rgw: fix not initialized pointer which cause rgw crash with ec data pool

In RGWPutObjProcessor_Atomic::complete_writing_data()
with pending_data_bl.length() > 0 and next_part_ofs==data_ofs
not initialized void *handle leads to invalid pointer librados::AioCompletion::pc
which leads to rgw crash.

Fixes: http://tracker.ceph.com/issues/20542
Signed-off-by: Aleksei Gutikov <aleksey.gutikov@synesis.ru>
(cherry picked from commit 3e938dd9fe681fa8652dc4b0ec1dc56781d884c0)

Conflicts:
src/rgw/rgw_rados.cc (trivial resolution)

rgw: don't do unneccesary write if buffer with zero length

Don't do unneccesary write if buffer with zero length,
or there will be one more shadow stripe rados object with size 0

Signed-off-by: fang yuxiang <fang.yuxiang@eisoo.com>
(cherry picked from commit 4e21f7ea2167be7793273f31faeb0aa95f0fa30a)

Merge pull request #15448 from smithfarm/wip-19615-jewel

jewel: rgw: multisite: bucket zonegroup redirect not working

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

Merge pull request #15197 from smithfarm/wip-20007-jewel

jewel: rgw: remove unnecessary 'error in read_id for object name: default'

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

Merge pull request #15051 from asheplyakov/19910-bp-jewel

jewel: build/ops: extended, customizable systemd ceph-disk timeout

Reviewed-by: David Disseldorp <ddiss@suse.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #15428 from smithfarm/wip-20162-jewel

jewel: build/ops: deb: Fix logrotate packaging

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

Merge pull request #15447 from smithfarm/wip-19613-jewel

jewel: rgw: multisite: rest api fails to decode large period on "period commit"

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

Merge pull request #15196 from smithfarm/wip-20006-jewel

jewel: build/ops: rpm: fix python-Sphinx package name for SUSE

Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: Ken Dreyer <kdreyer@redhat.com>
Reviewed-by: Abhishek Lekshmanan <abhishek@suse.com>

qa/workunits/rados/test/test-upgrade-v11.0.0: skip Quota test

This makes a pool go full which confounds the wait for healthy check
during restarts.

Signed-off-by: Sage Weil <sage@redhat.com>

qa/workunits/rados/test-upgrade-v11.0.0: skip api_io *CmpExt*

Signed-off-by: Sage Weil <sage@redhat.com>

qa/workunits/rados/test-upgrade-11.0.0: skip RoundTrip3

This does the crc op, which isn't in jewel.

Signed-off-by: Sage Weil <sage@redhat.com>

Merge pull request #16343 from dillaman/wip-20630-jewel

jewel: tests: qa/tasks: rbd-mirror daemon not properly run in foreground mode

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

Merge pull request #13507 from shinobu-x/wip-18468-jewel

jewel: mon: Disallow enabling 'hashpspool' option to a pool without some kind of --i-understand-this-will-remap-all-pgs flag

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

Merge pull request #16289 from cbodley/wip-20513

jewel: rgw: cls: ceph::timespan tag_timeout wrong units

qa/tasks: rbd-mirror daemon not properly run in foreground mode

Fixes: http://tracker.ceph.com/issues/20630
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
(cherry picked from commit 4fa1918717b25a5ffafbf649eedcfe7d5ab829c2)

10.2.9

Merge pull request #16282 from smithfarm/wip-20599-jewel

jewel: cephfs: Damaged MDS with 10.2.8

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

rgw: allow larger payload for period commit

testing with 3 zonegroups and 3 zones each, the period json grew larger
than 4k and caused decode failures on period commit

updated to use the new config variable rgw_max_put_param_size

Fixes: http://tracker.ceph.com/issues/19505
Signed-off-by: Casey Bodley <cbodley@redhat.com>
(cherry picked from commit 7f2871fe59d933b03f37fde40f1781b2320d0d50)

Conflicts:
    src/rgw/rgw_rest_realm.cc - in RGWOp_Period_Post::execute(), set max_size
        explicitly instead of backporting rgw_max_put_param_size

rgw: cls: ceph::timespan tag_timeout wrong units

In rgw_dir_suggest(), the ceph::timespan tag_timeout is intended
to be a value in seconds, but has been taken as something much
smaller (millis?).  The incorrect time scale likely induces a race
condition with object deletes.

