Merge pull request #42366 from liewegas/mon-down-grace

mon: allow a MON_DOWN grace period after cluster mkfs

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

Merge pull request #42315 from TheJJ/fix-op-duration-calculation

osd: fix op event duration calculation

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

Merge pull request #42385 from tchaikov/wip-crimson-test-alien

test/crimson/test_alien_echo: fixes

Reviewed-by: Chunmei Liu <chunmei.liu@intel.com>

Merge pull request #42351 from SoftIron/softiron-distro

build/qa: Add SoftIron Linux as valid distro in build and test scripts

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

test/crimson/test_alien_echo: msgr->stop() before shutdown

it is required to remove all registered dispatchers before shutting down a
messenger. otherwise assert(dispatchers.empty()) in SocketMessenger::shutdown() fails.

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

test/crimson/test_alien_echo: drop msgr v1 support

crimson msgr dropped the v1 support back in
7a7e869c0a7d96e1e242929755c3af73085d00ac. so there is no point to offer
an option to use v1 protocol.

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

Merge PR #42338 into master

* refs/pull/42338/head:
mon/LogMonitor: tolerate external_log_to in the future

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

mon: allow a MON_DOWN grace period after cluster mkfs

During teuthology tests, the initial cluster bootstrap often starts up
the mon sbut doesn't include all mons in the initial quorum, due to
mon startup misalignment and random delays.  Provide a short grace period
where we will not raise a MON_DOWN alert even though the quorum is not
complete.

Fixes: https://tracker.ceph.com/issues/43584
Signed-off-by: Sage Weil <sage@newdream.net>

Merge PR #42341 into master

* refs/pull/42341/head:
doc/cephadm: operations.rst typo

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

Merge pull request #42200 from cbodley/wip-47781

rgw: RadosBucket::get_bucket_info() updates RGWBucketEnt

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

Merge pull request #36213 from joke-lee/BulkUpload_md5

rgw: fix md5 not match for RGWBulkUploadOp upload when enable rgw com…

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

Merge pull request #42126 from paulreece42/wip-copyobject-urldecode

rgw: url_decode before parsing copysource in copyobject

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

Merge pull request #41579 from trociny/wip-50932

rgw: allow to set ssl options and ciphers for beast frontend

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

Merge pull request #42195 from xrmeng8756/master

rgw: fail as expected when set/delete-bucket-website attempted on a non-exis…

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

Merge pull request #42359 from idryomov/wip-15.2.13-linkage

doc/releases: add linkage for 15.2.13

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

doc/releases: add linkage for 15.2.13

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

Merge PR #42343 into master

* refs/pull/42343/head:
qa/suites/upgrade/pacific-x: do distro before tasks
qa/suites/orch/cephadm/smoke: centos 8.2 -> 8.3

Reviewed-by: Neha Ojha <nojha@redhat.com>

Merge PR #42073 into master

* refs/pull/42073/head:
doc/mgr/nfs: fix 'export apply', pool name
PendingReleaseNotes: document workaround for NFS storage change
qa/tasks/mgr/test_orchestrator_cli: fix test
qa/suites/orch/cephadm/mgr-nfs-upgrade: add test for nfs migration
mgr/cephadm: migrate nfs grace file
mgr/nfs: migrate pre-pacific nfs.ganesha-foo clusters to nfs.foo
doc/cephfs/fs-nfs-exports: document new export apply capabilities
qa/tasks/cephfs/test_nfs: define NFS_POOL_NAME
mgr/nfs: use NFS_POOL_NAME in test_nfs.py
mgr/nfs: test export apply on JSON list
mgr/nfs: add test for ganesha conf apply/import
qa/tasks/cephfs/test_nfs: retry mount a few times
mgr/cephadm: migrate all legacy nfs exports to new .nfs pool
mgr/nfs: adjust cephfs export caps if necessary
python-common: don't accept pool/ns for NFSServiceSpec
mgr/orchestrator: drop rados_config_location ServiceDescription property
mgr/cephadm: move rados_config_location() out of NFSServiceSpec
mgr/nfs: change nfs pool to .nfs
mgr/nfs/export: accept a JSON or ganesha EXPORT config
mgr/nfs: allow 'nfs export apply' to take a list of exports
python-common: remove pool + namespace from NFSServiceSpec
mgr/nfs: used fixed pool + ns
mgr/rook: used fixed pool + ns
mgr/dashboard: use fixed pool + ns
mgr/cephadm: always use fixed pool and namespace
mgr/nfs: adjust test to match pool name
mgr/nfs: always create ganesha pool with well-defined name

