Merge pull request #18838 from shinobu-x/wip-21973-luminous

luminous: [test] UpdateFeatures RPC message should be included in test_notify.py

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #18840 from shinobu-x/wip-21970-luminous

luminous: [journal] tags are not being expired if no other clients are registered

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #18949 from jcsp/wip-luminous-rgw-config-desc

luminous: options.cc: document rgw config options

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #19051 from shinobu-x/wip-22185-luminous

luminous: rbd: abort in listing mapped nbd devices when running in a container

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #20062 from cbodley/wip-luminous-rgw-bilog-trim

luminous: rgw multisite: automated trimming for bucket index logs

Reviewed-by: Orit Wasserman <owasserm@redhat.com>

Merge pull request #19848 from cbodley/wip-luminous-boost-166

luminous: backport beast frontend and boost 1.66 update

Merge pull request #19500 from shinobu-x/wip-22401-luminous

luminous: rgw: make HTTP dechunking compatible with Amazon S3

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #17816 from smithfarm/wip-hadoop-s3a

luminous: tests: qa/rgw: hadoop-s3a suite targets centos_latest

Reviewed-by: Casey Bodley <cbodley@redhat.com>
Reviewed-by: Yuri Weinstein <yweins@redhat.com>

Merge pull request #18972 from linuxbox2/luminous-22084

luminous: rgw: Fix swift object expiry not deleting objects

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #19086 from shinobu-x/wip-22210-luminous

luminous: radosgw-admin zonegroup get and zone get should return defaults when there is no realm

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #19088 from pdvian/wip-22177-luminous

luminous: rgw: check going_down() when lifecycle processing

Reviewed-by: Casey Bodley <cbodley@redhat.com>
Reviewed-by: Amit Kumar <amitkuma@redhat.com>

Merge pull request #19636 from shinobu-x/wip-22506-luminous

luminous: rgw usage trim only trims a few entries

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #19784 from linuxbox2/luminous-20201

luminous: RGW: S3 POST policy should not require Content-Type

Reviewed-by: Casey Bodley <cbodley@redhat.com>
Reviewed-by: Abhishek Lekshmanan <abhishek.lekshmanan@gmail.com>

Merge pull request #19799 from smithfarm/wip-22581-luminous

luminous: rgw: multisite: 'radosgw-admin sync error list' contains temporary EBUSY errors

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #19867 from smithfarm/wip-22591-luminous

luminous: rgw: refuses upload when Content-Type missing from POST policy

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #20051 from ceph/revert-19538-wip-22434-luminous

Revert "luminous: rgw: user stats increased after bucket reshard"

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #20056 from linuxbox2/luminous-rgw-latchfix

luminous: rgw_file: alternate fix deadlock on lru eviction

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #20073 from smithfarm/wip-22767-luminous

luminous: rgw: librgw: fix shutdown error with resources uncleaned

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #20075 from smithfarm/wip-22773-luminous

luminous: rgw: file deadlock on lru evicting

Reviewed-by: Matt Benjamin <mbenjami@redhat.com>

Merge pull request #20107 from dreamhost/wip-22792

luminous: rgw: When a system object is created exclusively, do not distribute the

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #20136 from dillaman/wip-22815-luminous

luminous: qa/workunits/rbd: switch devstack to pike release

Reviewed-by: Mykola Golub <mgolub@suse.com>

qa/workunits/rbd: switch devstack to pike release

Signed-off-by: Jason Dillaman <dillaman@redhat.com>
(cherry picked from commit 7ff1e0bc2193acd009a74db81cf8e8cc8639ad12)

Merge pull request #19105 from batrick/i22228

luminous: client: anchor Inode while trimming caps

Reviewed-by: Yan, Zheng <zyan@redhat.com>
Reviewed-by: Yuri Weinstein <yweins@redhat.com>

Merge pull request #19360 from ukernel/luminous-22219

luminous: mds: ignore export pin for unlinked directory

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #19646 from pdvian/wip-22503-luminous

luminous: mds: properly eval locks after importing inode

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #20055 from dzafman/wip-22724

luminous: miscounting degraded objects and PG stuck in recovery_unfound

Reviewed-by: Josh Durgin <jdurgin@redhat.com>

Merge branch 'luminous' into wip-22724

Merge pull request #20081 from dzafman/wip-recovering-luminous

luminous: osd/ReplicatedPG.cc: recover_replicas: object added to missing set for backfill, but is not in recovering, error!