Fixes: http://tracker.ceph.com/issues/20380
Signed-off-by: Matt Benjamin <mbenjamin@redhat.com>
(cherry picked from commit e6763ebd6124c1b4736c5c7850077df7644bc8f5)

Revert "osdc/Journaler: make header write_pos align to boundary of flushed entry"

This reverts commit 2e299b50de4a297fee2aec21290632336d239857.

Signed-off-by: Nathan Cutler <ncutler@suse.com>

Revert "osdc/Journaler: avoid executing on_safe contexts prematurely"

This reverts commit 06cf9f365033f7913051bdf4060f0bc6fc0444d7.

Signed-off-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #16273 from smithfarm/wip-jewel-pending-release-notes

jewel: doc: clarify status of jewel PendingReleaseNotes

doc: zero PendingReleaseNotes in preparation for v10.2.9

Signed-off-by: Nathan Cutler <ncutler@suse.com>

doc: clarify status of jewel PendingReleaseNotes

Status as of 10.2.8 release

Signed-off-by: Nathan Cutler <ncutler@suse.com>

10.2.8

Merge pull request #14710 from smithfarm/wip-start-race

tests: rados: sleep before ceph tell osd.0 flush_pg_stats after restart

Reviewed-by: Sage Weil <sage@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: David Zafman <dzafman@redhat.com>

Merge pull request #16089 from ceph/wip_fix_point_jewel

qa/Fixed upgrade sequence to 10.2.0 -> 10.2.7 -> latest -x (10.2.8)

Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #16088 from smithfarm/wip-fix-client-upgrade-centos

tests: run upgrade/client-upgrade on latest CentOS 7.3

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

Fixed upgrade sequence to 10.2.0 -> 10.2.7 -> latest -x (10.2.8)

Signed-off-by: Yuri Weinstein <yweinste@redhat.com>

tests: run upgrade/client-upgrade on latest CentOS 7.3

Before this patch, all centos jobs were failing because there are no longer any
CentOS 7.2 machines in Sepia.

Signed-off-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #15504 from Vicente-Cheng/wip-20151-jewel

jewel: ceph-disk: do not setup_statedir on trigger

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

Merge pull request #15904 from smithfarm/wip-20413-jewel

jewel: tests: upgrade:hammer-x/stress-split-erasure-code-x86_64 fails in 10.2.8 integration testing

Reviewed-by: Brad Hubbard <bhubbard@redhat.com>

Merge pull request #14930 from smithfarm/wip-19829-jewel

jewel: tests: New upgrade test for #14930

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

Merge pull request #14392 from asheplyakov/19508-jewel

jewel: osd: pg_pool_t::encode(): be compatible with Hammer <= 0.94.6

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

tests: upgrade/hammer-x/v0-94-6-mon-overload: tweak packages list

Include some hammer dependencies that aren't in the jewel default packages
list, and exclude some java packages that may not be in the hammer repo and are
not needed for the upgrade test in any case.

N.B.: This cannot be cherry-picked from master because upgrade/hammer-x was
dropped in master.

Signed-off-by: Nathan Cutler <ncutler@suse.com>

tests: upgrade/hammer-x: new v0-94-6-mon-overload subsuite

This is not a cherry-pick from master because direct upgrades
from hammer to kraken+ are not supported.

Fixes: http://tracker.ceph.com/issues/19829
References: http://tracker.ceph.com/issues/19508
Signed-off-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #15936 from batrick/i20412

qa: enable quotas for pre-luminous quota tests

Reviewed-by: John Spray <john.spray@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

qa: enable quotas for pre-luminous quota tests

This cannot be cherry-picked from master because the config option is removed
since 0f250a889dba2100d3afcea0a18e4f6a8d086b86.