Reviewed-by: Varsha Rao <varao@redhat.com>

Merge pull request #42290 from javicacheiro/fix_cephadm_selinux

cephadm: fetch the real selinux status

Reviewed-by: Adam King <adking@redhat.com>
Reviewed-by: Daniel Pivonka <dpivonka@redhat.com>

Merge pull request #42220 from dsavineau/issue_51546

mgr/cephadm/iscsi: check if dashboard is enabled

Reviewed-by: Adam King <adking@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: Sebastian Wagner <sewagner@redhat.com>

doc/mgr/nfs: fix 'export apply', pool name

Signed-off-by: Sage Weil <sage@newdream.net>

Merge pull request #42339 from cbodley/wip-revert-vstart-system

Revert "src/vstart: create rgw user with system flag"

Reviewed-by: Varsha Rao <varao@redhat.com>

Merge pull request #41300 from rhcs-dashboard/force-maintenance-test-error

mgr/dashboard: Fix test_error force maintenance dashboard check

Reviewed-by: Avan Thakkar <athakkar@redhat.com>
Reviewed-by: Ernesto Puerta <epuertat@redhat.com>
Reviewed-by: Nizamudeen A <nia@redhat.com>

qa: Add SoftIron Linux as valid debian-based distro in test scripts

Signed-off-by: Danny Abukalam <danny@softiron.com>

build: Add SoftIron Linux as valid debian-based distro in install-deps

Signed-off-by: Danny Abukalam <danny@softiron.com>

Merge pull request #42334 from tchaikov/wip-crimson-logging

crimson/common/log: print out logger.debug() when log level >=6

Reviewed-by: Mark Nelson <mnelson@readhat.com>
Reviewed-by: Radoslaw Zarzynski <rzarzyns@redhat.com>
Reviewed-by: Chunmei Liu <chunmei.liu@intel.com>

Merge pull request #42337 from rzarzynski/wip-crimson-alienstore-dedicated-subsys

crimson/os: give AlienStore its own debug subsystem.

Reviewed-by: Mark Nelson <mnelson@readhat.com>
Reviewed-by: Chunmei Liu <chunmei.liu@intel.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>

Merge pull request #42251 from adamemerson/wip-robustly-invalidate

rgw: Robustly Invalidate

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

qa/suites/upgrade/pacific-x: do distro before tasks

Otherwise, we isntall new podman at the end, and the
container-selinux-policy package install triggers a bunch of selinux
errors.

Fixes: https://tracker.ceph.com/issues/50151
Signed-off-by: Sage Weil <sage@newdream.net>

PendingReleaseNotes: document workaround for NFS storage change

Signed-off-by: Sage Weil <sage@newdream.net>

qa/tasks/mgr/test_orchestrator_cli: fix test

Signed-off-by: Sage Weil <sage@newdream.net>

qa/suites/orch/cephadm/mgr-nfs-upgrade: add test for nfs migration

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/cephadm: migrate nfs grace file

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/nfs: migrate pre-pacific nfs.ganesha-foo clusters to nfs.foo

In octopus, the service is nfs.ganesha-$id instead of nfs.$id

Signed-off-by: Sage Weil <sage@newdream.net>

doc/cephfs/fs-nfs-exports: document new export apply capabilities

- import list
- import ganesha conf

Signed-off-by: Sage Weil <sage@newdream.net>

qa/tasks/cephfs/test_nfs: define NFS_POOL_NAME

We can't import from mgr_module.py from here, sadly.

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/nfs: use NFS_POOL_NAME in test_nfs.py

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/nfs: test export apply on JSON list

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/nfs: add test for ganesha conf apply/import

Signed-off-by: Sage Weil <sage@newdream.net>

qa/tasks/cephfs/test_nfs: retry mount a few times

It may take a moment for a ganesha to (re)configure itself with a new
export.  If a mount fails, retry a couple times.

