common/TrackedOp: Move tracking_enabled check into register_inflight_op()

Make tracking_enabled (and the lock private.)*

Signed-off-by: David Zafman <dzafman@redhat.com>
(Partial cherry picked from commit 4e8e6172e8f6cc2adc63a429c32f5edd03982f5e)

* Backport: Exclude locking since tracking_enabled can't change after launch

common/TrackedOp: Handle dump racing with constructor

Use is_tracked to prevent TrackedOp::dump() from trying to call
virtual function while still in OpRequest constructor.

Fixes: #8885
Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit ad13e05499669a79bde9a219ba1089f929e0388e)

Conflicts:
src/common/TrackedOp.cc
src/common/TrackedOp.h
Hammer can't use atomic<bool> so use atomic_t instead

common/TrackedOp: Missed locking when examining "events"

Another crash caused by a dump_ops_in_flight similar to #8885

Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit be098c83143226e502a0587c9215cff941dbace1)

Conflicts:
src/common/TrackedOp.h (trivial)

CLEANUP: Move locking into dump_ops_in_flight()/dump_historic_ops()

Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit 17c645f70708134f50374cadb5505b0babe2265a)

Conflicts:
src/mds/MDSRank.cc (variation of code exists in mds/MDS.cc)
src/mon/Monitor.cc (no equivalent code here)

mds, osd: Fix missing locking for "dump_blocked_ops"

Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit be5bed840f21da998be3b8374a8e33f208258e1b)

Conflicts:
src/mds/MDSRank.cc (missing dump_blocked_ops in mds)

osd: cleanup: Specify both template types for create_request()

Code consistency

Signed-off-by: David Zafman <dzafman@redhat.com>
(cherry picked from commit cecdc49e6d2f38d5eb9f85218484f7a02bd2e1b7)

osd: add dump_blocked_ops asok command.

From ceph health we can see there are blocked ops and from 'ceph health
detai' we can see which osd has blocked ops. Later we only in log of osd
to see the detail message of blocked ops.
Now via asok dump blocked ops.

Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
(cherry picked from commit cf4e8c655c4a0e6c367c88338cd6e51a3474fa2f)

Conflicts:
src/common/TrackedOp.h (trivial)

common/TrackedOp: Should lock ops_history_lock when access shutdown.

Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
(cherry picked from commit 7fe72a2e7d116093ccff6683666e6289f5f12421)

common/TrackedOp: checking in flight ops fix

Return earlier if tracking is not enabled when checking in flight ops.

Signed-off-by: Zhiqiang Wang <zhiqiang.wang@intel.com>
(cherry picked from commit 7509a6caa4964d71625c43f2ddc10e6274b6a21b)

common/OpTracker: don't dump ops if tracking is not enabled

If tracking is disabled, there is nothing to dump. No need to go through
each sharded list. This is true for both in flight and historical ops.

Signed-off-by: Zhiqiang Wang <zhiqiang.wang@intel.com>
(cherry picked from commit ec3c40952bbc4b495875e08b57d764a7487c6c09)

common/TrackedOp: check tracking_enabled for event "initiated/done".

Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
(cherry picked from commit d7585c020de41f25e6a598661844169f2bdbb2b4)

common/TrackedOp: clean up code make look good.

Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
(cherry picked from commit edefd05edce6728774bb396ea4e6641ea53463cd)

Merge pull request #12006 from SUSE/wip-17905-hammer

hammer: mon: MonmapMonitor should return success when MON will be removed

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

mon: MonmapMonitor: return success when monitor will be removed

Fixes: http://tracker.ceph.com/issues/17725
Signed-off-by: Joao Eduardo Luis <joao@suse.de>
(cherry picked from commit c9d46cfbf2512bc3495c6901de2b8f711bef9bae)

Conflicts:
src/mon/MonmapMonitor.cc (just inserted the "err = 0" assignment
        immediately after the error message line)

Merge pull request #11372 from ceph/wip-17386-hammer

hammer: doc: add "Upgrading to Hammer" section

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

doc: add "to Hammer" upgrade section

Fixes: http://tracker.ceph.com/issues/17534
Signed-off-by: Kefu Chai <kchai@redhat.com>

Merge pull request #11253 from ceph/wip-offline-split-hammer2

ceph-objectstore-tool: add a way to split filestore directories offline

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

ceph-objectstore-tool: add a way to split filestore directories offline

Use the usual split code, and split each dir that meets the
usual split criteria.

