From 19fcec84d8d7d21e796c7624e521b60d28ee21ed Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Sun, 7 Apr 2024 20:45:59 +0200 Subject: Adding upstream version 16.2.11+ds. Signed-off-by: Daniel Baumann --- src/pybind/mgr/cephadm/migrations.py | 333 +++++++++++++++++++++++++++++++++++ 1 file changed, 333 insertions(+) create mode 100644 src/pybind/mgr/cephadm/migrations.py (limited to 'src/pybind/mgr/cephadm/migrations.py') diff --git a/src/pybind/mgr/cephadm/migrations.py b/src/pybind/mgr/cephadm/migrations.py new file mode 100644 index 000000000..f4a3056b2 --- /dev/null +++ b/src/pybind/mgr/cephadm/migrations.py @@ -0,0 +1,333 @@ +import json +import logging +from typing import TYPE_CHECKING, Iterator, Optional, Dict, Any + +from ceph.deployment.service_spec import PlacementSpec, ServiceSpec, HostPlacementSpec +from cephadm.schedule import HostAssignment +import rados + +from mgr_module import NFS_POOL_NAME +from orchestrator import OrchestratorError, DaemonDescription + +if TYPE_CHECKING: + from .module import CephadmOrchestrator + +LAST_MIGRATION = 5 + +logger = logging.getLogger(__name__) + + +class Migrations: + def __init__(self, mgr: "CephadmOrchestrator"): + self.mgr = mgr + + # Why having a global counter, instead of spec versions? + # + # for the first migration: + # The specs don't change in (this) migration. but the scheduler here. + # Adding the version to the specs at this time just felt wrong to me. + # + # And the specs are only another part of cephadm which needs potential upgrades. + # We have the cache, the inventory, the config store, the upgrade (imagine changing the + # upgrade code, while an old upgrade is still in progress), naming of daemons, + # fs-layout of the daemons, etc. + if self.mgr.migration_current is None: + self.set(LAST_MIGRATION) + + v = mgr.get_store('nfs_migration_queue') + self.nfs_migration_queue = json.loads(v) if v else [] + + # for some migrations, we don't need to do anything except for + # incrementing migration_current. + # let's try to shortcut things here. + self.migrate(True) + + def set(self, val: int) -> None: + self.mgr.set_module_option('migration_current', val) + self.mgr.migration_current = val + + def is_migration_ongoing(self) -> bool: + return self.mgr.migration_current != LAST_MIGRATION + + def verify_no_migration(self) -> None: + if self.is_migration_ongoing(): + # this is raised in module.serve() + raise OrchestratorError( + "cephadm migration still ongoing. Please wait, until the migration is complete.") + + def migrate(self, startup: bool = False) -> None: + if self.mgr.migration_current == 0: + if self.migrate_0_1(): + self.set(1) + + if self.mgr.migration_current == 1: + if self.migrate_1_2(): + self.set(2) + + if self.mgr.migration_current == 2 and not startup: + if self.migrate_2_3(): + self.set(3) + + if self.mgr.migration_current == 3: + if self.migrate_3_4(): + self.set(4) + + if self.mgr.migration_current == 4: + if self.migrate_4_5(): + self.set(5) + + def migrate_0_1(self) -> bool: + """ + Migration 0 -> 1 + New scheduler that takes PlacementSpec as the bound and not as recommendation. + I.e. the new scheduler won't suggest any new placements outside of the hosts + specified by label etc. + + Which means, we have to make sure, we're not removing any daemons directly after + upgrading to the new scheduler. + + There is a potential race here: + 1. user updates his spec to remove daemons + 2. mgr gets upgraded to new scheduler, before the old scheduler removed the daemon + 3. now, we're converting the spec to explicit placement, thus reverting (1.) + I think this is ok. + """ + + def interesting_specs() -> Iterator[ServiceSpec]: + for s in self.mgr.spec_store.all_specs.values(): + if s.unmanaged: + continue + p = s.placement + if p is None: + continue + if p.count is None: + continue + if not p.hosts and not p.host_pattern and not p.label: + continue + yield s + + def convert_to_explicit(spec: ServiceSpec) -> None: + existing_daemons = self.mgr.cache.get_daemons_by_service(spec.service_name()) + placements, to_add, to_remove = HostAssignment( + spec=spec, + hosts=self.mgr.inventory.all_specs(), + unreachable_hosts=self.mgr._unreachable_hosts(), + daemons=existing_daemons, + ).place() + + # We have to migrate, only if the new scheduler would remove daemons + if len(placements) >= len(existing_daemons): + return + + def to_hostname(d: DaemonDescription) -> HostPlacementSpec: + if d.hostname in old_hosts: + return old_hosts[d.hostname] + else: + assert d.hostname + return HostPlacementSpec(d.hostname, '', '') + + old_hosts = {h.hostname: h for h in spec.placement.hosts} + new_hosts = [to_hostname(d) for d in existing_daemons] + + new_placement = PlacementSpec( + hosts=new_hosts, + count=spec.placement.count + ) + + new_spec = ServiceSpec.from_json(spec.to_json()) + new_spec.placement = new_placement + + logger.info(f"Migrating {spec.one_line_str()} to explicit placement") + + self.mgr.spec_store.save(new_spec) + + specs = list(interesting_specs()) + if not specs: + return True # nothing to do. shortcut + + if not self.mgr.cache.daemon_cache_filled(): + logger.info("Unable to migrate yet. Daemon Cache still incomplete.") + return False + + for spec in specs: + convert_to_explicit(spec) + + return True + + def migrate_1_2(self) -> bool: + """ + After 15.2.4, we unified some service IDs: MONs, MGRs etc no longer have a service id. + Which means, the service names changed: + + mon.foo -> mon + mgr.foo -> mgr + + This fixes the data structure consistency + """ + bad_specs = {} + for name, spec in self.mgr.spec_store.all_specs.items(): + if name != spec.service_name(): + bad_specs[name] = (spec.service_name(), spec) + + for old, (new, old_spec) in bad_specs.items(): + if new not in self.mgr.spec_store.all_specs: + spec = old_spec + else: + spec = self.mgr.spec_store.all_specs[new] + spec.unmanaged = True + self.mgr.spec_store.save(spec) + self.mgr.spec_store.finally_rm(old) + + return True + + def migrate_2_3(self) -> bool: + if self.nfs_migration_queue: + from nfs.cluster import create_ganesha_pool + + create_ganesha_pool(self.mgr) + for service_id, pool, ns in self.nfs_migration_queue: + if pool != '.nfs': + self.migrate_nfs_spec(service_id, pool, ns) + self.nfs_migration_queue = [] + self.mgr.log.info('Done migrating all NFS services') + return True + + def migrate_nfs_spec(self, service_id: str, pool: str, ns: Optional[str]) -> None: + renamed = False + if service_id.startswith('ganesha-'): + service_id = service_id[8:] + renamed = True + + self.mgr.log.info( + f'Migrating nfs.{service_id} from legacy pool {pool} namespace {ns}' + ) + + # read exports + ioctx = self.mgr.rados.open_ioctx(pool) + if ns is not None: + ioctx.set_namespace(ns) + object_iterator = ioctx.list_objects() + exports = [] + while True: + try: + obj = object_iterator.__next__() + if obj.key.startswith('export-'): + self.mgr.log.debug(f'reading {obj.key}') + exports.append(obj.read().decode()) + except StopIteration: + break + self.mgr.log.info(f'Found {len(exports)} exports for legacy nfs.{service_id}') + + # copy grace file + if service_id != ns: + try: + grace = ioctx.read("grace") + new_ioctx = self.mgr.rados.open_ioctx(NFS_POOL_NAME) + new_ioctx.set_namespace(service_id) + new_ioctx.write_full("grace", grace) + self.mgr.log.info('Migrated nfs-ganesha grace file') + except rados.ObjectNotFound: + self.mgr.log.debug('failed to read old grace file; skipping') + + if renamed and f'nfs.ganesha-{service_id}' in self.mgr.spec_store: + # rename from nfs.ganesha-* to nfs.