diff options
Diffstat (limited to 'src/ceph-volume/ceph_volume/devices/lvm/batch.py')
| -rw-r--r-- | src/ceph-volume/ceph_volume/devices/lvm/batch.py | 654 |
1 files changed, 654 insertions, 0 deletions
diff --git a/src/ceph-volume/ceph_volume/devices/lvm/batch.py b/src/ceph-volume/ceph_volume/devices/lvm/batch.py new file mode 100644 index 000000000..c97d3a25b --- /dev/null +++ b/src/ceph-volume/ceph_volume/devices/lvm/batch.py @@ -0,0 +1,654 @@ +import argparse +from collections import namedtuple +import json +import logging +from textwrap import dedent +from ceph_volume import terminal, decorators +from ceph_volume.util import disk, prompt_bool, arg_validators, templates +from ceph_volume.util import prepare +from . import common +from .create import Create +from .prepare import Prepare + +mlogger = terminal.MultiLogger(__name__) +logger = logging.getLogger(__name__) + + +device_list_template = """ + * {path: <25} {size: <10} {state}""" + + +def device_formatter(devices): + lines = [] + for path, details in devices: + lines.append(device_list_template.format( + path=path, size=details['human_readable_size'], + state='solid' if details['rotational'] == '0' else 'rotational') + ) + + return ''.join(lines) + + +def ensure_disjoint_device_lists(data, db=[], wal=[], journal=[]): + # check that all device lists are disjoint with each other + if not all([set(data).isdisjoint(set(db)), + set(data).isdisjoint(set(wal)), + set(data).isdisjoint(set(journal)), + set(db).isdisjoint(set(wal))]): + raise Exception('Device lists are not disjoint') + + +def separate_devices_from_lvs(devices): + phys = [] + lvm = [] + for d in devices: + phys.append(d) if d.is_device else lvm.append(d) + return phys, lvm + + +def get_physical_osds(devices, args): + ''' + Goes through passed physical devices and assigns OSDs + ''' + data_slots = args.osds_per_device + if args.data_slots: + data_slots = max(args.data_slots, args.osds_per_device) + rel_data_size = 1.0 / data_slots + mlogger.debug('relative data size: {}'.format(rel_data_size)) + ret = [] + for dev in devices: + if dev.available_lvm: + dev_size = dev.vg_size[0] + abs_size = disk.Size(b=int(dev_size * rel_data_size)) + free_size = dev.vg_free[0] + for _ in range(args.osds_per_device): + if abs_size > free_size: + break + free_size -= abs_size.b + osd_id = None + if args.osd_ids: + osd_id = args.osd_ids.pop() + ret.append(Batch.OSD(dev.path, + rel_data_size, + abs_size, + args.osds_per_device, + osd_id, + 'dmcrypt' if args.dmcrypt else None, + dev.symlink)) + return ret + + +def get_lvm_osds(lvs, args): + ''' + Goes through passed LVs and assigns planned osds + ''' + ret = [] + for lv in lvs: + if lv.used_by_ceph: + continue + osd_id = None + if args.osd_ids: + osd_id = args.osd_ids.pop() + osd = Batch.OSD("{}/{}".format(lv.vg_name, lv.lv_name), + 100.0, + disk.Size(b=int(lv.lvs[0].lv_size)), + 1, + osd_id, + 'dmcrypt' if args.dmcrypt else None) + ret.append(osd) + return ret + + +def get_physical_fast_allocs(devices, type_, fast_slots_per_device, new_osds, args): + requested_slots = getattr(args, '{}_slots'.format(type_)) + if not requested_slots or requested_slots < fast_slots_per_device: + if requested_slots: + mlogger.info('{}_slots argument is too small, ignoring'.format(type_)) + requested_slots = fast_slots_per_device + + requested_size = getattr(args, '{}_size'.format(type_), 0) + if not requested_size or requested_size == 0: + # no size argument was specified, check ceph.conf + get_size_fct = getattr(prepare, 'get_{}_size'.format(type_)) + requested_size = get_size_fct(lv_format=False) + + ret = [] + vg_device_map = group_devices_by_vg(devices) + for vg_devices in vg_device_map.values(): + for dev