osd/OSD: osd_fast_shutdown_notify_mon not quite right

When osd_fast_shutdown and osd_fast_shutdown_notify_mon set as true, OSD marked as Down
it should be marked as Dead,

Fixed: https://tracker.ceph.com/issues/53327

Signed-off-by: Nitzan Mordechai <nmordech@redhat.com>
nd

nd

osd: make osd_fast_shutdown_notify_mon option true by default

osd_fast_shutdown_notify_mon option is false by default. So users suffer
from error log flood, slow ops, and the long I/O timeouts on voluntary OS
shutdown before they are aware of the existence of this option. Let's
make this option true by default.

Fixes: https://tracker.ceph.com/issues/53328
Signed-off-by: Satoru Takeuchi <satoru.takeuchi@gmail.com>

Merge pull request #45253 from tchaikov/wip-crimson-is-future

crimson/common: specialize seastar::is_future<> for interruptible_fut…

Reviewed-by: Xuehan Xu <xxhdx1985126@gmail.com>
Reviewed-by: Radoslaw Zarzynski <rzarzyns@redhat.com>

Merge pull request #45258 from tchaikov/wip-cmake-disable-dpdk-warnings

cmake: pass RTE_DEVEL_BUILD=n when building dpdk

Reviewed-by: Ilya Dryomov <idryomov@redhat.com>

Merge pull request #45255 from kamoltat/wip-ksirivad-add-jenkins-window

.github/pull_request_template.md: added test windows command

Reviewed-by: Neha Ojha <nojha@redhat.com>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>

cmake: pass RTE_DEVEL_BUILD=n when building dpdk

ceph is still using the Makefile based building system for building
DPDK. and DPDK enables -Werror if RTE_DEVEL_BUILD is 'y' which is
enabled by default when the dpdk is built from a git repo.

but newer GCC is more picky than the older versions, to prevent
the possible FTBFS when we switch to newer GCC for building old
branches whose dpdk submodule might be include the changes addressing
those warnings. let's just disable this option.

the only effect of this option is to add -Werror to CFLAGS. but
the building warnings from DPDK is not our focus when developing
Ceph in the most cases. so it should be fine.

see also
https://github.com/ceph/dpdk/blob/eac901ce29be559b1bb5c5da33fe2bf5c0b4bfd6/doc/build-sdk-quick.txt#L18

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

.github/pull_request_template.md: added test windows command

Added `jenkins test windows` to the

Show available Jenkins Commands section

Signed-off-by: Kamoltat <ksirivad@redhat.com>

test/crimson: use std::for_each() for applying plain function

let's just use std::for_each() for applying function not returning future
to a range, otherwise we'd have

../src/test/crimson/seastore/test_transaction_manager.cc:950:5: error: no matching function for call to 'parallel_for_each'
    seastar::parallel_for_each(
    ^~~~~~~~~~~~~~~~~~~~~~~~~~
../src/seastar/include/seastar/core/loop.hh:559:1: note: candidate template ignored: constraints not satisfied [with Iterator = boost::iterators::counting_iterator<unsigned int, boost::use_default, boost::use_default>, Sentinel = boost::iterators::counting_iterator<unsigned int, boost::use_default, boost::use_default>, Func = (lambda at ../src/test/crimson/seastore/test_transaction_manager.cc:953:7)]
parallel_for_each(Iterator begin, Sentinel end, Func&& func) noexcept {
^
../src/seastar/include/seastar/core/loop.hh:553:78: note: because type constraint 'std::same_as<void, future<> >' was not satisfied:

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

crimson/common: specialize seastar::is_future<> for interruptible_future<>

this ensures that seastar helper functions are able to accept
interruptible_future<> in place of vanilla seastar::future<>.

for instance, if concepts is enabled, seastar::when_all_succeed()
has a set of constraits which include:

is_future<std::remove_reference_t<decltype(*i)>>::value

where i is the iterator passed to this function. so, to ensure
seastar::when_all_succeed() works with interruptible_future,
we need to teach the is_future<> to accept interruptible_future
as well.

