osd/PrimaryLogPG: can_create false for cache ops

Some cache ops do writes, but none of them *create* objects, so we
should not set can_create.

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

osd/PrimaryLogPG: drop odd maintain_ssc hack

This was added two years ago by b61f3e43f1c9d43daa0dedd7cbd0fe5787cbdfbb,
but I can see why it is needed here.  The test case for that
original bug (from a3442cf07841a7269713e30ca270f2f9c356a598) passes
fine.

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

ceph-objectstore-tool: drop pre-snapset support for setting up snapmapper

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

ceph-objectstore-tool: drop head_exists corruption

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

ceph_test_rados_api_pool: make LibRadosPools.PoolList tolerate long pool names

There was a baked-in assumption that the first pool was <20 chars.
No more!

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

osd/osd_types: object_info_t: remove legacy_snaps

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

osd/PrimaryLogPG: drop last legacy_snaps users (just debug output!)

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

osd/osd_types: SnapSet: remove head_exists

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

osd: drop last head_exists = true calls

No more users!

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

osd/PrimaryLogPG: find_object_context: drop !head_exists path

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

osd/PrimaryLogPG: do_osd_ops: head_exists is always true for LIST_SNAPS

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

osd/osd_types: SnapSet: drop is_legacy

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

osd/PrimaryLogPG: start_flush: drop legacy snapset handling

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

osd/osd_types: default head_exists = true

This will be helpful as we start removing code that would otherwise
set this to true in some convoluted way.

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

osd/PrimaryLogPG: finish_promote: drop legacy snapset handling

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

osd/PrimaryLogPG: do_copy_get: drop legacy snapset path

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

osd/PrimaryLogPG: _delete_oid: drop legacy snapset handling

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

osd/PrimaryLogPG: do_osd_ops: drop legacy snapset handling from LIST_SNAPS

This is so much simpler, yay!

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

osd/PrimaryLogPG: on_local_recover: drop legacy snapset handling

This was there to deal with jewel peers, which are no more.

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

osd/PG: _scan_snaps: drop legacy snapset handling

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

osd/PrimaryLogPG: find_object_context: drop legacy snapset handling

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

osd/PrimaryLogPG: issue_repop: drop obsolete condition on head_obc

The only user is trim_object, which sets obc = clone, head_obc =
head.

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

osd/PrimaryLogPG: snapset_obc -> head_obc

This is only used for trim_object, now!

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

osd/PrimaryLogPG: do_op: drop snapset_obc lookup on snapdir

We'll never find it.  This also means we can drop the goofy hack
to clear it for cache tiers.

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

osd/PrimaryLogPG: trim_object: ctx->snapset_obc -> head_obc

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

osd/PrimaryLogPG: trim_object: simplify a bit more for post-snapdir world

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

osd/PrimaryLogPG: trim_object: drop legacy SnapSet handling

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

osd/PrimaryLogPG: trim_object: rename oid, obc vars {snap->head}

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

osd/PG: chunky_scrub: degeneralize end boundary condition to head

We can swap has_snapset() for is_head(), and drop the get_head()
on hobject_t's we know are head.

(The original constraint on the scrub boundary is still needed;
the clones get created to the left of both head and snapdir.)

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

osd/PrimaryLogPG: do_osd_ops: update CACHE_EVICT comment

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

osd/PrimaryLogPG: do_osd_ops: update LIST_SNAPS comment

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

osd/PrimaryLogPG: start_copy: simplify assert

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

osd/PrimaryLogPG: start_flush: update comment for a post-snapdir world

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

osd/PrimaryLogPG: find_object_context: drop snapdir handling

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

osd/PrimaryLogPG: do_op: move up LIST_SNAPS specialness

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

osd/PrimaryLogPG: do_op: use head on for LIST_SNAPS

Use head oid even for LIST_SNAPS op.  This doesn't really affect
LIST_SNAPS but it *does* mean that we never call find_object_context
on a snapdir oid.

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

osd/PrimaryLogPG: do_op: move and strength LIST_SNAPs vs snapdir check

Verify that *only* the LIST_SNAPS op uses an oid with snapdir set.

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

osd/PrimaryLogPG: add_object_context_to_pg_stat: drop snapdir handling

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

osd/PrimaryLogPG: get_snapset_context: don't look on snapdir

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

osd/PrimaryLogPG: recover_missing: no need to look for snapdir

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

osd/PrimaryLogPG: recover_replicas: don;t care about missing snapdir

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

osd/PrimaryLogPG: recover_backfill: remove snapdir hackery

Phew!