Fixes: http://tracker.ceph.com/issues/20412
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #15933 from smithfarm/wip-hammer-jewel-x

jewel: tests: drop upgrade/hammer-jewel-x

tests: drop upgrade/hammer-jewel-x

This suite doesn't have any test logic in it. Its existence in the jewel branch
appears to be an oversight.

This cannot be cherry-picked from master because the upgrade/hammer-jewel-x
suite is present (and justified) in master and is not currently being dropped
there.

Signed-off-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #15438 from Vicente-Cheng/wip-20027-jewel

jewel: mds: issue new caps when sending reply to client

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

Merge pull request #15000 from jan--f/wip-19846-jewel

jewel: cephfs: normalize file open flags internally used by cephfs

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

Merge pull request #15383 from asheplyakov/20014-bp-jewel

jewel: cls/rgw: list_plain_entries() stops before bi_log entries

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

Merge pull request #15870 from smithfarm/wip-swift-task-move-jewel

tests: move swift.py task from teuthology to ceph, phase one (jewel)

Merge pull request #15842 from smithfarm/wip-sortbitwise-jewel

qa/suites/upgrade/hammer-x: set "sortbitwise" for jewel clusters

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

Merge pull request #15468 from smithfarm/wip-20140-jewel

jewel: cephfs: Journaler may execute on_safe contexts prematurely

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

tests: swift.py: tweak imports

The ".." form only works within the teuthology repo. With swift.py now in the
Ceph repo, we have to be explicit.

Error message was: "ValueError: Attempted relative import beyond toplevel
package

Signed-off-by: Nathan Cutler <ncutler@suse.com>

tests: upgrade/hammer-x/stress-split: tweak packages list

Include some hammer dependencies that aren't in the jewel default packages
list, and exclude some java packages that may not be in the hammer repo and are
not needed for the upgrade test in any case.

N.B.: This cannot be cherry-picked from master because upgrade/hammer-x was
dropped in master.

Signed-off-by: Nathan Cutler <ncutler@suse.com>

tests: swift.py: clone the ceph-jewel branch

The master branch of ceph/swift.git contains tests that are incompatible with
Jewel and Hammer. The ceph-jewel branch omits these tests.

Signed-off-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #15472 from smithfarm/wip-20148-jewel

jewel: mds: Too many stat ops when trying to probe a large file

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

Merge branch 'master' of /home/smithfarm/src/ceph/upstream/teuthology into wip-swift-task-move-jewel

tests: move swift.py task to qa/tasks

In preparation for moving this task from ceph/teuthology.git into ceph/ceph.git

The move is necessary because jewel-specific changes are needed, yet teuthology
does not maintain a separate branch for jewel. Also, swift.py is a
Ceph-specific task so it makes more sense to have it in Ceph.

Signed-off-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #15529 from badone/wip-async-sleep-timer-fix-jewel

jewel:  osd: Implement asynchronous scrub sleep

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

qa/suites/upgrade/hammer-x: set "sortbitwise" for jewel clusters

Inspired by 3734280522a913ca8340ebc98b80978f63bade6f

This cannot be cherry-picked from master because master does not have
qa/suites/upgrade/hammer-x

Signed-off-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #15824 from tchaikov/jewel

qa/workunits/rados/test-upgrade-*: whitelist tests the right way

Reviewed-by: Sage Weil <sage@redhat.com>

qa/workunits/rados/test-upgrade-*: whitelist tests the right way

--gtest_filter=POSTIVE_PATTERNS[-NEGATIVE_PATTERNS], so we cannot add
multiple exclusive patterns using -pattern:-pattern, instead, we should
use: -pattern:pattern

Signed-off-by: Kefu Chai <kchai@redhat.com>
Conflicts:
        qa/workunits/rados/test-upgrade-v11.0.0.sh: this change is not
    cherry-picked from master, because the clone-range op was removed
    from master. and only supported in pre-luminous releases.