in vg_devices: + if not dev.available_lvm: + continue + # any LV present is considered a taken slot + occupied_slots = len(dev.lvs) + dev_size = dev.vg_size[0] + # this only looks at the first vg on device, unsure if there is a better + # way + abs_size = disk.Size(b=int(dev_size / requested_slots)) + free_size = dev.vg_free[0] + relative_size = int(abs_size) / dev_size + if requested_size: + if requested_size <= abs_size: + abs_size = requested_size + relative_size = int(abs_size) / dev_size + else: + mlogger.error( + '{} was requested for {}, but only {} can be fulfilled'.format( + requested_size, + '{}_size'.format(type_), + abs_size, + )) + exit(1) + while abs_size <= free_size and len(ret) < new_osds and occupied_slots < fast_slots_per_device: + free_size -= abs_size.b + occupied_slots += 1 + ret.append((dev.path, relative_size, abs_size, requested_slots)) + return ret + +def group_devices_by_vg(devices): + result = dict() + result['unused_devices'] = [] + for dev in devices: + if len(dev.vgs) > 0: + # already using assumption that a PV only belongs to single VG in other places + vg_name = dev.vgs[0].name + if vg_name in result: + result[vg_name].append(dev) + else: + result[vg_name] = [dev] + else: + result['unused_devices'].append(dev) + return result + +def get_lvm_fast_allocs(lvs): + return [("{}/{}".format(d.vg_name, d.lv_name), 100.0, + disk.Size(b=int(d.lvs[0].lv_size)), 1) for d in lvs if not + d.used_by_ceph] + + +class Batch(object): + + help = 'Automatically size devices for multi-OSD provisioning with minimal interaction' + + _help = dedent(""" + Automatically size devices ready for OSD provisioning based on default strategies. + + Usage: + + ceph-volume lvm batch [DEVICE...] + + Devices can be physical block devices or LVs. + Optional reporting on possible outcomes is enabled with --report + + ceph-volume lvm batch --report [DEVICE...] + """) + + def __init__(self, argv): + parser = argparse.ArgumentParser( + prog='ceph-volume lvm batch', + formatter_class=argparse.RawDescriptionHelpFormatter, + description=self._help, + ) + + parser.add_argument( + 'devices', + metavar='DEVICES', + nargs='*', + type=arg_validators.ValidBatchDataDevice(), + default=[], + help='Devices to provision OSDs', + ) + parser.add_argument( + '--db-devices', + nargs='*', + type=arg_validators.ValidBatchDevice(), + default=[], + help='Devices to provision OSDs db volumes', + ) + parser.add_argument( + '--wal-devices', + nargs='*', + type=arg_validators.ValidBatchDevice(), + default=[], + help='Devices to provision OSDs wal volumes', + ) + parser.add_argument( + '--journal-devices', + nargs='*', + type=arg_validators.ValidBatchDevice(), + default=[], + help='Devices to provision OSDs journal volumes', + ) + parser.add_argument( + '--auto', + action='store_true', + help=('deploy multi-device OSDs if rotational and non-rotational drives ' + 'are passed in DEVICES'), + default=True + ) + parser.add_argument( + '--no-auto', + action='store_false', + dest='auto', + help=('deploy standalone OSDs if rotational and non-rotational drives ' + 'are passed in DEVICES'), + ) + parser.add_argument( + '--bluestore', + action='store_true', + help='bluestore objectstore (default)', + ) + parser.add_argument( + '--filestore', + action='store_true', + help='filestore objectstore', + ) + parser.add_argument( + '--report', + action='store_true', + help='Only report on OSD that would be created and exit', + ) + parser.add_argument( + '--yes', + action='store_true', + help='Avoid prompting for confirmation when provisioning', + ) + parser.add_argument( + '--format', + help='output format, defaults to "pretty"', + default='pretty', + choices=['json', 'json-pretty', 