This can be run with lower than usual split settings, to avoid costly
online splits. To make sure the directories are not merged again, use
a load merge threshold (e.g. filestore merge threshold = 1), and
adjust the split multiplier accordingly.

Fixes: http://tracker.ceph.com/issues/17220
Signed-off-by: Josh Durgin <jdurgin@redhat.com>
(cherry picked from commit e7b0428e0e8d8f5459311dc698d94a3ac4f04684)

Conflicts:
src/tools/ceph_objectstore_tool.cc
* remove c++11 auto usage
* change HashIndex::list_subdirs() call to use set instead of vector
* adjust to hammer signature of coll_t::is_pg()

Merge pull request #11140 from cbodley/wip-radosgw-admin-man-hammer

hammer: add orphan options to radosgw-admin --help and man page

Reviewed-by: Ken Dreyer <kdreyer@redhat.com>

man: add orphan options to radosgw-admin.8

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

Add two options to radosgw-admin.rst manpage

Add '--job-id' and '--max-concurrent-ios' to Orphan Search Options

Fixes: http://tracker.ceph.com/issues/17281
Signed-off-by: Thomas Serlin <tserlin@redhat.com>
(cherry picked from commit 697f30d86f49b73c981c06375ab2937570b1db01)

radosgw-admin: add "--orphan-stale-secs" to --help

The radosgw-admin --help did not include the description of the
`--orphan-stale-secs` option of the `orphans find` command. The option
sets the number of seconds to wait before declaring an object to be an
orphan.

Fixes: http://tracker.ceph.com/issues/17280
Signed-off-by: Ken Dreyer <kdreyer@redhat.com>
(cherry picked from commit 354059ae43b4f4cc797da1669715399cd96a4738)

rgw_admin: improve the orphans find help

Adding  a few missing commandline switches and improve the help message

Signed-off-by: Abhishek Lekshmanan <abhishek@suse.com>
(cherry picked from commit 27e4c46f741a5f908d5c8ca1af227b3581c697f6)

doc: add "--orphan-stale-secs" to radosgw-admin(8)

The radosgw-admin(8) manual page did not include the description of the
`--orphan-stale-secs` option of the `orphans find` command. The option sets
the number of seconds to wait before declaring an object to be an
orphan.

Fixes: http://tracker.ceph.com/issues/17280
Signed-off-by: Ken Dreyer <kdreyer@redhat.com>
(cherry picked from commit a676c516069cc448591018ecf4f7d1f7f7bc3bfd)

Merge pull request #10839 from oritwas/hammer-default-quota

Hammer default quota

Reviewed-by: Daniel Gryniewicz <dang@redhat.com>

0.94.9

Merge pull request #10913 from ceph/wip-revert-random-uuid-hammer

hammer: revert use of boost for uuids

Reviewed-by: Alfredo Deza <adeza@redhat.com>

rgw: Add documentation for RGW default quota

Fixes http://tracker.ceph.com/issues/16447
Cherry picked from e3ab3476146cf0fe604783e9e5a0d63237271735

This was modified to use size in kB, rather than size in bytes, since
Jewel has not yet been converted to use size in bytes.

Signed-off-by: Pavan Rallabhandi <PRallabhandi@walmartlabs.com>
Signed-off-by: Daniel Gryniewicz <dang@redhat.com>
(cherry picked from commit 11328d7dcc6634ab81ebce4801ebc243a27b84b0)

rgw: Do not bother to check against conf values during quota handling

Fixes http://tracker.ceph.com/issues/16447
Cherry picked from 17d2c1712a5b72315a47ab0f8380331bfd478c0b

This was modified to use size in kB, rather than size in bytes, since
Jewel has not yet been converted to use size in bytes.