Reviewed-by: Josh Durgin <jdurgin@redhat.com>

Merge pull request #18008 from ukernel/luminous-21584

luminous: mds: fix CDir::log_mark_dirty()

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Jos Collin <jcollin@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #18782 from ukernel/luminous-21985

luminous: mds: fix MDS_FEATURE_INCOMPAT_FILE_LAYOUT_V2 definition

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #18783 from ukernel/luminous-21975

luminous: mds: trim 'N' log segments according to how many log segments are there

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #18831 from shinobu-x/wip-22076-luminous

luminous: cephfs: "ceph tell mds" commands result in "File exists" errors on client admin socket

Reviewed-by: Jos Collin <jcollin@redhat.com>
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #18835 from shinobu-x/wip-22074-luminous

luminous: mds: don't check gid when none specified in auth caps

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #18869 from shinobu-x/wip-21952-luminous

luminous: mds: no assertion on inode being purging in find_ino_peers()

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #18871 from shinobu-x/wip-21947-luminous

luminous: mds: preserve order of requests during recovery of multimds cluster

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #18912 from shinobu-x/wip-22077-luminous

luminous: src/mds/MDCache.cc: 7421: FAILED assert(CInode::count() == inode_map.size() + snap_inode_map.size())

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #19055 from shinobu-x/wip-22192-luminous

luminous: MDSMonitor: monitor gives constant "is now active in filesystem cephfs as rank" cluster log info messages

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Yan, Zheng <zyan@redhat.com>

Merge pull request #19326 from ukernel/luminous-22263

luminous: mds: handle client reconnect gather race

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #19480 from smithfarm/wip-22407-luminous

luminous: cephfs: client: implement delegation support in userland cephfs

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>

Merge pull request #19775 from ukernel/luminous-19578

luminous: mds: track dirty dentries in separate list

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #19827 from pdvian/wip-22587-luminous

luminous: mds: reduce debugging level for balancer messages

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Amit Kumar <amitkuma@redhat.com>

Merge pull request #20082 from smithfarm/wip-21525-luminous

luminous: cephfs: client: dual client segfault with racing ceph_shutdown

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>

Merge pull request #20085 from pdvian/wip-22763-luminous

luminous: cephfs-journal-tool: add "set pool_id" option

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #20086 from pdvian/wip-22765-luminous

luminous: client: avoid recursive lock in ll_get_vino

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

qa/rgw: hadoop suite targets centos_latest

Signed-off-by: Casey Bodley <cbodley@redhat.com>
(cherry picked from commit a97d5bea42a40909aa4e9672505b9ad994cec184)

Conflicts:
    qa/suites/rgw/hadoop-s3a/s3a-hadoop.yaml (changes already backported
        via 6b5f212b9170dbd99779527b41be233b9d40b5e4)

qa/rgw: use overrides for hadoop version

instead of having two separate tasks, add v27.yaml and v28.yaml with
overrides for hadoop_version

Signed-off-by: Casey Bodley <cbodley@redhat.com>
(cherry picked from commit 4a953a16d7b80d343899de39a8cf55dc761847de)

Merge pull request #18019 from ukernel/luminous-21091

luminous: mds: fix StrayManager::truncate()

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: John Spray <john.spray@redhat.com>

Merge pull request #19485 from dillaman/wip-21788-luminous

luminous: librbd: refresh image after applying new/removing old metadata

Reviewed-by: Mykola Golub <mgolub@suse.com>

Merge pull request #18834 from shinobu-x/wip-22073-luminous

luminous: [api] compare-and-write methods not properly advertised

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #19058 from shinobu-x/wip-22190-luminous

luminous: class rbd.Image discard----OSError: [errno 2147483648] error discarding region

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #19066 from pdvian/wip-22172-luminous

luminous: [rbd-nbd] Fedora does not register resize events

Reviewed-by: Jason Dillaman <dillaman@redhat.com>
Reviewed-by: Amit Kumar <amitkuma@redhat.com>