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

osd/PrimaryLogPG: scrub_snapshot_metadata: no snapdirs

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

osd/PrimaryLogPG: remove some snapdir-specific head_exists cases

We don't care about these any more.

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

osd/PrimaryLogPG: scrub_snapshot_metadata: remove snapset conversion

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

osd/PrimaryLogPG: scrub_snapshot_metadata: no need for snapdir

The object never exists; no need to look for it.

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

osd/PrimaryLogPG: finish_ctx: remove snapdir handling

- we neither need to delete or create snapdir objects
- snapset is only ever written to head; move bufferlist bss and
encode call there
- drop legacy_snapset encoding path

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

osd/PrimaryLogPG: make_writeable: remove legacy SnapSet handling

The SnapSet will now never be 'legacy'.

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

osd/PrimaryLogPG: make_writeable: whitespace

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

osd/ReplicatedBackend: prepare_pull: simplify assert

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

osd/ReplicatedBackend: never missing snapdirs

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

osd/PrimaryLogPG: maybe_await_blocked_{snapset->head}

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

osd/PrimaryLogPG: do_op: no need to block on missing/degraded snapdir

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

osd: PrimaryLogPG: no snapdirs objects during pgls

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

osd: scrub: no need to expect snapdir objects

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

Merge pull request #17621 from idryomov/wip-krbd-xfstests-blockdev

qa/suites/krbd: enable xfstests blockdev tests

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

Merge pull request #17453 from kungf/eversion_t_opt_max

osd_types: define max in eversion_t::max() to static

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

Merge pull request #17571 from jcsp/wip-21260

mon: fix dropping mgr metadata for active mgr (#21260)

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

Merge pull request #17572 from xiexingguo/wip-os-tool

tools/ceph_objectstore_tool: fix 'dup' unable to duplicate meta PG

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

Merge pull request #17585 from jdurgin/wip-recovery-delete-encoding-master

MOSDPGRecoveryDelete[Reply]: bump header version

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

Merge pull request #17588 from kungf/aio_submit_size_change

os/bluestore: shrink aio submit size to pending value

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

Merge pull request #17599 from tchaikov/wip-encoding-optional

encoding: reset optional<> if it is uninitialized

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

Merge pull request #17600 from jecluis/wip-21309

mon/OSDMonitor: don't create pgs if pool was deleted

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

Merge pull request #17298 from xiexingguo/wip-pg-null-obc-access

osd/PrimaryLogPG: maybe_handle_manifest_detail - sanity check obc existence

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

Merge pull request #17559 from dillaman/wip-21181

qa/workunits/rbd: relax greps to support upgrade formatting change

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

Merge pull request #17324 from smithfarm/wip-rbd-symlink

build/ops: rpm: drop legacy librbd.so.1 symlink in /usr/lib64/qemu

Reviewed-by: Ken Dreyer <kdreyer@redhat.com>
Reviewed-by: Ademar de Souza Reis Jr <areis@redhat.com>
Reviewed-by: Boris Ranto <branto@redhat.com>

Merge pull request #17512 from tchaikov/wip-chmod-x

common,mon,osd,pybind: silence warning and remove executable mode bit

Reviewed-by: J. Eric Ivancich <ivancich@redhat.com>

Merge pull request #17539 from tchaikov/wip-21204

mon,option: set default value for mon_dns_srv_name

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

Merge pull request #17549 from iliul/fix-texttable-align

tools: Modify "rados df" header's alignment

Reviewed-by: xie xingguo <xie.xingguo@zte.com.cn>

Merge pull request #17096 from ceph/wip-ansbile-fixes

[qa/tests]: misc ceph-ansible fixes and udpate

Merge pull request #17513 from Liuchang0812/wip-max-avail-in-df

mon: incorrect MAX AVAIL in "ceph df"

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

Merge PR #17533 into master

* refs/remotes/upstream/pull/17533/head:
qa: log ceph-fuse kill/cleanup

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

Merge PR #17506 into master

* refs/remotes/upstream/pull/17506/head:
mds: reset Formatter on error

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

Merge PR #17466 into master

* refs/remotes/upstream/pull/17466/head:
qa: whitelist expected MDS_READ_ONLY for kcephfs test

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

Merge PR #17456 into master

* refs/remotes/upstream/pull/17456/head:
mds: print subtrees only after merge

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

Merge PR #17452 into master

* refs/remotes/upstream/pull/17452/head:
Make standby-replay mds not migrate empty subtree when trimming cache
Make standby-replay mds avoid initiating subtree export

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

Merge PR #17384 into master

* refs/remotes/upstream/pull/17384/head:
ceph.in: validate service glob

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

Merge PR #16752 into master

* refs/remotes/upstream/pull/16752/head:
client: NULL check before append value to bufferlist

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Jos Collin <jcollin@redhat.com>
Reviewed-by: Piotr Dałek <piotr.dalek@corp.ovh.com>

Stop the mgr on node which is not client, this will ensure
the client.0's mgr is active mgr for workunit to work.