Signed-off-by: Pavan Rallabhandi <PRallabhandi@walmartlabs.com>
Signed-off-by: Daniel Gryniewicz <dang@redhat.com>
(cherry picked from commit c85e9d4e1e3f367150314c08995c7a4d418910ff)

rgw: Let the default quota settings take effect during user creation

Fixes http://tracker.ceph.com/issues/16447
Cherry picked from 400d7c982f18efd7cf96acfc3a63078791d1ec0a

Signed-off-by: Pavan Rallabhandi <PRallabhandi@walmartlabs.com>
Signed-off-by: Daniel Gryniewicz <dang@redhat.com>
(cherry picked from commit e07fa1befabd56a3f353bcca50229620abd08dec)

Merge pull request #10898 from SUSE/wip-release-notes-hammer

doc: release-notes: add missing hammer releases

Reviewed-by: Loic Dachary <ldachary@redhat.com>

Revert "moved to use boost uuid implementation, based on commit 4fe89a7b14c97b2ed7f357132901beb2bdcec551"

This reverts commit 174de7fce8080df6e02b363e7821e8cca6f8157f.

Revert "uuid: use boost::random:random_device"

This reverts commit 187d308b5b0cb0512b442bbad1bddfda8ef9203f.

This breaks precise builds.

doc: release-notes.rst: add missing line to v0.94.8

see 46264a3b6cb7f490680c79bf0957c36aeaea5f6c and
b83d25d2918cbd880fbaeec976bffcea8868d5d9 in master

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

doc: add missing changelogs up to 0.94.8

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

doc: release-notes: add missing hammer releases

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

0.94.8

Merge pull request #10750 from ceph/wip-14799-hammer

hammer: rocksdb do not link against tcmalloc if it's disabled

Reviewed-by: Loic Dachary <ldachary@redhat.com>

rocksdb: disable tcmalloc if disabled

the rocksdb/configure.ac does not support --with-tcmalloc before,
and the updated rocksdb commit has the change for '--with-tcmalloc'
option, so let's pick it up.

Fixes: http://tracker.ceph.com/issues/14799
Signed-off-by: Kefu Chai <kchai@redhat.com>

ceph.spec: respect CEPH_EXTRA_CONFIGURE_ARGS

Signed-off-by: Sage Weil <sage@redhat.com>
(cherry picked from commit 51abff11688f0201b8f4076ac515e4515929d4cb)
Signed-off-by: Brad Hubbard <bhubbard@redhat.com>

Merge pull request #10718 from oritwas/wip-14328-hammer

hammer: rgw: the swift key remains after removing a subuser

Merge pull request #10509 from yehudasa/wip-15557-hammer

rgw: Have a flavor of bucket deletion to bypass GC and to trigger

rgw: fix subuser rm command failure

Fixes: 14375
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
(cherry picked from commit bbc06e70b49b2917ee63985727cb92724fb99209)

rgw: add a method to purge all associate keys when removing a subuser

Fixes: #12890
When removing a subuser, make sure all of its keys, including the swift key and possible s3 keys, are also deleted.

Signed-off-by: Sangdi Xu <xu.sangdi@h3c.com>
(cherry picked from commit e7b7e1afc7a81c3f97976f7442fbdc5118b532b5)

Merge pull request #10291 from badone/wip-13829-hammer

hammer: config set with negative value results in "error setting 'filestore_merge_threshold' to '-40': (22) Invalid argument"

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #10699 from ktdreyer/wip-15867-nroff

jewel: doc: fix by-parttypeuuid in ceph-disk(8) nroff

Reviewed-by: Loic Dachary <ldachary@redhat.com>

doc: fix by-parttypeuuid in ceph-disk(8) nroff

Commit 221efb0b893adbfd7a19df171cf967fee87afcc7 altered the rST source
for the ceph-disk man page. In Hammer, we also have to modify the nroff
sources, because static copies of the generated man pages are stored in
Git.