Merge pull request #19107 from pdvian/wip-22208-luminous

luminous: rbd: disk usage on empty pool no longer returns an error message

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #19305 from pdvian/wip-21700-luminous

luminous: rbd-mirror: Allow a different data-pool to be used on the secondary cluster

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #19447 from smithfarm/wip-21646-luminous

luminous: librbd: Image-meta should be dynamically refreshed

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #19479 from smithfarm/wip-22376-luminous

luminous: rbd: Python RBD metadata_get does not work

Reviewed-by: Jason Dillaman <dillaman@redhat.com>
Reviewed-by: Amit Kumar <amitkuma@redhat.com>

Merge pull request #19484 from dillaman/wip-21644-luminous

luminous: rbd-mirror: sync image metadata when transfering remote image

Reviewed-by: Mykola Golub <mgolub@mirantis.com>

Merge pull request #19503 from shinobu-x/wip-22395-luminous

luminous: librbd: cannot clone all image-metas if we have more than 64 key/value pairs

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #19504 from shinobu-x/wip-22393-luminous

luminous: librbd: cannot copy all image-metas if we have more than 64 key/value pairs

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #19554 from shinobu-x/wip-22454-luminous

 luminous: cluster resource agent ocf:ceph:rbd - wrong permissions

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #19625 from shinobu-x/wip-22497-luminous

 luminous: [rbd-mirror] new pools might not be detected

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #19800 from smithfarm/wip-21641-luminous

luminous: rbd ls -l crashes with SIGABRT

Reviewed-by: Mykola Golub <mgolub@mirantis.com>

Merge pull request #19802 from smithfarm/wip-21690-luminous

luminous: tests: rbd_mirror_helpers.sh request_resync_image function saves image id to wrong variable

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #19853 from pdvian/wip-22593-luminous

luminous: rbd: librbd: filter out potential race with image rename

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Revert "luminous: rgw: user stats increased after bucket reshard"

Signed-off-by: Orit Wasserman <owasserm@redhat.com>

When a system object is created exclusively, do not distribute the
cache information to the other rados gateways and instead let them
lazily retrieve the information when they need it.

Signed-off-by: J. Eric Ivancich <ivancich@redhat.com>
(cherry picked from commit 647ce3387312fc683660c1f3c7571c577379be1c)
Signed-off-by: Robin H. Johnson <robin.johnson@dreamhost.com>

Merge pull request #18722 from shinobu-x/wip-luminous-21946

luminous: mds: set PRIO_USEFUL on num_sessions counter

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>

Merge pull request #18730 from shinobu-x/wip-luminous-22030

luminous: List of filesystems does not get refreshed after a filesystem deletion

Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #18754 from shinobu-x/wip-21636-luminous

luminous: ceph-monstore-tool --readable mode doesn't understand FSMap, MgrMap

Reviewed-by: Josh Durgin <jdurgin@redhat.com>

Merge pull request #20053 from dillaman/wip-22740-luminous

luminous: qa/suites/upgrade: disable broken kraken upgrade cls_rbd test

Reviewed-By: Yuri Weinstein <yuriw@redhat.com>

Merge pull request #18564 from kmroz/wip-21688-luminous

luminous: Possible deadlock in 'list_children' when refresh is required

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #18698 from shinobu-x/wip-luminous-21808

luminous: rbd: fix crash during map

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #19038 from pdvian/wip-22183-luminous

luminous: rgw: set sync_from_all as true when no value is seen

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #19446 from smithfarm/wip-22404-luminous

luminous: doc: crush_ruleset is invalid command in luminous

Reviewed-By: Abhishek Lekshmanan <abhishek@suse.com>

Merge pull request #18747 from shinobu-x/wip-21794-luminous

luminous: backoff causes out of order op

Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: Yuri Weinstein <yweinste@redhat.com>