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

Merge pull request #45140 from dvanders/snaptrim_fix

osd: require osd_pg_max_concurrent_snap_trims > 0

Merge pull request #45239 from ronen-fr/wip-rf-54458-snaps

test: osd-scrub-snaps.sh: fix expected 'missing snaps' log string

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

Merge pull request #45042 from dvanders/report_leader

Monitor: forward report command to leader

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

Merge pull request #45035 from ssdohammer-sl/bug-fix-TierFlushDuringFlush

test: fix TierFlushDuringFlush to wait until dedup_tier is set on base pool

Merge pull request #44439 from myoungwon/fix-reference-leak-dedup-tier

osd: fix ObjectContextRef reference leak

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

Merge pull request #45222 from tchaikov/wip-cmake-cleanup

cmake: use string(JOIN ..) to compose URL for boost download

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

Merge pull request #45219 from idryomov/wip-ceph-config-get-doc

doc: document "ceph config get <who> <option>"

Reviewed-by: Anthony D'Atri <anthony.datri@gmail.com>

test: osd-scrub-snaps.sh: fix expected 'missing snaps' log string

Fix the expected log message to match the scrub code, by removing
the redundant part.

Fixes: https://tracker.ceph.com/issues/54458
Signed-off-by: Ronen Friedman <rfriedma@redhat.com>

Merge pull request #45090 from liu-chunmei/seastore-fix-sparse-read

crimson/seastore: fix object not exists when do fiemap

Reviewed-by: Samuel Just <sam.just@inktank.com>

Merge pull request #45209 from zdover23/wip-doc-2022-03-01-hardware-recs-alpha-links

doc/start: alphabetize hardware-recs links

Reviewed-by: Anthony D Atri <anthony.datri@gmail.com>

crimson: fix object not exist when do fiemap

Signed-off-by: chunmei-liu <chunmei.liu@intel.com>

crimson/seastore: make fiemap return errorator future

Signed-off-by: chunmei-liu <chunmei.liu@intel.com>

Merge pull request #45216 from guits/fix-54441

ceph-volume: fix generic activate

Merge pull request #45207 from pritha-srivastava/wip-rgw-data-sync-run

rgw: adding OPT_DATA_SYNC_RUN to gc_ops_list so that gc gets

Reviewed-by: Adam C. Emerson <aemerson@redhat.com>
Reviewed-by: Casey Bodley <cbodley@redhat.com>
Reviewed-by: Matt Benjamin <mbenjami@redhat.com>

Merge pull request #44656 from NitzanMordhai/fix_44092_config_get

mon/ConfigMonitor: fix config get key with whitespace

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

Merge pull request #44479 from leeq2016/leeq2016_0106

common: Use thread-local pointer variables to save the *shard

Reviewed-by: Ronen Friedman <rfriedma@redhat.com>

Merge pull request #44425 from sqiu110/fix-ENABLE_COVERAGE

cmake: link ceph-common against gcov if coverage is enabled

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

Merge pull request #43058 from k0ste/feature_52512

mgr/prometheus: Added `avail_raw` field for Pools DF Prometheus mgr module

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

cmake: use string(JOIN ..) to compose URL for boost download

* since the minimum required cmake version is now 3.16, there is no
  need to check for cmake version for using multiple URLs for
  downloading external project.
* use string(JOIN ..) to compose URL option for downloading boost

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

doc: document "ceph config get <who> <option>"

... and note the confusing asymmetry with "ceph config set <who>
<option> <value>".

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

Merge pull request #44923 from rhcs-dashboard/box-host-add

cephadm/box: default add hosts

Reviewed-by: Ernesto Puerta <epuertat@redhat.com>
Reviewed-by: Nizamudeen A <nia@redhat.com>
Reviewed-by: Pere Diaz Bou <pdiazbou@redhat.com>

Merge pull request #45080 from ceph/epuertat-patch-1

doc: fix ceph-volume ref

Merge pull request #45082 from ceph/epuertat-patch-3

doc: fix titles

rgw: adding OPT_DATA_SYNC_RUN to gc_ops_list so that gc gets
initialized for this command.

Fixes: https://tracker.ceph.com/issues/54433
Signed-off-by: Pritha Srivastava <prsrivas@redhat.com>

Merge pull request #45198 from tchaikov/wip-cxx-20

changes to compile with latest seastar and C++20

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

Merge pull request #45204 from liu-chunmei/seastore-fix-bad_alloc

crimson/seastore: fix bad_alloc

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

ceph-volume: fix generic activate

afd8be7eac5e996c3bd07656601a4534053e2516 broke it.
It has dropped`block_wal` and `block_db` from
`ceph_volume.devices.raw.activate.activate_bluestore` but
`activate.main.Activate.main` still passes those arguments when
calling `RAWActivate([]).activate()`

Fixes: https://tracker.ceph.com/issues/54441
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

Merge pull request #45215 from neha-ojha/wip-cf-add

doc/foundation.rst: Add ArvanCloud to the Foundation

Reviewed-by: Dan van der Ster <daniel.vanderster@cern.ch>

doc/foundation.rst: Add ArvanCloud to the Foundation

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

mgr/prometheus: added `avail_raw` field for Pools DF Prometheus mgr module

Fixes: https://tracker.ceph.com/issues/52512
Signed-off-by: Konstantin Shalygin <k0ste@k0ste.ru>

seastar: bump up seastar submodule to recent master

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

crimson/common/interruptible_future: ensure seastar::AllAreFutures<> is happy

if compiled with C++20, C++ concept is enabled. to ensure the
constraints required by when_all_impl() is fulfilled, we need to
specialize is_tuple_of_futures<> for interruptible_future.

