorError(
 195                     'Upgrade to %s (not %s) already in progress' %
 196                     (self.upgrade_state._target_name, target_name))
 197             if self.upgrade_state.paused:
 198                 self.upgrade_state.paused = False
 199                 self._save_upgrade_state()
 200                 return 'Resumed upgrade to %s' % self.target_image
 201             return 'Upgrade to %s in progress' % self.target_image
 202         self.mgr.log.info('Upgrade: Started with target %s' % target_name)
 203         self.upgrade_state = UpgradeState(
 204             target_name=target_name,
 205             progress_id=str(uuid.uuid4())
 206         )
 207         self._update_upgrade_progress(0.0)
 208         self._save_upgrade_state()
 209         self._clear_upgrade_health_checks()
 210         self.mgr.event.set()
 211         return 'Initiating upgrade to %s' % (target_name)
 212
 213     def upgrade_pause(self) -> str:
 214         if not self.upgrade_state:
 215             raise OrchestratorError('No upgrade in progress')
 216         if self.upgrade_state.paused:
 217             return 'Upgrade to %s already paused' % self.target_image
 218         self.upgrade_state.paused = True
 219         self.mgr.log.info('Upgrade: Paused upgrade to %s' % self.target_image)
 220         self._save_upgrade_state()
 221         return 'Paused upgrade to %s' % self.target_image
 222
 223     def upgrade_resume(self) -> str:
 224         if not self.upgrade_state:
 225             raise OrchestratorError('No upgrade in progress')
 226         if not self.upgrade_state.paused:
 227             return 'Upgrade to %s not paused' % self.target_image
 228         self.upgrade_state.paused = False
 229         self.mgr.log.info('Upgrade: Resumed upgrade to %s' % self.target_image)
 230         self._save_upgrade_state()
 231         self.mgr.event.set()
 232         return 'Resumed upgrade to %s' % self.target_image
 233
 234     def upgrade_stop(self) -> str:
 235         if not self.upgrade_state:
 236             return 'No upgrade in progress'
 237         if self.upgrade_state.progress_id:
 238             self.mgr.remote('progress', 'complete',
 239                             self.upgrade_state.progress_id)
 240         target_image = self.target_image
 241         self.mgr.log.info('Upgrade: Stopped')
 242         self.upgrade_state = None
 243         self._save_upgrade_state()
 244         self._clear_upgrade_health_checks()
 245         self.mgr.event.set()
 246         return 'Stopped upgrade to %s' % target_image
 247
 248     def continue_upgrade(self) -> bool:
 249         """
 250         Returns false, if nothing was done.
 251         :return:
 252         """
 253         if self.upgrade_state and not self.upgrade_state.paused:
 254             try:
 255                 self._do_upgrade()
 256             except Exception as e:
 257                 self._fail_upgrade('UPGRADE_EXCEPTION', {
 258                     'severity': 'error',
 259                     'summary': 'Upgrade: failed due to an unexpected exception',
 260                     'count': 1,
 261                     'detail': [f'Unexpected exception occurred during upgrade process: {str(e)}'],
 262                 })
 263                 return False
 264             return True
 265         return False
 266
 267     def _wait_for_ok_to_stop(
 268             self, s: DaemonDescription,
 269             known: Optional[List[str]] = None,  # NOTE: output argument!
 270     ) -> bool:
 271         # only wait a little bit; the service might go away for something
 272         assert s.daemon_type is not None
 273         assert s.daemon_id is not None
 274         tries = 4
 275         while tries > 0:
 276             if not self.upgrade_state or self.upgrade_state.paused:
 277                 return False
 278
 279             # setting force flag to retain old functionality.