'pretty'], + ) + parser.add_argument( + '--dmcrypt', + action='store_true', + help='Enable device encryption via dm-crypt', + ) + parser.add_argument( + '--crush-device-class', + dest='crush_device_class', + help='Crush device class to assign this OSD to', + default="" + ) + parser.add_argument( + '--no-systemd', + dest='no_systemd', + action='store_true', + help='Skip creating and enabling systemd units and starting OSD services', + ) + parser.add_argument( + '--osds-per-device', + type=int, + default=1, + help='Provision more than 1 (the default) OSD per device', + ) + parser.add_argument( + '--data-slots', + type=int, + help=('Provision more than 1 (the default) OSD slot per device' + ' if more slots then osds-per-device are specified, slots' + 'will stay unoccupied'), + ) + parser.add_argument( + '--block-db-size', + type=disk.Size.parse, + help='Set (or override) the "bluestore_block_db_size" value, in bytes' + ) + parser.add_argument( + '--block-db-slots', + type=int, + help='Provision slots on DB device, can remain unoccupied' + ) + parser.add_argument( + '--block-wal-size', + type=disk.Size.parse, + help='Set (or override) the "bluestore_block_wal_size" value, in bytes' + ) + parser.add_argument( + '--block-wal-slots', + type=int, + help='Provision slots on WAL device, can remain unoccupied' + ) + def journal_size_in_mb_hack(size): + # TODO give user time to adjust, then remove this + if size and size[-1].isdigit(): + mlogger.warning('DEPRECATION NOTICE') + mlogger.warning('--journal-size as integer is parsed as megabytes') + mlogger.warning('A future release will parse integers as bytes') + mlogger.warning('Add a "M" to explicitly pass a megabyte size') + size += 'M' + return disk.Size.parse(size) + parser.add_argument( + '--journal-size', + type=journal_size_in_mb_hack, + help='Override the "osd_journal_size" value, in megabytes' + ) + parser.add_argument( + '--journal-slots', + type=int, + help='Provision slots on journal device, can remain unoccupied' + ) + parser.add_argument( + '--prepare', + action='store_true', + help='Only prepare all OSDs, do not activate', + ) + parser.add_argument( + '--osd-ids', + nargs='*', + default=[], + help='Reuse existing OSD ids', + type=arg_validators.valid_osd_id + ) + self.args = parser.parse_args(argv) + self.parser = parser + for dev_list in ['', 'db_', 'wal_', 'journal_']: + setattr(self, '{}usable'.format(dev_list), []) + + def report(self, plan): + report = self._create_report(plan) + print(report) + + def _create_report(self, plan): + if self.args.format == 'pretty': + report = '' + report += templates.total_osds.format(total_osds=len(plan)) + + report += templates.osd_component_titles + for osd in plan: + report += templates.osd_header + report += osd.report() + return report + else: + json_report = [] + for osd in plan: + json_report.append(osd.report_json()) + if self.args.format == 'json': + return json.dumps(json_report) + elif self.args.format == 'json-pretty': + return json.dumps(json_report, indent=4, + sort_keys=True) + + def _check_slot_args(self): + ''' + checking if -slots args are consistent with other arguments + ''' + if self.args.data_slots and self.args.osds_per_device: + if self.args.data_slots < self.args.osds_per_device: + raise ValueError('data_slots is smaller then osds_per_device') + + def _sort_rotational_disks(self): + ''' + Helper for legacy auto behaviour. + Sorts drives into rotating and non-rotating, the latter being used for + db or journal. + ''' + mlogger.warning('DEPRECATION NOTICE') + mlogger.warning('You are using the legacy automatic disk sorting behavior') + mlogger.warning('The Pacific release will change the default to --no-auto') + rotating = [] + ssd = [] + for d in self.args.devices: + rotating.append(d) if d.rotational else ssd.append(d) + if ssd and not rotating: + # no need for additional sorting, we'll only deploy standalone on ssds + return + self.args.devices = rotating + if self.args.filestore: + self.args.journal_devices = ssd + else: + self.args.db_devices = ssd + + @decorators.needs_root + def main(self): + if not self.args.devices: + return self.parser.print_help() + + # Default to bluestore here since defaulting it in add_argument may + # cause both to be True + if not self.args.bluestore and not self.args.filestore: + self.args.bluestore = True + + if (self.args.auto and not self.args.db_devices and not + self.args.wal_devices and not self.args.journal_devices): + self._sort_rotational_disks() + + self._check_slot_args() + + ensure_disjoint_device_lists(self.args.devices, + self.args.db_devices, + self.args.wal_devices, + self.args.journal_devices) + + plan = self.get_plan(self.args) + + if self.args.report: + self.report(plan) + return 0 + + if not self.args.yes: + self.report(plan) + terminal.info('The above OSDs would be created if the operation continues') + if not prompt_bool('do you want to proceed? (yes/no)'): + terminal.error('aborting OSD provisioning') + raise SystemExit(0) + + self._execute(plan) + + def _execute(self, plan): + defaults = common.get_default_args() + global_args = [ + 'bluestore', + 'filestore', + 'dmcrypt', + 'crush_device_class', + 'no_systemd', + ] + defaults.update({arg: getattr(self.args, arg) for arg in global_args}) + for osd in plan: + args = osd.get_args(defaults) + if self.args.prepare: + p = Prepare([]) + p.safe_prepare(argparse.Namespace(**args)) + else: + c = Create([]) + c.create(argparse.Namespace(**args)) + + + def get_plan(self, args): + if args.bluestore: + plan = self.get_deployment_layout(args, args.devices, args.db_devices, + args.wal_devices) + elif args.filestore: + plan = self.get_deployment_layout(args, args.devices, args.journal_devices) + return plan + + def get_deployment_layout(self, args, devices, fast_devices=[], + very_fast_devices=[]): + ''' + The methods here are mostly just organization, error reporting and + setting up of (default) args. The heavy lifting code for the deployment + layout can be found in the static get_*_osds and get_*_fast_allocs + functions. + ''' + plan = [] + phys_devs, lvm_devs = separate_devices_from_lvs(devices) + mlogger.debug(('passed data devices: {} physical,' + ' {} LVM').format(len(phys_devs), len(lvm_devs))) + + plan.extend(get_physical_osds(phys_devs, args)) + + plan.extend(get_lvm_osds(lvm_devs, args)) + + num_osds = len(plan) + if num_osds == 0: + mlogger.info('All data devices are unavailable') + return plan + requested_osds = args.osds_per_device * len(phys_devs) + len(lvm_devs) + + fast_type = 'block_db' if args.bluestore else 'journal' + fast_allocations = self.fast_allocations(fast_devices, + requested_osds, + num_osds, + fast_type) + if fast_devices and not fast_allocations: + mlogger.info('{} fast devices were passed, but none are available'.format(len(fast_devices))) + return [] + if fast_devices and not len(fast_allocations) == num_osds: + mlogger.error('{} fast allocations != {} num_osds'.format( + len(fast_allocations), num_osds)) + exit(1) + + very_fast_allocations = self.fast_allocations(very_fast_devices, + requested_osds, + num_osds, + 'block_wal') + if very_fast_devices and not very_fast_allocations: + mlogger.info('{} very fast devices were passed, but none are available'.format(len(very_fast_devices))) + return [] + if very_fast_devices and not len(very_fast_allocations) == num_osds: + mlogger.error('{} very fast allocations != {} num_osds'.format( + len(very_fast_allocations), num_osds)) + exit(1) + + for osd in plan: + if fast_devices: + osd.add_fast_device(*fast_allocations.pop(), + type_=fast_type) + if very_fast_devices and args.bluestore: + osd.add_very_fast_device(*very_fast_allocations.pop()) + return plan + + def fast_allocations(self, devices, requested_osds, new_osds, type_): + ret = [] + if not devices: + return ret + phys_devs, lvm_devs = separate_devices_from_lvs(devices) + mlogger.debug(('passed {} devices: {} physical,' + ' {} LVM').format(type_, len(phys_devs), len(lvm_devs))) + + ret.extend(get_lvm_fast_allocs(lvm_devs)) + + # fill up uneven distributions across fast devices: 5 osds and 2 fast + # devices? create 3 slots on each device rather then deploying + # heterogeneous osds + slot_divider = max(1, len(phys_devs)) + if (requested_osds - len(lvm_devs)) % slot_divider: + fast_slots_per_device = int((requested_osds - len(lvm_devs)) / slot_divider) + 1 + else: + fast_slots_per_device = int((requested_osds - len(lvm_devs)) / slot_divider) + + + ret.extend(get_physical_fast_allocs(phys_devs, + type_, + fast_slots_per_device, + new_osds, + self.args)) + return ret + + class OSD(object): + ''' + This class simply stores info about to-be-deployed OSDs and provides an + easy way to retrieve the necessary create arguments. + ''' + VolSpec = namedtuple('VolSpec', + ['path', + 'rel_size', + 'abs_size', + 'slots', + 'type_']) + + def __init__(self, + data_path, + rel_size, + abs_size, + slots, + id_, + encryption, + symlink=None): + self.id_ = id_ + self.data = self.VolSpec(path=data_path, + rel_size=rel_size, + abs_size=abs_size, + slots=slots, + type_='data') + self.fast = None + self.very_fast = None + self.encryption = encryption + self.symlink = symlink + + def add_fast_device(self, path, rel_size, abs_size, slots, type_): + self.fast = self.VolSpec(path=path, + rel_size=rel_size, + abs_size=abs_size, + slots=slots, + type_=type_) + + def add_very_fast_device(self, path, rel_size, abs_size, slots): + self.very_fast = self.VolSpec(path=path, + rel_size=rel_size, + abs_size=abs_size, + slots=slots, + type_='block_wal') + + def _get_osd_plan(self): + plan = { + 'data': self.data.path, + 'data_size': self.data.abs_size, + 'encryption': self.encryption, + } + if self.fast: + type_ = self.fast.type_.replace('.', '_') + plan.update( + { + type_: self.fast.path, + '{}_size'.format(type_): self.fast.abs_size, + }) + if self.very_fast: + plan.update( + { + 'block_wal': self.very_fast.path, + 'block_wal_size': self.very_fast.abs_size, + }) + if self.id_: + plan.update({'osd_id': self.id_}) + return plan + + def get_args(self, defaults): + my_defaults = defaults.copy() + my_defaults.update(self._get_osd_plan()) + return my_defaults + + def report(self): + report = '' + if self.id_: + report += templates.osd_reused_id.format( + id_=self.id_) + if self.encryption: + report += templates.osd_encryption.format( + enc=self.encryption) + path = self.data.path + if self.symlink: + path = f'{self.symlink} -> {self.data.path}' + report += templates.osd_component.format( + _type=self.data.type_, + path=path, + size=self.data.abs_size, + percent=self.data.rel_size) + if self.fast: + report += templates.osd_component.format( + _type=self.fast.type_, + path=self.fast.path, + size=self.fast.abs_size, + percent=self.fast.rel_size) + if self.very_fast: + report += templates.osd_component.format( + _type=self.very_fast.type_, + path=self.very_fast.path, + size=self.very_fast.abs_size, + percent=self.very_fast.rel_size) + return report + + def report_json(self): + # cast all values to string so that the report can be dumped in to + # json.dumps + return {k: str(v) for k, v in self._get_osd_plan().items()} |