Merge pull request #14661 from smithfarm/wip-19575-jewel

jewel: rgw: unsafe access in RGWListBucket_ObjStore_SWIFT::send_response()

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

Merge pull request #15778 from tchaikov/wip-upgrade-without-clone-range-jewel

qa/workunits/rados/test-upgrade-*: whitelist tests for master

Reviewed-by: Sage Weil <sage@redhat.com>

jewel: osd: pg_pool_t::encode(): be compatible with Hammer <= 0.94.6

This patch is necessary for Jewel only since direct upgrades from Hammer
to Kraken and newer are not supported.

Fixes: http://tracker.ceph.com/issues/19508
Signed-off-by: Alexey Sheplyakov <asheplyakov@mirantis.com>

qa/workunits/rados/test-upgrade-*: whitelist tests for master

The jewel-x upgrade test now runs this script against a mixed cluster on
a machine with code from master installed.  That means we have to
skip any new tests that will fail on a mixed cluster. CloneRange was
removed in 0d7b0b7.

Signed-off-by: Kefu Chai <kchai@redhat.com>
Conflicts:
    qa/workunits/rados/test-upgrade-v11.0.0.sh: this change is not
cherry-picked from master, because the clone-range op was removed from
master. and only supported in pre-luminous releases.

mds: issue new caps when sending reply to client

After Locker::issue_new_caps() adds new Capability data struct,
do not issue caps immediately. Let CInode::encode_inodestate()
do the job instead. This can avoid various races that early reply
is not allowed, caps that haven't been sent to client gets revoked.

Fixes: http://tracker.ceph.com/issues/19635
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
(cherry picked from commit 799703a4acb49db0b6cc99a23e4326767e694c3a)

ceph-disk: do not setup_statedir on trigger

trigger may run when statedir is unavailable and does not use it.

Fixes: http://tracker.ceph.com/issues/19941
Signed-off-by: Loic Dachary <loic@dachary.org>
(cherry picked from commit 16bfbdd3d9988523bba31aace516c303057daa58)

tests: rados: sleep before ceph tell osd.0 flush_pg_stats after restart

Even though we wait for HEALTH_OK after restarting the daemons, they are not
ready to respond to flush_pg_stats.

The reason why the osd is not ready for "tell" command after "ceph health"
shows that the cluster is "HEALTH_OK" is that the monitor fails to be notified
that the osd in question is not up in "heatbeat_interval". Because infernalis
does not have the osd_fast_fail_on_connection_refused support, the monitor
needs longer to detect that an osd is down, and osd_heartbeat_grace is used to
determine if an osd is down.

References: http://tracker.ceph.com/issues/16239
Signed-off-by: Nathan Cutler <ncutler@suse.com>
Signed-off-by: Kefu Chai <kchai@redhat.com>

rgw: fix crash when listing objects via swift

Fixes: http://tracker.ceph.com/issues/19249
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
(cherry picked from commit a9ec5e8ce184e19c009863db4d3519f9d8af91bd)

Conflicts:
        src/rgw/rgw_rest_swift.cc ("key" element of RGWObjEnt struct
           is not a reference; fix)

(cherry picked from commit 92b35155ff7b7492f3c50bf4f2ff0ffef2ca1c55)

Merge pull request #14752 from cbodley/wip-19474

jewel: rgw: allow system users to read SLO parts

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

build/ops: rpm: fix python-Sphinx package name for SUSE

This commit moves "BuildRequires: python-sphinx" down to the RH/CentOS/Fedora
distro conditional and adds a "BuildRequires: python-Sphinx" to the SUSE
conditional.