 280             # note that known is an output argument for ok_to_stop()
 281             r = self.mgr.cephadm_services[daemon_type_to_service(s.daemon_type)].ok_to_stop([
 282                 s.daemon_id], known=known, force=True)
 283
 284             if not r.retval:
 285                 logger.info(f'Upgrade: {r.stdout}')
 286                 return True
 287             logger.info(f'Upgrade: {r.stderr}')
 288
 289             time.sleep(15)
 290             tries -= 1
 291         return False
 292
 293     def _clear_upgrade_health_checks(self) -> None:
 294         for k in self.UPGRADE_ERRORS:
 295             if k in self.mgr.health_checks:
 296                 del self.mgr.health_checks[k]
 297         self.mgr.set_health_checks(self.mgr.health_checks)
 298
 299     def _fail_upgrade(self, alert_id: str, alert: dict) -> None:
 300         assert alert_id in self.UPGRADE_ERRORS
 301         if not self.upgrade_state:
 302             # this could happen if the user canceled the upgrade while we
 303             # were doing something
 304             return
 305
 306         logger.error('Upgrade: Paused due to %s: %s' % (alert_id,
 307                                                         alert['summary']))
 308         self.upgrade_state.error = alert_id + ': ' + alert['summary']
 309         self.upgrade_state.paused = True
 310         self._save_upgrade_state()
 311         self.mgr.health_checks[alert_id] = alert
 312         self.mgr.set_health_checks(self.mgr.health_checks)
 313
 314     def _update_upgrade_progress(self, progress: float) -> None:
 315         if not self.upgrade_state:
 316             assert False, 'No upgrade in progress'
 317
 318         if not self.upgrade_state.progress_id:
 319             self.upgrade_state.progress_id = str(uuid.uuid4())
 320             self._save_upgrade_state()
 321         self.mgr.remote('progress', 'update', self.upgrade_state.progress_id,
 322                         ev_msg='Upgrade to %s' % (
 323                             self.upgrade_state.target_version or self.target_image
 324                         ),
 325                         ev_progress=progress,
 326                         add_to_ceph_s=True)
 327
 328     def _save_upgrade_state(self) -> None:
 329         if not self.upgrade_state:
 330             self.mgr.set_store('upgrade_state', None)
 331             return
 332         self.mgr.set_store('upgrade_state', json.dumps(self.upgrade_state.to_json()))
 333
 334     def get_distinct_container_image_settings(self) -> Dict[str, str]:
 335         # get all distinct container_image settings
 336         image_settings = {}
 337         ret, out, err = self.mgr.check_mon_command({
 338             'prefix': 'config dump',
 339             'format': 'json',
 340         })
 341         config = json.loads(out)
 342         for opt in config:
 343             if opt['name'] == 'container_image':
 344                 image_settings[opt['section']] = opt['value']
 345         return image_settings
 346
 347     def _prepare_for_mds_upgrade(
 348         self,
 349         target_major: str,
 350         need_upgrade: List[DaemonDescription]
 351     ) -> bool:
 352         # are any daemons running a different major version?
 353         scale_down = False
 354         for name, info in self.mgr.get("mds_metadata").items():
 355             version = info.get("ceph_version_short")
 356             major_version = None
 357             if version:
 358                 major_version = version.split('.')[0]
 359             if not major_version:
 360                 self.mgr.log.info('Upgrade: mds.%s version is not known, will retry' % name)
 361                 time.sleep(5)
 362                 return False
 363             if int(major_version) < int(target_major):
 364                 scale_down = True
 365
 366         if not scale_down:
 367             self.mgr.log.debug('Upgrade: All MDS daemons run same major version')
 368             return True
 369
 370         # scale down all filesystems to 1 MDS
 371         assert self.upgrade_state
 372         if not self.upgrade_state.fs_original_max_mds:
 373             self.upgrade_state.fs_original_max_mds = {}
 374         fsmap = self.mgr.get("fs_map")
 375         continue_upgrade = True
 376         for i in fsmap.get('filesystems', []):
 377             fs = i["mdsmap"]
 378             fs_id = i["id"]
 379             fs_name = fs["fs_name"]
 380
 381             # scale down this filesystem?