Merge pull request #20087 from dzafman/wip-misc-luminous

qa/standalone: Minor fixes for ceph_objectstore_tool.py test

Reviewed-by: Kefu Chai <kchai@redhat.com>

test: Cleanup test-erasure-eio.sh code

Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit 69b5fc54feb59f8b0a26a3ca3e925980c91b5b75)

test: Add replicated recovery/backfill test

Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit c2572bee3c6256419e0b265e9e2829e7f3afb76d)

osd: Better recovery/backfill restart for mark_unfound_lost delete/revert

Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit 689bff354a6bfa1cf47b0d0a04fa9f1b2ef68f75)

osd, mon: Add new pg states recovery_unfound and backfill_unfound

Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit 7f8b0ce9e681f727d8217e3ed74a1a3355f364f3)

Conflicts:
src/osd/PG.h (trivial)
src/osd/osd_types.h (trivial)

osd: Add new UnfoundBackfill and UnfoundRecovery pg transitions

Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit bb2bcb95f51abc206e005e44ef383ee45b8f2209)

Conflicts:
src/osd/PG.cc (trivial, no MBackfillReserve::CANCEL)
src/osd/PG.h (trivial)

fixup! osd: Add new UnfoundBackfill and UnfoundRecovery pg transitions

test: Fix ceph-objectstore-tool usage check

Caused by: c7b7a1f04f78fa62890c567d0ca53874c8d75eb7

Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit 0364ae104afc15effadfe968b5f60221704d933c)

test: ceph_objectstore_tool.py bad --filestore option to vstart

Luminous doesn't have --filestore option

Caused by: fd7d53bacb10d5b2094f1f579c96899d8a0a4b4f

Signed-off-by: David Zafman <dzafman@redhat.com>

client: avoid recursive lock in ll_get_vino

Fixes: http://tracker.ceph.com/issues/22629
Signed-off-by: dongdong tao <tdd21151186@gmail.com>
(cherry picked from commit fdfbe40c3f16c14dc9c3352f3eea66d34138ee9d)

cephfs-journal-tool: add usage help info for header set <field>

Signed-off-by: dongdong tao <tdd21151186@gmail.com>
(cherry picked from commit 2b06bff950a78b15aca083b20eaa812b4214e8d4)

mds: add error log info

Signed-off-by: dongdong tao <tdd21151186@gmail.com>
(cherry picked from commit a7c2ecd6391d22c5312a9e568ff742cff252749e)

cephfs-journal-tool: add "set pool_id" option

Fixes: http://tracker.ceph.com/issues/22631
Signed-off-by: dongdong tao <tdd21151186@gmail.com>
(cherry picked from commit b3a5e313117028b23133f3e0d72e0b6b7159a178)

Merge pull request #18742 from shinobu-x/wip-21924-luminous

luminous: ceph_test_objectstore fails ObjectStore/StoreTest.Synthetic/1 (filestore) buffer content mismatch

Reviewed-by: Josh Durgin <jdurgin@redhat.com>

Merge pull request #18744 from shinobu-x/wip-21922-luminous

luminous: Objecter::C_ObjectOperation_sparse_read throws/catches exceptions on -ENOENT

Reviewed-by: Josh Durgin <jdurgin@redhat.com>

Merge pull request #18745 from shinobu-x/wip-21921-luminous

luminous: Objecter::_send_op unnecessarily constructs costly hobject_t

Reviewed-by: Josh Durgin <jdurgin@redhat.com>

test: make the LibCephFS.ShutdownRacer test even more thrashy

Have each thread do the startup and shutdown in a loop for a specified
number of times.

Tracker: http://tracker.ceph.com/issues/21512
Signed-off-by: Jeff Layton <jlayton@redhat.com>
(cherry picked from commit f877e365a42f8f34af6fc3382593ef09101a50d0)

lockdep: free_ids and lock_ref hashes must be truly global

It's possible for the teardown of g_lockdep_ceph_ctx to occur, followed
by a new context being registered as the lockdep context. When that
occurs, we can end up reusing lock id's that were previously handed out
to consumers. We need for those IDs to be persistent across lockdep
enablement and disablement.

Make both the free_ids table, and the lock_refs map persistent across
lockdep_unregister_ceph_context and lockdep_register_ceph_context cycles.
Entries in those tables will only be deleted by the destruction of the
associated mutex.