Signed-off-by: Vasu Kulkarni <vasu@redhat.com>

Merge pull request #16272 from liewegas/wip-balancer

mgr/balancer: mgr module to automatically balance PGs across OSDs

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

os/bluestore: shrink aio submit size to pending value

Signed-off-by: kungf <yang.wang@easystack.cn>

Merge pull request #17403 from ceph/wip-bz1485011

ceph-volume Persist non-lv devices for journals

Reviewed-by: Andrew Schoen aschoen@redhat.com

Merge pull request #17603 from liewegas/wip-migration-twiddle

doc: restructure bluestore migration insructions

Reviewed-by: David Galloway <dgallowa@redhat.com>

doc: restructure bluestore migration insructions

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

Merge pull request #17602 from ceph/wip-blumig-doc

doc: Updates to bluestore migration doc

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

doc/rados/operations/bluestore-migration: Add instruction for evacuating
host

Signed-off-by: David Galloway <dgallowa@redhat.com>

Merge pull request #17011 from liewegas/wip-doc-bluestore-migration

doc/rados/operations/bluestore-migration: update docs a bit

Merge pull request #17582 from ivancich/wip-fix-mclock-unittests

test/common: unittest_mclock_priority_queue builds with "make" command

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

doc/rados/operations/bluestore-migration: Fix typo in bluestore migration doc

Signed-off-by: David Galloway <dgallowa@redhat.com>

mon: fix dropping mgr metadata for active mgr

drop_standby() was killing it and it was only getting added
back in certain locations.  Instead, make the metadata
drop conditional and only do it in the places we're
really dropping the daemon, not when we're promoting
it to active.

Fixes: http://tracker.ceph.com/issues/21260
Signed-off-by: John Spray <john.spray@redhat.com>

qa/tasks/mgr: regression test for 21260

(aka http://tracker.ceph.com/issues/21260)

Fixes: http://tracker.ceph.com/issues/21260
Signed-off-by: John Spray <john.spray@redhat.com>

qa/suites/krbd: enable xfstests blockdev tests

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

mon/OSDMonitor: don't create pgs if pool was deleted

A pool may be deleted while in the process of creating pgs. If that
happens, we should not queue additional pgs from that pool to be
created.

Fixes: http://tracker.ceph.com/issues/21309
Signed-off-by: Joao Eduardo Luis <joao@suse.de>

Merge pull request #17569 from liewegas/wip-21293

Revert "os/bluestore: pass strict flag to bluestore_blob_use_tracker_t::equal()"

Reviewed-by: xie xingguo <xie.xingguo@zte.com.cn>

Merge pull request #17352 from liewegas/wip-21171

os/bluestore: drop deferred_submit_lock, fix aio leak

Reviewed-by: xie xingguo <xie.xingguo@zte.com.cn>
Reviewed-by: Pan Liu <liupan1111@gmail.com>

encoding: reset optional<> if it is uninitialized

* should reset it, in case we reuse it after initializing it.
* initialize the value of `p` using the C++11 style initializer, so it
  is zero-initialized.
* revert 2a83ef3c which disables a warning of:
 ./include/encoding.h:317:7: warning: 't' may be used uninitialized in
    this function [-Wmaybe-uninitialized]
 where the `t` is the temporary variable for initializing the value of
 `p`.

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

Merge pull request #17593 from shashalu/fix-clean-up-apply_epoch

rgw: drop unused function apply_epoch

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

Merge pull request #17590 from xiaoxichen/minor_doc_fix

doc: fix typo on specify db block device

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

Merge pull request #17573 from jcsp/wip-21292

mgr/restful: cleaner message when not configured

Reviewed-by: Boris Ranto <branto@redhat.com>

rgw: drop unused function apply_epoch

Signed-off-by: Shasha Lu <lu.shasha@eisoo.com>