 382             if fs["max_mds"] > 1:
 383                 self.mgr.log.info('Upgrade: Scaling down filesystem %s' % (
 384                     fs_name
 385                 ))
 386                 if fs_id not in self.upgrade_state.fs_original_max_mds:
 387                     self.upgrade_state.fs_original_max_mds[fs_id] = fs['max_mds']
 388                     self._save_upgrade_state()
 389                 ret, out, err = self.mgr.check_mon_command({
 390                     'prefix': 'fs set',
 391                     'fs_name': fs_name,
 392                     'var': 'max_mds',
 393                     'val': '1',
 394                 })
 395                 continue_upgrade = False
 396                 continue
 397
 398             if len(fs['info']) > 1:
 399                 self.mgr.log.info('Upgrade: Waiting for fs %s to scale down to 1 MDS' % (fs_name))
 400                 time.sleep(10)
 401                 continue_upgrade = False
 402                 continue
 403
 404             lone_mds = list(fs['info'].values())[0]
 405             if lone_mds['state'] != 'up:active':
 406                 self.mgr.log.info('Upgrade: Waiting for mds.%s to be up:active (currently %s)' % (
 407                     lone_mds['name'],
 408                     lone_mds['state'],
 409                 ))
 410                 time.sleep(10)
 411                 continue_upgrade = False
 412                 continue
 413
 414         return continue_upgrade
 415
 416     def _do_upgrade(self):
 417         # type: () -> None
 418         if not self.upgrade_state:
 419             logger.debug('_do_upgrade no state, exiting')
 420             return
 421
 422         target_image = self.target_image
 423         target_id = self.upgrade_state.target_id
 424         target_digests = self.upgrade_state.target_digests
 425         target_version = self.upgrade_state.target_version
 426
 427         first = False
 428         if not target_id or not target_version or not target_digests:
 429             # need to learn the container hash
 430             logger.info('Upgrade: First pull of %s' % target_image)
 431             self.upgrade_info_str = 'Doing first pull of %s image' % (target_image)
 432             try:
 433                 target_id, target_version, target_digests = CephadmServe(self.mgr)._get_container_image_info(
 434                     target_image)
 435             except OrchestratorError as e:
 436                 self._fail_upgrade('UPGRADE_FAILED_PULL', {
 437                     'severity': 'warning',
 438                     'summary': 'Upgrade: failed to pull target image',
 439                     'count': 1,
 440                     'detail': [str(e)],
 441                 })
 442                 return
 443             if not target_version:
 444                 self._fail_upgrade('UPGRADE_FAILED_PULL', {
 445                     'severity': 'warning',
 446                     'summary': 'Upgrade: failed to pull target image',
 447                     'count': 1,
 448                     'detail': ['unable to extract ceph version from container'],
 449                 })
 450                 return
 451             self.upgrade_state.target_id = target_id
 452             # extract the version portion of 'ceph version {version} ({sha1})'
 453             self.upgrade_state.target_version = target_version.split(' ')[2]
 454             self.upgrade_state.target_digests = target_digests
 455             self._save_upgrade_state()
 456             target_image = self.target_image
 457             first = True
 458
 459         if target_digests is None:
 460             target_digests = []
 461         if target_version.startswith('ceph version '):
 462             # tolerate/fix upgrade state from older version
 463             self.upgrade_state.target_version = target_version.split(' ')[2]
 464             target_version = self.upgrade_state.target_version
 465         (target_major, _) = target_version.split('.', 1)
 466         target_major_name = self.mgr.lookup_release_name(int(target_major))
 467
 468         if first:
 469             logger.info('Upgrade: Target is version %s (%s)' % (
 470                 target_version, target_major_name))
 471             logger.info('Upgrade: Target container is %s, digests %s' % (
 472                 target_image, target_digests))
 473
 474         version_error = self._check_target_version(target_version)
 475         if version_error:
 476             self._fail_upgrade('UPGRADE_BAD_TARGET_VERSION', {
 477                 'severity': 'error',
 478                 'summary': f'Upgrade: cannot upgrade/downgrade to {target_version}',
 479                 'count': 1,
 480                 'detail': [version_error],
 481             })
 482             return
 483
 484         image_settings = self.get_distinct_container_image_settings()
 485
 486         daemons = [d for d in self.mgr.cache.get_daemons() if d.daemon_type in CEPH_UPGRADE_ORDER]
 487         done = 0
 488         for daemon_type in CEPH_UPGRADE_ORDER:
 489             logger.debug('Upgrade: Checking %s daemons' % daemon_type)
 490
 491             need_upgrade_self = False
 492             need_upgrade: List[Tuple[DaemonDescription, bool]] = []
 493             need_upgrade_deployer: List[Tuple[DaemonDescription, bool]] = []
 494             for d in daemons:
 495                 if d.daemon_type != daemon_type:
 496                     continue
 497                 assert d.daemon_type is not None
 498                 assert d.daemon_id is not None
 499                 correct_digest = False
 500                 if (any(d in target_digests for d in (d.container_image_digests or []))
 501                         or d.daemon_type in MONITORING_STACK_TYPES):
 502                     logger.debug('daemon %s.%s container digest correct' % (
 503                         daemon_type, d.daemon_id))
 504                     correct_digest = True
 505                     if any(d in target_digests for d in (d.deployed_by or [])):
 506                         logger.debug('daemon %s.%s deployed by correct version' % (
 507                             d.daemon_type, d.daemon_id))
 508                         done += 1
 509                         continue
 510
 511                 if self.mgr.daemon_is_self(d.daemon_type, d.daemon_id):
 512                     logger.info('Upgrade: Need to upgrade myself (mgr.%s)' %
 513                                 self.mgr.get_mgr_id())
 514                     need_upgrade_self = True
 515                     continue
 516
 517                 if correct_digest:
 518                     logger.debug('daemon %s.%s not deployed by correct version' % (
 519                         d.daemon_type, d.daemon_id))
 520                     need_upgrade_deployer.append((d, True))
 521                 else:
 522                     logger.debug('daemon %s.%s not correct (%s, %s, %s)' % (
 523                         daemon_type, d.daemon_id,
 524                         d.container_image_name, d.container_image_digests, d.version))
 525                     need_upgrade.append((d, False))
 526
 527             if not need_upgrade_self:
 528                 # only after the mgr itself is upgraded can we expect daemons to have
 529                 # deployed_by == target_digests
 530                 need_upgrade += need_upgrade_deployer
 531
 532             # prepare filesystems for daemon upgrades?
 533             if (
 534                 daemon_type == 'mds'
 535                 and need_upgrade
 536                 and not self._prepare_for_mds_upgrade(target_major, [d_entry[0] for d_entry in need_upgrade])
 537             ):
 538                 return
 539
 540             if need_upgrade:
 541                 self.upgrade_info_str = 'Currently upgrading %s daemons' % (daemon_type)
 542
 543             to_upgrade: List[Tuple[DaemonDescription, bool]] = []
 544             known_ok_to_stop: List[str] = []
 545             for d_entry in need_upgrade:
 546                 d = d_entry[0]
 547                 assert d.daemon_type is not None
 548                 assert d.daemon_id is not None
 549                 assert d.hostname is not None
 550
 551                 if not d.container_image_id:
 552                     if d.container_image_name == target_image:
 553                         logger.debug(
 554                             'daemon %s has unknown container_image_id but has correct image name' % (d.name()))
 555                         continue
 556
 557                 if known_ok_to_stop:
 558                     if d.name() in known_ok_to_stop:
 559                         logger.info(f'Upgrade: {d.name()} is also safe to restart')
 560                         to_upgrade.append(d_entry)
 561                     continue
 562
 563                 if d.daemon_type in ['mon', 'osd', 'mds']:
 564                     # NOTE: known_ok_to_stop is an output argument for