When lockdep_unregister is called, do the refcounting like we normally
would, but only clear out the state when the lockid is registered
in the lock_names hash.

Finally, we do still need to handle the case where g_lockdep has gone
false even when there are outstanding references after the decrement.
Only log the message if that's not the case.

With this, we can deal with the case of multiple clients enabling and
disabling lockdep in an unsynchronized way.

Tracker: http://tracker.ceph.com/issues/21512
Signed-off-by: Jeff Layton <jlayton@redhat.com>
(cherry picked from commit 82fe4e94bcb706434f9215bc8405ff60770cf14f)

common: add a clear_g_str_vec() function to clear g_str_vec

Prefix str_vec and str_vec_lock with "g_" to make it clear that they are
truly global values. Add a new clear_g_str_vec function to allow it to
be explicitly cleaned out by callers that need that functionality
(mostly testcase for now).

Tracker: http://tracker.ceph.com/issues/21512
Signed-off-by: Jeff Layton <jlayton@redhat.com>
(cherry picked from commit f72ace0a5666db49bb390ff4ec6808e6e55a779f)

Merge pull request #18746 from shinobu-x/wip-21916-luminous

luminous: msg/async/AsyncConnection.cc: 1835: FAILED assert(state == STATE_CLOSED)

Reviewed-by: Josh Durgin <jdurgin@redhat.com>

common: make it safe to call env_to_vec multiple times

After it has been called once and we have outstanding CephContexts with
pointers into str_vec, we can't call get_str_vec on it again.

Add a static local mutex to protect access to str_vec.

Tracker: http://tracker.ceph.com/issues/21512
Signed-off-by: Jeff Layton <jlayton@redhat.com>
(cherry picked from commit 90e898de5f4b8d22f1a6d0e2aedf9e8c50cf72d5)

lockdep: fix Mutex tests to disable lockdep properly

...and make g_lockdep a bool.

Signed-off-by: Jeff Layton <jlayton@redhat.com>
(cherry picked from commit 0cd0bd778a6149d3e2fe657060e223be3675aed7)

client: test shutdown race

Spawn threads that bring up a bunch of ceph_mounts with individual
CephContext objects, and then tear them down in parallel.

Tracker: http://tracker.ceph.com/issues/20988
Signed-off-by: Jeff Layton <jlayton@redhat.com>
(cherry picked from commit 8252f3168a1bc90add7f7515c549e9eecffa54b7)

client: fix signed/unsigned comparison compiler warning

The build says:

src/client/Client.cc: In member function ‘void Client::trim_caps(MetaSession*, int)’:
src/client/Client.cc:4121:22: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
   if (s->caps.size() > max)
       ~~~~~~~~~~~~~~~^~~~~

Signed-off-by: Jeff Layton <jlayton@redhat.com>
(cherry picked from commit e057b6770b5545f327990cbaf17e7a391e3e8a50)

lockdep: fix races with concurrent lockdep teardown

If the cct is unregistered while other threads are flogging mutexes,
then we can hit all sorts of bugs. Ensure that we handle that
situation sanely, by checking that g_lockdep is still set after
we take the lockdep_mutex.

Also, remove an assertion from lockdep_unregister, and just turn it into
an immediate return. It's possible to have a call to
lockdep_unregister_ceph_context, and then a call to
lockdep_register_ceph_context while a mutex is being held by another
task.

In that case, it's possible the lock does not exist in the map
when we go to unregister it. That's not a bug though, just a natural
consequence of that series of actions.

Tracker: http://tracker.ceph.com/issues/20988
Signed-off-by: Jeff Layton <jlayton@redhat.com>
(cherry picked from commit 75f41a95782a7ee83a243d91963e8d591402f8a6)

lockdep: don't take lockdep_mutex twice for new lock registrations

We can do it under the same mutex, which should be more efficient.

Signed-off-by: Jeff Layton <jlayton@redhat.com>
(cherry picked from commit 01863bb6fb62ea89aa3e21e43bf4dc4f3da9cfcb)

Merge pull request #18749 from shinobu-x/wip-21785-luminous

luminous: OSDMap cache assert on shutdown

Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Greg Farnum <gfarnum@redhat.com>