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/cephadm: migrate all legacy nfs exports to new .nfs pool

Migrate all past NFS pools, whether they were created by mgr/nfs or by
the dashboard, to the new mgr/nfs .nfs pool.

Since this migrations relies on RADOS being available, we have to be a bit
careful here: we only attempt the migration from serve(), not during
module init.

After the exports are re-imported, we destroy existing ganesha daemons so
that new ones will get recreated.  This ensures the (new) daemons have
cephx keys to access the new pool.

Note that no attempt is made to clean up the old NFS pools.  This is out
of paranoia: if something goes wrong, the old NFS configuration data will
still be there.

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/nfs: adjust cephfs export caps if necessary

If we are importing an old export, we may find that the cephx user
existed but with the wrong caps.  Adjust caps in that case!

Signed-off-by: Sage Weil <sage@newdream.net>

python-common: don't accept pool/ns for NFSServiceSpec

All users are now gone.

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/orchestrator: drop rados_config_location ServiceDescription property

I'm not sure what this was intended to be used for...

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/cephadm: move rados_config_location() out of NFSServiceSpec

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/nfs: change nfs pool to .nfs

This is a new pool that we can migrate all past NFS configuration to,
simplifying the migration process (and also allowing us to pick a
.-prefixed name).

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/nfs/export: accept a JSON or ganesha EXPORT config

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/nfs: allow 'nfs export apply' to take a list of exports

Signed-off-by: Sage Weil <sage@newdream.net>

python-common: remove pool + namespace from NFSServiceSpec

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/nfs: used fixed pool + ns

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/rook: used fixed pool + ns

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/dashboard: use fixed pool + ns

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/cephadm: always use fixed pool and namespace

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/nfs: adjust test to match pool name

Signed-off-by: Sage Weil <sage@newdream.net>

mgr/nfs: always create ganesha pool with well-defined name

Signed-off-by: Sage Weil <sage@newdream.net>

Merge PR #42041 into master

* refs/pull/42041/head:
mgr/restful: ignore min/max_size
test/crush: drop min/max_size refs
qa/workunits/mon/pool_ops: remove test for min/max_size check
qa: scrub a few remaining mentions of ruleset
qa/standalone/mon/osd-*: fix tests
PendingReleaseNotes: note min/max_size removal
mgr/dashboard: remove max/min_size and ruleset
mon/OSDMonitor: fix calls to CrushTester
crush: eliminate min_size and max_size
test/cli/crushtool: reunumber rulesets in test maps
crushtool: require min/max or num-rep for --test
crush: remove last traces of 'ruleset'
test/cli/crushtool: use 'id' instead of 'ruleset' in crush inputs
crushtool: take --min-rep and --max-rep explicitly
crush/CrushTester: drop --ruleset
doc: scrub 'ruleset' from docs
src/erasure-code: rule, not ruleset
mon/OSDMonitor: remove check_crush_rule() callers
mon/OSDMonitor: rule, not ruleset
crushtool: remove check for overlapped ruels
crush/CrushWrapper: get_osd_pool_default_crush_replicated_ruleset -> rule
crush: remove find_rule()
mon/OSDMonitor: use pool's crush rule directly
osd/OSDMap: drop checks for ruleset == ruleid
osd/OSDMap: use pool's crush rule_id directly
mon/PGMap: use pool's crush_rule directly
mon/OSDMonitor: remove crush ruleset->rule rewrite

Reviewed-by: Ernesto Puerta <epuertat@redhat.com>
Reviewed-by: Avan Thakkar <athakkar@redhat.com>

doc/cephadm: operations.rst typo

s/any hosts that is/any host/

Signed-off-by: Zac Dover <zac.dover@gmail.com>

Merge pull request #42332 from rzarzynski/wip-crimson-remotepeeringevent-sends-when-active

crimson/osd: send peering messages only when OSD is active.

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

rgw: Robust notify invalidates on cache timeout

This avoids a potential race condition in which updates are delayed.

Fixes: https://tracker.ceph.com/issues/51674
Signed-off-by: Adam C. Emerson <aemerson@redhat.com>

rgw: Rename REMOVE_OBJ to INVALIDATE_OBJ

Also rename ObjectCache::remove to ObjectCache::invalidate_remove

Since we're depending on these message types/functions having
invalidate semantics but NOT caching a negative result, rename and
leave a comment for clarity.