Fixes: http://tracker.ceph.com/issues/15867
Signed-off-by: Ken Dreyer <kdreyer@redhat.com>

Merge pull request #9741 from SUSE/wip-16343-hammer

hammer: boost uuid makes valgrind complain

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Sage Weil <sage@redhat.com>
Reviewed-by: Ken Dreyer <kdreyer@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #10554 from Vicente-Cheng/wip-16815-hammer

hammer: radosgw-admin region-map set is not reporting the bucket quota correctly
Reviewed-by: Orit Wasserman <owasserm@redhat.com>

Merge pull request #10555 from Vicente-Cheng/wip-16317-hammer

hammer: rgw: bucket listing following object delete is partial
Reviewed-by: Orit Wasserman <owasserm@redhat.com>

Merge pull request #10530 from yehudasa/wip-16412-hammer

rgw: remove bucket index objects when deleting the bucket
Reviewed-by: Orit Wasserman <owasserm@redhat.com>

Merge pull request #8755 from kmroz/wip-hammer-convert-plain-object-to-versioned

hammer: rgw: convert plain object to versioned (with null version) when removing
Reviewed-by: Orit Wasserman <owasserm@redhat.com>

Merge pull request #9271 from XinzeChi/wip-fix-digest-hammer

osd: fix omap digest compare when scrub

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: David Zafman <dzafman@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #8805 from SUSE/wip-15482-hammer

hammer: is_split crash in handle_pg_create

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

Merge pull request #8593 from renhwsky/wip-15478-hammer

hammer: Pool quota alarm is not in effect

Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #7542 from SUSE/wip-14668-hammer

hammer: Wrong ceph get mdsmap assertion

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

Merge pull request #8042 from ukernel/hammer-15050

hammer: mds: fix stray purging in 'stripe_count > 1' case

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

Merge pull request #8804 from SUSE/wip-15512-hammer

hammer: Double decreased the count to trim caps which will cause failing to respond to cache pressure

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

Merge pull request #9404 from SUSE/wip-16082-hammer

hammer: mds: wrongly treat symlink inode as normal file/dir when symlink inode is stale on kcephfs

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

Merge pull request #10198 from SUSE/wip-16626-hammer

hammer: Failing file operations on kernel based cephfs mount point leaves unaccessible file behind on hammer 0.94.7

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

rgw: reset return code in when iterating over the bucket the objects

Fixes: #14826
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
(cherry picked from commit e23fdcbbce47aaf041b9f8a0812c8a729aa5014b)

rgw: fix compilation

Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
(cherry picked from commit 9dfbae6d65bf6f8ab7cdf8597d556505b81cedd9)

rgw: add bucket_quota to RGWRegionMap::encode_json

Signed-off-by: Orit Wasserman <owasserm@redhat.com>
(cherry picked from commit e18b72f7e97a65af681e368d3706a465603df89d)

rgw: Have a flavor of bucket deletion to bypass GC and to trigger
object deletions async.

Fixes: http://tracker.ceph.com/issues/15557
Signed-off-by: Pavan Rallabhandi <PRallabhandi@walmartlabs.com>
(cherry picked from commit b7a69fca248afeef1de1278890076693b16cf6d3)

Conflicts:
src/rgw/rgw_admin.cc
src/rgw/rgw_bucket.cc
src/rgw/rgw_bucket.h
src/test/cli/radosgw-admin/help.t

rgw: remove bucket index objects when deleting the bucket

Fixes: http://tracker.ceph.com/issues/16412
Signed-off-by: Orit Wasserman <owasserm@redhat.com>
(cherry picked from commit 3ae276390641ad5fc4fef0c03971db95948880b4)

Merge pull request #10182 from gaurav36/wip-configure-xfs-issue-hammer

hammer: configure: Add -D_LARGEFILE64_SOURCE to Linux build.