 565                     # _wait_for_ok_to_stop
 566                     if not self._wait_for_ok_to_stop(d, known_ok_to_stop):
 567                         return
 568
 569                 to_upgrade.append(d_entry)
 570
 571                 # if we don't have a list of others to consider, stop now
 572                 if not known_ok_to_stop:
 573                     break
 574
 575             num = 1
 576             for d_entry in to_upgrade:
 577                 d = d_entry[0]
 578                 assert d.daemon_type is not None
 579                 assert d.daemon_id is not None
 580                 assert d.hostname is not None
 581
 582                 self._update_upgrade_progress(done / len(daemons))
 583
 584                 # make sure host has latest container image
 585                 out, errs, code = CephadmServe(self.mgr)._run_cephadm(
 586                     d.hostname, '', 'inspect-image', [],
 587                     image=target_image, no_fsid=True, error_ok=True)
 588                 if code or not any(d in target_digests for d in json.loads(''.join(out)).get('repo_digests', [])):
 589                     logger.info('Upgrade: Pulling %s on %s' % (target_image,
 590                                                                d.hostname))
 591                     self.upgrade_info_str = 'Pulling %s image on host %s' % (
 592                         target_image, d.hostname)
 593                     out, errs, code = CephadmServe(self.mgr)._run_cephadm(
 594                         d.hostname, '', 'pull', [],
 595                         image=target_image, no_fsid=True, error_ok=True)
 596                     if code:
 597                         self._fail_upgrade('UPGRADE_FAILED_PULL', {
 598                             'severity': 'warning',
 599                             'summary': 'Upgrade: failed to pull target image',
 600                             'count': 1,
 601                             'detail': [
 602                                 'failed to pull %s on host %s' % (target_image,
 603                                                                   d.hostname)],
 604                         })
 605                         return
 606                     r = json.loads(''.join(out))
 607                     if not any(d in target_digests for d in r.get('repo_digests', [])):
 608                         logger.info('Upgrade: image %s pull on %s got new digests %s (not %s), restarting' % (
 609                             target_image, d.hostname, r['repo_digests'], target_digests))
 610                         self.upgrade_info_str = 'Image %s pull on %s got new digests %s (not %s), restarting' % (
 611                             target_image, d.hostname, r['repo_digests'], target_digests)
 612                         self.upgrade_state.target_digests = r['repo_digests']
 613                         self._save_upgrade_state()
 614                         return
 615
 616                     self.upgrade_info_str = 'Currently upgrading %s daemons' % (daemon_type)
 617
 618                 if len(to_upgrade) > 1:
 619                     logger.info('Upgrade: Updating %s.%s (%d/%d)' %
 620                                 (d.daemon_type, d.daemon_id, num, len(to_upgrade)))
 621                 else:
 622                     logger.info('Upgrade: Updating %s.%s' %
 623                                 (d.daemon_type, d.daemon_id))
 624                 action = 'Upgrading' if not d_entry[1] else 'Redeploying'
 625                 try:
 626                     daemon_spec = CephadmDaemonDeploySpec.from_daemon_description(d)
 627                     self.mgr._daemon_action(
 628                         daemon_spec,
 629                         'redeploy',
 630                         image=target_image if not d_entry[1] else None
 631                     )
 632                 except Exception as e:
 633                     self._fail_upgrade('UPGRADE_REDEPLOY_DAEMON', {
 634                         'severity': 'warning',
 635                         'summary': f'{action} daemon {d.name()} on host {d.hostname} failed.',
 636                         'count': 1,
 637                         'detail': [
 638                             f'Upgrade daemon: {d.name()}: {e}'
 639                         ],
 640                     })
 641                     return
 642                 num += 1
 643             if to_upgrade:
 644                 return
 645
 646             # complete mon upgrade?