Signed-off-by: Jan Matejek <jmatejek@suse.com>
Signed-off-by: Nathan Cutler <ncutler@suse.com>
(cherry picked from commit 0688f57d1c35692cdddf219d9ac3571d03dbb8e0)

rgw: remove unnecessary output

a lot of radosgw-admin command will output
"error in read_id for object name: default : (2) No such file or directory"
when the zone named 'default' is not exist:

Signed-off-by: weiqiaomiao <wei.qiaomiao@zte.com.cn>
(cherry picked from commit 4e9c7be08b9828db2795536ce170a675d25bc1ed)

rgw: fix for zonegroup redirect url

local dest_url variable was shadowing the one in the enclosing scope, so
the changes were not applied and no Location header was written on redirect

Fixes: http://tracker.ceph.com/issues/19488
Signed-off-by: Casey Bodley <cbodley@redhat.com>
(cherry picked from commit 542e188a40f0495720b48308372366951ae41e62)

rgw: use zonegroup's master zone endpoints for bucket redirect

if no zonegroup endpoints are set, fall back to master zone endpoints

Fixes: http://tracker.ceph.com/issues/19488
Signed-off-by: Casey Bodley <cbodley@redhat.com>
(cherry picked from commit 92f63c6392bdc4633a2e57cb3867051bb1a3fd55)

client: update the 'approaching max_size' code

The old 'approaching max_size' code expects MDS set max_size to
'2 x reported_size'. This is no longer true. The new code reports
file size when half of previous max_size increment has been used.

Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
(cherry picked from commit 9316b0442c6f828dcf8da952e4c7a63c4db1398d)

Conflicts:
        src/client/Client.cc - in jewel, second argument to check_caps() is
            a bool (see 0df562a8e13 which is not in jewel)

mds: limit client writable range increment

For very large file, setting the writable range to '2 * file_size'
causes file recovery to run a long time. To recover a 1T file, Filer
needs to probe 2T~1T range.

Fixes: http://tracker.ceph.com/issues/19955
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
(cherry picked from commit 538f35bef944b18e9bca2b15ed7f4e8807ef0554)

Conflicts:
        src/mds/Locker.h - in jewel, file_update_finish() has different
            arguments than it does in master

build/ops: deb: fix logrotate packaging

This minimal jewel-only fix is not cherry-picked from master because
the Debian packaging was refactored between jewel and master and the
master fix is totally different.

Fixes: http://tracker.ceph.com/issues/20316
Signed-off-by: Nathan Cutler <ncutler@suse.com>

osdc/Journaler: avoid executing on_safe contexts prematurely

Journaler::_do_flush() can skip flushing some data when prezered
journal space isn't enough. Before updating Journaler::next_safe_pos,
we need to check if Journaler::_do_flush() has flushed enough data.

Fixes: http://tracker.ceph.com/issues/20055
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
(cherry picked from commit 6511e7a9e35a14216c03cd6921ca4d232274f953)

osdc/Journaler: make header write_pos align to boundary of flushed entry

This can speed up the process that detects and drops partial written
entry in the log tail.

Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
(cherry picked from commit 8ae2962b79903e217fda83cea4140af64b5d6883)

Conflicts:
src/osdc/Journaler.cc - 8d4f6b92cba is not being backported to jewel
src/osdc/Journaler.h - Journaler::Journaler initializer list is different in jewel, compared to master

Merge pull request #14672 from smithfarm/wip-19334-jewel

jewel: MDS heartbeat timeout during rejoin, when working with large amount of caps/inodes

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

pybind: fix cephfs.OSError initialization

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "cephfs.pyx", line 672, in cephfs.LibCephFS.open (/home/zhyan/Ceph/ceph-2/build/src/pybind/cephfs/pyrex/cephfs.c:10160)
  File "cephfs.pyx", line 155, in cephfs.OSError.__init__ (/home/zhyan/Ceph/ceph-2/build/src/pybind/cephfs/pyrex/cephfs.c:1889)
TypeError: __init__() takes exactly 3 positional arguments (2 given)

Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
(cherry picked from commit e6493f64ba4592b8dca54ece4464efa6c7f331a7)

pybind: fix open flags calculation

(O_WRONLY | O_RDWR) is invaild open flags

Fixes: http://tracker.ceph.com/issues/19890
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
(cherry picked from commit 2c25c99cb4572ffae97555a56b24a4c4097dcdec)