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Casey Bodley <cbodley@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #7961 from dzafman/wip-15002

hammer: objects unfound after repair (fixed by repeering the pg)

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #9125 from dzafman/wip-15635

hammer: [MON] Fixed calculation of %USED

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #8464 from dzafman/wip-13039-15315

hammer: "FAILED assert(pg->info.history.same_interval_since == p.same_interval_since)" in upgrade:hammer-x-jewel-distro-basic-openstack

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #8573 from Vicente-Cheng/wip-15293-hammer

hammer: ECBackend: send write and read sub ops on behalf of client ops at normal priority

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #8665 from Vicente-Cheng/wip-15525-hammer

hammer: ReplicatedPG.cc: 430: FAILED assert(!actingbackfill.empty()): old watch timeout tries to queue repop on replica

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #8806 from SUSE/wip-15320-hammer

hammer: LibRadosMiscConnectFailure.ConnectFailure (not so intermittent) failure in upgrade/hammer-x

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #9001 from Vicente-Cheng/wip-15730-hammer

hammer: osd: acting_primary not updated on split

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #9400 from Vicente-Cheng/wip-15796-hammer

hammer: LibRadosWatchNotifyPPTests/LibRadosWatchNotifyPP.WatchNotify2Timeout/1 segv

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #9416 from chardan/wip-15770-hammer

hammer: mon/OSDMonistor: improve reweight_by_utilization() logic

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #10189 from dreamhost/wip-16618-hammer

rgw: fix multi-delete query param parsing.

Reviewed-by: Yehuda Sadeh <yehuda@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #8379 from Vicente-Cheng/wip-15283-hammer

hammer: RGW shouldn't send Content-Type nor Content-Length for 304 responses

Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #8313 from SUSE/wip-15257-hammer

hammer: rgw: radosgw server abort when user passed bad parameters to set quota

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

Merge pull request #8528 from wido/hammer-issue-15348

rgw: Set Access-Control-Allow-Origin to a Asterisk if allowed in a rule

Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #9671 from ceph/hammer-rgw-copy-if-newer

hammer: rgw: backport rgwx-copy-if-newer for radosgw-agent

Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #9961 from SUSE/wip-16190-hammer

hammer: subuser rm fails with status 125

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

Merge pull request #9402 from SUSE/wip-15966-hammer

hammer: No Last-Modified, Content-Size and X-Object-Manifest headers if no segments in DLO manifest

Reviewed-by: Orit Wasserman <owasserm@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

mon/OSDMonitor: avoid potential expensive grace calculation

The grace calculation during check_failure() is now very complicated
and time-consuming. Therefore we shall skip this when it is possible.

Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
(cherry picked from commit 3557903d5d57642179b2ae137bedc389974b1956)

 Conflicts:
src/mon/OSDMonitor.cc

Resolved by choosing the move-to-top implementation. Removed unused vars.

mon/OSDMonitor: improve reweight_by_utilization() logic

By calling reweight_by_utilization() method, we are aiming at an evener result
of utilization among all osds. To achieve this, we shall decrease weights of
osds which are currently overloaded, and try to increase weights of osds which
are currently underloaded when it is possible.
However, we can't do this all at a time in order to avoid a massive pg migrations
between osds. Thus we introduce a max_osds limit to smooth the progress.

The problem here is that we have sorted the utilization of all osds in a descending
manner and we always try to decrease the weights of the most overloaded osds
since they are most likely to encounter a nearfull/full transition soon, but
we won't increase the weights from the most underloaded(least utilized by contrast)
at the same time, which I think is not quite reasonable.

Actually, the best thing would probably be to iterate over teh low and high osds
in parallel, and do the ones that are furthest from the average first.

Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
(cherry picked from commit e7a32534ebc9e27f955ff2d7a8d1db511383301e)

 Conflicts:
src/mon/OSDMonitor.cc

Resolved by picking the lambda implemenation.
NOTE: Because hammer does not support C++11, the lambda functionality from the
current master has been moved into the "Sorter" function object.