Fixes: https://tracker.ceph.com/issues/51674
Signed-off-by: Adam C. Emerson <aemerson@redhat.com>

rgw: distribute() takes RGWCacheNotifyInfo

So we don't have to parse the bufferlist back out to find what object
to throw out of the cache.

Fixes: https://tracker.ceph.com/issues/51674
Signed-off-by: Adam C. Emerson <aemerson@redhat.com>

Revert "src/vstart: create rgw user with system flag"

This reverts commit 0d0cd4c6b0e8bbeaa71f54c831d67ef880e58016.

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

mon/LogMonitor: tolerate external_log_to in the future

This can be fast-forwarded by a sync_force or other mon thrashing events.

In any case, if we see it in the future, it is easy to rewind.

Signed-off-by: Sage Weil <sage@newdream.net>

crimson/os: give AlienStore its own debug subsystem.

For the sake of decoupling it from the BlueStore's logging.

Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>

Merge pull request #42327 from zdover23/wip-doc-cephadm-troubleshooting-1-of-x-2021-07-15

doc/cephadm: rewrite troubleshooting 1 of x

Reviewed-by: Sebastian Wagner <sewagner@redhat.com>

qa/suites/orch/cephadm/smoke: centos 8.2 -> 8.3

Signed-off-by: Sage Weil <sage@newdream.net>

Merge pull request #42324 from adamemerson/wip-51661

rgw: Don't segfault on datalog trim

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

mgr/cephadm/iscsi: check if dashboard is enabled

When the mgr dashboard module isn't enabled then the iSCSI service deletion
is stuck and the cluster state goes ERR.
The `ceph dashboard` commands aren't available when the mgr dashboard module
isnt' enabled.

Closes: https://tracker.ceph.com/issues/51546
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>

Fetch the actually running selinux status.

The HostFacts should return the **actual** selinux mode in which the
kernel is running.

The actual mode can be different from the one in the configuration
if the server has not been rebooted or if the mode was changed
after boot using setenforce.

Instead of reading _selinux_path_list we should look at the output of
sestatus or getenforce.

The _selinux_path_list attribute is no longer needed.

Fixes: https://tracker.ceph.com/issues/51632
Signed-off-by: Javier Cacheiro <javier.cacheiro.lopez@cesga.es>

crimson/common/log: print out logger.debug() when log level >=6

less chatty when debug level is 5. this matches the behavior of classic
osd.

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

rgw: Don't segfault on datalog trim

Synchronous (or yielded, basically other-than AioCompletion trim)
would try to dereference the past-the-end iterator if we were trimming
to a point in the most recent generation.

https://tracker.ceph.com/issues/51661
Signed-off-by: Adam C. Emerson <aemerson@redhat.com>

crimson/osd: send peering messages only when OSD is active.

Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>

crimson/osd: assert OSD in active when dispatching peering msgs.

Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>

mgr/dashboard: Fix test_error force maintenance dashboard check

Fixes: https://tracker.ceph.com/issues/50771
Signed-off-by: Nizamudeen A <nia@redhat.com>

Merge pull request #41490 from idryomov/wip-rbd-pwl-ssd-tailp

librbd/cache/pwl/ssd: fix first_free_entry and m_first_free_entry corruption

Reviewed-by: Mahati Chamarthy <mahati.chamarthy@intel.com>
Reviewed-by: Mykola Golub <mgolub@suse.com>

Merge pull request #42145 from hualongfeng/fix_get_context_error

librbd/cache/pwl/ssd: fix use-after-free on C_BlockIORequest

Reviewed-by: Mahati Chamarthy <mahati.chamarthy@intel.com>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>

Merge pull request #42244 from zdover23/wip-doc-cephadm-operations-data-location-and-health-checks-2021-07-08

doc/cephadm: operations: Data location & ...

Reviewed-by: Sebastian Wagner <sewagner@redhat.com>

Merge pull request #42288 from mgfritch/cephadm-test-fixup-mock

cephadm: use CephadmContext rather than MagicMock

Reviewed-by: Sebastian Wagner <sewagner@redhat.com>

Merge pull request #42310 from rzarzynski/wip-crimson-remotepeeringevent-waits-on-active

crimson/osd: remote peering requests wait for OSD activation.