Merge pull request #14677 from smithfarm/wip-19665-jewel

jewel: mds: C_MDSInternalNoop::complete doesn't free itself

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

Merge pull request #15466 from smithfarm/wip-19762-jewel

jewel: cephfs: non-local quota changes not visible until some IO is done

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

Merge pull request #14700 from smithfarm/wip-19709-jewel

jewel: mds: enable start when session ino info is corrupt

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

Merge pull request #14685 from smithfarm/wip-19675-jewel

jewel: cephfs: Test failure: test_data_isolated (tasks.cephfs.test_volume_client.TestVolumeClient)

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

Merge pull request #14684 from smithfarm/wip-19673-jewel

jewel: cephfs: mds is crushed, after I set about 400 64KB xattr kv pairs to a file

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

Merge pull request #14683 from smithfarm/wip-19671-jewel

jewel: cephfs: MDS assert failed when shutting down

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

Merge pull request #14682 from smithfarm/wip-19668-jewel

jewel: cephfs: MDS goes readonly writing backtrace for a file whose data pool has been removed

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

Merge pull request #14679 from smithfarm/wip-19666-jewel

jewel: cephfs: The mount point break off when mds switch hanppened.

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

Merge pull request #14676 from smithfarm/wip-19619-jewel

jewel: cephfs: MDS server crashes due to inconsistent metadata.

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

Merge pull request #14674 from smithfarm/wip-19482-jewel

jewel: cephfs: No output for ceph mds rmfailed 0 --yes-i-really-mean-it command

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

Merge pull request #14671 from smithfarm/wip-19044-jewel

jewel: tests: buffer overflow in test LibCephFS.DirLs

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

Merge pull request #14670 from smithfarm/wip-18949-jewel

jewel: mds: avoid reusing deleted inode in StrayManager::_purge_stray_logged

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

Merge pull request #14669 from smithfarm/wip-18900-jewel

jewel: cephfs: Test failure: test_open_inode

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

Merge pull request #14668 from smithfarm/wip-18705-jewel

jewel: mds: fragment space check can cause replayed request fail

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

Merge pull request #14698 from smithfarm/wip-19677-jewel

jewel: cephfs: ceph-fuse does not recover after lost connection to MDS

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

Merge pull request #14766 from smithfarm/wip-19757-jewel

jewel: rgw: fix failed to create bucket if a non-master zonegroup has a single zone

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

Merge pull request #14787 from linuxbox2/jewel-rgw-shard-limit-ck

jewel: rgw: add bucket size limit check to radosgw-admin

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

Merge pull request #14789 from mdw-at-linuxbox/wip-jewel-rgw-rvk

jewel: rgw: swift: disable revocation thread if sleep == 0 || cache_size == 0

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

Merge pull request #14815 from smithfarm/wip-19786-jewel

jewel: rgw: failure to create s3 type subuser from admin rest api

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

Merge pull request #15312 from theanalyst/wip-20078

jewel: rgw: only append zonegroups to rest params if not empty

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

Merge pull request #15382 from theanalyst/wip-mem-leak2

jewel: rgw:fix memory leaks in data/md sync

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

osd: Move scrub sleep timer to osdservice

PR 14886 erroneously creates a scrub sleep timer for every pg resulting
in a proliferation of threads. Move the timer to the osd service so
there can be only one.

Fixes: http://tracker.ceph.com/issues/19986
Signed-off-by: Brad Hubbard <bhubbard@redhat.com>
(cherry picked from commit f110a82437df79dc20207d296e8229fc0e9ce18b)

Conflicts:
        src/osd/PG.cc - ceph_clock_now requires a CephContext argmunent
        in Jewel

osd: Implement asynchronous scrub sleep

Rather than blocking the main op queue just do an async sleep.

Fixes: http://tracker.ceph.com/issues/19497
Signed-off-by: Brad Hubbard <bhubbard@redhat.com>
(cherry picked from commit 7af3e86c2e4992db35637864b83832535c94d0e6)