Merge pull request #10038 from tchaikov/wip-16297-hammer

mon: Monitor: validate prefix on handle_command()

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #9090 from tchaikov/wip-13990-hammer

hammer: osd: remove all stale osdmaps in handle_osd_map()

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #10227 from SUSE/wip-15081-hammer

hammer: When logging to a file fails, ceph logs excessively to stderr

Reviewed-by: Samuel Just <sjust@redhat.com>

Merge pull request #9893 from Vicente-Cheng/wip-16430-hammer

hammer: OSDMonitor: drop pg temps from not the current primary

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #9238 from SUSE/wip-15962-hammer

hammer: ceph-osd valgrind invalid reads/writes

Reviewed-by: Samuel Just <sjust@redhat.com>

Merge pull request #8849 from SUSE/wip-14326-hammer

hammer: osd/PGLog.h: 88: FAILED assert(rollback_info_trimmed_to == head)

Reviewed-by: Samuel Just <sjust@redhat.com>

Merge pull request #8845 from SUSE/wip-15647-hammer

hammer: osd: rados cppool omap to ec pool crashes osd

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #8470 from dachary/wip-15403-api-test-timeout-hammer

hammer: tests: be more generous with test timeout

Reviewed-by: Loic Dachary <ldachary@redhat.com>

Merge pull request #9977 from shun-s/hammer

replcatedBackend: delete one of the repeated op->mark_started in ReplicatedBackend::sub_op_modify_impl

Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #8960 from badone/wip-hammer-rados-bench-units

hammer: Add units to rados bench output and clean up formatting

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

OSDMonitor: drop pg temp from sources other than the current primary

Fixes: http://tracker.ceph.com/issues/16127
Signed-off-by: Samuel Just <sjust@redhat.com>
(cherry picked from commit 1a07123c38e3fecb3fb2e43bbbae962d8411d287)

change:
            use NULL to replace the nullptr because we don't have C++11

osd: reset session->osdmap if session is not waiting for a map anymore

we should release the osdmap reference once we are done with it,
otherwise we might need to wait very long to update that reference with
a newer osdmap ref. this appears to be an OSDMap leak: it is held by an
quiet OSD::Session forever.

the osdmap is not reset in OSD::session_notify_pg_create(), because its
only caller is wake_pg_waiters(), which will call
dispatch_session_waiting() later. and dispatch_session_waiting() will
check the session->osdmap, and will also reset the osdmap if
session->waiting_for_pg.empty().

Fixes: http://tracker.ceph.com/issues/13990
Signed-off-by: Kefu Chai <kchai@redhat.com>
(cherry picked from commit 82b0af7cedc3071cd83ee53479f834c23c62b7d0)

qa: Add test for #13829

qa/workunits/cephtool/test.sh: add test for setting negative int options
src/test/daemon_config.cc: remove tests for failed negative values

Fixes: #13829
Signed-off-by: Brad Hubbard <bhubbard@redhat.com>
(cherry picked from commit 994ac2942d04584d1617e6d4bbd5b880b1ea0448)

common: Allow config set with negative value

A recent commit disabled negative values but they are required for variables
such as filestore_merge_threshold.

Modified patch to remove C++11 specific elements so it will build for hammer

Fixes: #13829
Signed-off-by: Kefu Chai <kchai@redhat.com>
Signed-off-by: Brad Hubbard <bhubbard@redhat.com>
(cherry picked from commit 8b777a0c346bc70fd10d07e89368b3785b58f10e)

log: do not repeat errors to stderr

If we get an error writing to the log, log it only once to stderr.
This avoids generating, say, 72 GB of ENOSPC errors in
teuthology.log when /var/log fills up.

Fixes: #14616
Signed-off-by: Sage Weil <sage@redhat.com>
(cherry picked from commit d9ac0474b864afda58a44b9012cca4bbc6aaf509)

Conflicts:
src/log/Log.cc (drop m_uid and m_gid which are not used in hammer;
order of do_stderr, do_syslog, do_fd conditional blocks is reversed in
hammer; drop irrelevant speed optimization code from
5bfe05aebfefdff9022f0eb990805758e0edb1dc)

mds: disallow 'open truncate' non-regular inode

Signed-off-by: Yan, Zheng <zyan@redhat.com>
(cherry picked from commit 0e4b6f2332bb4822cf324587a94144f1c98e4b97)

Conflicts:
src/mds/Server.cc (hammer has cur->inode.inline_version - in master
this has been changed to cur->inode.inline_data.version)