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

rgw: set default ssl options for beast frontend

to 'no_sslv2:no_sslv3:no_tlsv1:no_tlsv1_1'

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

rgw: allow to set ssl options and ciphers for beast frontend

Two new conf keys are added for "beast" framework:

- ssl_options: a colon separated list of ssl context options,
  documented in boost's ssl::context_base;

- ssl_ciphers: a colon separated list of ciphers, documented
  in openssl's ciphers(1) manual.

Example:

  rgw frontends = beast ...  ssl_options=default_workarounds:no_tlsv1:no_tlsv1_1 ssl_ciphers=HIGH:!aNULL:!MD5

Fixes: https://tracker.ceph.com/issues/50932
Signed-off-by: Mykola Golub <mgolub@suse.com>

librbd/cache/pwl/ssd: fix use-after-free on C_BlockIORequest

In setup_schedule_append() function, its first expression
will cause the req to be deleted, and subsequent use of
the variable req becomes an illegal operation. And due to
delete, rep->m_image_ctx will be empty, so it lead to
segfault in AbstractWriteLog::get_context().
So pass the `req` into `schedule_append()` function.

Fixes: https://tracker.ceph.com/issues/50951
Signed-off-by: Hualong Feng <hualong.feng@intel.com>

crimson/osd: move the wait-for-active logic into OSDState.

Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>

Merge pull request #42314 from rzarzynski/wip-crimson-single-do_peering_event

crimson/osd: consider epoch_sent when ignoring old peering event.

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

Merge pull request #42271 from xxhdx1985126/wip-seastore-writev

crimson/os/seastore: use writev to avoid extra memory copy

Reviewed-by: Yingxin Cheng <yingxin.cheng@intel.com>
Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>

Merge pull request #42311 from rzarzynski/wip-crimson-fix-premature-activation

crimson/osd: prevent premature OSD activation.

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

Merge pull request #42202 from cyx1231st/wip-seastore-metrics

crimson/os/seastore: implement label-based profiling with metrics

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

doc/cephadm: rewrite troubleshooting 1 of x

This PR improves the readability and format
of the troubleshooting.rst file. This also
makes a change to the markdown of one of the
sub-subsections so that it is made of tildes
(~) instead of carets (^), because that's
the RST standard.

Signed-off-by: Zac Dover <zac.dover@gmail.com>

crimson/os/seastore: use writev to avoid extra memory copy

Signed-off-by: Xuehan Xu <xxhdx1985126@gmail.com>

crimson/os/seastore/cache: improve logs for transaction conflict/reset

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os/seastore/cache: count invalidated transactions

Labeled by source and extent-type.

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os/seastore/cache: count committed transactions

Labeled by source.

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os/seastore/cache: count created transactions

Labeled by source.

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os/seastore: introduce src_t to classify transactions

* INIT - READ            => INIT
* INIT - MUTATE          => INIT
* SEASTORE/TEST - READ   => READ
* SEASTORE/TEST - MUTATE => MUTATE
* CLEANER - READ         => CLEANER
* CLEANER - MUTATE       => CLEANER

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os/seastore: consolidate the static member CachedExtent::TYPE

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

Merge pull request #42279 from myoungwon/wip-fix-51627

osd: fix to recover adjacent clone when set_chunk is called

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

Merge pull request #42285 from orozery/append-frame-claim-append

msg/async/ProtocolV2: optimize append_frame

Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>

Merge pull request #42218 from ifed01/wip-ifed-compact-after-upgrade

os/bluestore: compact db after bulk omap naming upgrade.

Reviewed-by: Adam Kupczyk <akupczyk@redhat.com>

Merge pull request #42197 from ifed01/wip-ifed-fix-spurious-read-errors-warn

os/bluestore: respect bluestore_warn_on_spurious_read_errors setting

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

Merge pull request #42291 from dsavineau/issue_51620

cephadm: ensure sysctl_dir exist

Reviewed-by: Sebastian Wagner <sewagner@redhat.com>
Reviewed-by: Michael Fritch <mfritch@suse.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>