*. This will destroy old daemons and + # deploy new ones. + self.mgr.log.info(f'Replacing nfs.ganesha-{service_id} with nfs.{service_id}') + spec = self.mgr.spec_store[f'nfs.ganesha-{service_id}'].spec + self.mgr.spec_store.rm(f'nfs.ganesha-{service_id}') + spec.service_id = service_id + self.mgr.spec_store.save(spec, True) + + # We have to remove the old daemons here as well, otherwise we'll end up with a port conflict. + daemons = [d.name() + for d in self.mgr.cache.get_daemons_by_service(f'nfs.ganesha-{service_id}')] + self.mgr.log.info(f'Removing old nfs.ganesha-{service_id} daemons {daemons}') + self.mgr.remove_daemons(daemons) + else: + # redeploy all ganesha daemons to ensures that the daemon + # cephx are correct AND container configs are set up properly + daemons = [d.name() for d in self.mgr.cache.get_daemons_by_service(f'nfs.{service_id}')] + self.mgr.log.info(f'Removing old nfs.{service_id} daemons {daemons}') + self.mgr.remove_daemons(daemons) + + # re-save service spec (without pool and namespace properties!) + spec = self.mgr.spec_store[f'nfs.{service_id}'].spec + self.mgr.spec_store.save(spec) + + # import exports + for export in exports: + ex = '' + for line in export.splitlines(): + if ( + line.startswith(' secret_access_key =') + or line.startswith(' user_id =') + ): + continue + ex += line + '\n' + self.mgr.log.debug(f'importing export: {ex}') + ret, out, err = self.mgr.mon_command({ + 'prefix': 'nfs export apply', + 'cluster_id': service_id + }, inbuf=ex) + if ret: + self.mgr.log.warning(f'Failed to migrate export ({ret}): {err}\nExport was:\n{ex}') + self.mgr.log.info(f'Done migrating nfs.{service_id}') + + def migrate_3_4(self) -> bool: + # We can't set any host with the _admin label, but we're + # going to warn when calling `ceph orch host rm...` + if 'client.admin' not in self.mgr.keys.keys: + self.mgr._client_keyring_set( + entity='client.admin', + placement='label:_admin', + ) + return True + + def migrate_4_5(self) -> bool: + registry_url = self.mgr.get_module_option('registry_url') + registry_username = self.mgr.get_module_option('registry_username') + registry_password = self.mgr.get_module_option('registry_password') + if registry_url and registry_username and registry_password: + + registry_credentials = {'url': registry_url, + 'username': registry_username, 'password': registry_password} + self.mgr.set_store('registry_credentials', json.dumps(registry_credentials)) + + self.mgr.set_module_option('registry_url', None) + self.mgr.check_mon_command({ + 'prefix': 'config rm', + 'who': 'mgr', + 'key': 'mgr/cephadm/registry_url', + }) + self.mgr.set_module_option('registry_username', None) + self.mgr.check_mon_command({ + 'prefix': 'config rm', + 'who': 'mgr', + 'key': 'mgr/cephadm/registry_username', + }) + self.mgr.set_module_option('registry_password', None) + self.mgr.check_mon_command({ + 'prefix': 'config rm', + 'who': 'mgr', + 'key': 'mgr/cephadm/registry_password', + }) + + self.mgr.log.info('Done migrating registry login info') + return True + + +def queue_migrate_nfs_spec(mgr: "CephadmOrchestrator", spec_dict: Dict[Any, Any]) -> None: + """ + After 16.2.5 we dropped the NFSServiceSpec pool and namespace properties. + Queue up a migration to process later, once we are sure that RADOS is available + and so on. + """ + service_id = spec_dict['spec']['service_id'] + args = spec_dict['spec'].get('spec', {}) + pool = args.pop('pool', 'nfs-ganesha') + ns = args.pop('namespace', service_id) + queued = mgr.get_store('nfs_migration_queue') or '[]' + ls = json.loads(queued) + ls.append([service_id, pool, ns]) + mgr.set_store('nfs_migration_queue', json.dumps(ls)) + mgr.log.info(f'Queued nfs.{service_id} for migration') -- cgit v1.2.3