 647             if daemon_type == 'mon':
 648                 if not self.mgr.get("have_local_config_map"):
 649                     logger.info('Upgrade: Restarting mgr now that mons are running pacific')
 650                     need_upgrade_self = True
 651
 652             if need_upgrade_self:
 653                 try:
 654                     self.mgr.mgr_service.fail_over()
 655                 except OrchestratorError as e:
 656                     self._fail_upgrade('UPGRADE_NO_STANDBY_MGR', {
 657                         'severity': 'warning',
 658                         'summary': f'Upgrade: {e}',
 659                         'count': 1,
 660                         'detail': [
 661                             'The upgrade process needs to upgrade the mgr, '
 662                             'but it needs at least one standby to proceed.',
 663                         ],
 664                     })
 665                     return
 666
 667                 return  # unreachable code, as fail_over never returns
 668             elif daemon_type == 'mgr':
 669                 if 'UPGRADE_NO_STANDBY_MGR' in self.mgr.health_checks:
 670                     del self.mgr.health_checks['UPGRADE_NO_STANDBY_MGR']
 671                     self.mgr.set_health_checks(self.mgr.health_checks)
 672
 673             # make sure 'ceph versions' agrees
 674             ret, out_ver, err = self.mgr.check_mon_command({
 675                 'prefix': 'versions',
 676             })
 677             j = json.loads(out_ver)
 678             for version, count in j.get(daemon_type, {}).items():
 679                 short_version = version.split(' ')[2]
 680                 if short_version != target_version:
 681                     logger.warning(
 682                         'Upgrade: %d %s daemon(s) are %s != target %s' %
 683                         (count, daemon_type, short_version, target_version))
 684
 685             # push down configs
 686             daemon_type_section = name_to_config_section(daemon_type)
 687             if image_settings.get(daemon_type_section) != target_image:
 688                 logger.info('Upgrade: Setting container_image for all %s' %
 689                             daemon_type)
 690                 self.mgr.set_container_image(daemon_type_section, target_image)
 691             to_clean = []
 692             for section in image_settings.keys():
 693                 if section.startswith(name_to_config_section(daemon_type) + '.'):
 694                     to_clean.append(section)
 695             if to_clean:
 696                 logger.debug('Upgrade: Cleaning up container_image for %s' %
 697                              to_clean)
 698                 for section in to_clean:
 699                     ret, image, err = self.mgr.check_mon_command({
 700                         'prefix': 'config rm',
 701                         'name': 'container_image',
 702                         'who': section,
 703                     })
 704
 705             logger.debug('Upgrade: All %s daemons are up to date.' % daemon_type)
 706
 707             # complete osd upgrade?
 708             if daemon_type == 'osd':
 709                 osdmap = self.mgr.get("osd_map")
 710                 osd_min_name = osdmap.get("require_osd_release", "argonaut")
 711                 osd_min = ceph_release_to_major(osd_min_name)
 712                 if osd_min < int(target_major):
 713                     logger.info(
 714                         f'Upgrade: Setting require_osd_release to {target_major} {target_major_name}')
 715                     ret, _, err = self.mgr.check_mon_command({
 716                         'prefix': 'osd require-osd-release',
 717                         'release': target_major_name,
 718                     })
 719
 720             # complete mds upgrade?
 721             if daemon_type == 'mds' and self.upgrade_state.fs_original_max_mds:
 722                 for i in self.mgr.get("fs_map")['filesystems']:
 723                     fs_id = i["id"]
 724                     fs_name = i['mdsmap']['fs_name']
 725                     new_max = self.upgrade_state.fs_original_max_mds.get(fs_id)
 726                     if new_max:
 727                         self.mgr.log.info('Upgrade: Scaling up filesystem %s max_mds to %d' % (
 728                             fs_name, new_max
 729                         ))
 730                         ret, _, err = self.mgr.check_mon_command({
 731                             'prefix': 'fs set',
 732                             'fs_name': fs_name,
 733                             'var': 'max_mds',
 734                             'val': str(new_max),
 735                         })
 736
 737                 self.upgrade_state.fs_original_max_mds = {}
 738                 self._save_upgrade_state()
 739
 740         # clean up
 741         logger.info('Upgrade: Finalizing container_image settings')
 742         self.mgr.set_container_image('global', target_image)
 743
 744         for daemon_type in CEPH_UPGRADE_ORDER:
 745             ret, image, err = self.mgr.check_mon_command({
 746                 'prefix': 'config rm',
 747                 'name': 'container_image',
 748                 'who': name_to_config_section(daemon_type),
 749             })
 750
 751         logger.info('Upgrade: Complete!')
 752         if self.upgrade_state.progress_id:
 753             self.mgr.remote('progress', 'complete',
 754                             self.upgrade_state.progress_id)
 755         self.upgrade_state = None
 756         self._save_upgrade_state()
 757         return