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

include/xlist: define iterator without std::iterator<>

std::iterator<> is deprecated in C++17, so let's just
define the required traits directly.

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

crimson/os/seastore: use iter_cend()

to avoid the pain of fixing the FTBFS with C++20 like

/var/ssd/ceph/src/crimson/os/seastore/omap_manager/btree/omap_btree_node_impl.cc:141:19: error: use of overloaded operator '!=' is ambiguous (with operand types 'crimson::os::seastore::omap_manager::StringKVInnerN$
  assert(child_pt != iter_end());
         ~~~~~~~~ ^  ~~~~~~~~~~
/usr/include/assert.h:93:27: note: expanded from macro 'assert'
     (static_cast <bool> (expr)                                         \
                          ^~~~
/var/ssd/ceph/src/crimson/os/seastore/omap_manager/btree/string_kv_node_layout.h:380:10: note: candidate function
    bool operator==(const iter_t &rhs) const {
         ^
/var/ssd/ceph/src/crimson/os/seastore/omap_manager/btree/string_kv_node_layout.h:380:10: note: candidate function (with reversed parameter order)

also, to be explicit is advisable.

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

crimson: include <seastar/core/metrics.hh>

after the restructure of seastar source tree, we have to include
metrics.hh for accessing the metrics types.

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

crimson/admin/osd_admin: avoid using _format()

see also 3fd91d7e6d9315244aeff02070721ef4a021b0b3

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

crimson/osd: use reference when structured binding

to silence warning from clang13 with -std=c++20, like

ceph/src/crimson/osd/replicated_recovery_backend.cc:1098:21: warning: loop variable '[off, len]' creates a copy from type 'const std::pair<const unsigned long, unsigned long>' [-Wrange-loop-construct]
    for (const auto [off, len] : extents) {
                    ^
ceph/src/crimson/osd/replicated_recovery_backend.cc:1098:10: note: use reference type 'const std::pair<const unsigned long, unsigned long> &' to prevent copying
    for (const auto [off, len] : extents) {
         ^~~~~~~~~~~~~~~~~~~~~~~
                    &

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

include/rados: define iterator without std::iterator<>

std::iterator<> is deprecated in C++17, so let's just
define the required traits directly.

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

crimson/os: define iterator without std::iterator<>

std::iterator<> is deprecated in C++17, so let's just
define the required traits directly.

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

crimson/os/seastore: remove default ctor for delta_t

otherwise the aggregate initialization for this type would fail to
work on C++20, as the user-declared ctor is defined for this struct.

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

crimson/os: stop using seastar::metrics::label

seastar::metrics::label was dropped in a recent change. quote from
the commit message which removed this type:

>     In practice, it's an over-kill and just add information in the reporting
>    layer.

see also da395f31989a1ee333e489ffb251fc1152f12f0a in seastar repo

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

include/buffer: include <memory>

to address following FTBFS:

/usr/bin/ccache /usr/bin/clang++-13 -DBOOST_ALL_NO_LIB -DBOOST_ASIO_DISABLE_CONCEPTS -DBOOST_ASIO_DISABLE_THREAD_KEYWORD_EXTENSION -DBOOST_ASIO_USE_TS_EXECUTOR_AS_DEFAULT -DBOOST_PROGRAM_OPTIONS_DYN_LINK -DBOOST_T$
In file included from /var/ssd/ceph/src/crimson/os/seastore/seastore_types.cc:4:
In file included from /var/ssd/ceph/src/crimson/os/seastore/seastore_types.h:14:
In file included from /var/ssd/ceph/src/include/denc.h:47:
/var/ssd/ceph/src/include/buffer.h:98:37: error: no template named 'unique_ptr' in namespace 'std'; did you mean 'boost::movelib::unique_ptr'?
struct unique_leakable_ptr : public std::unique_ptr<T, ceph::nop_delete<T>> {
                                    ^~~~~~~~~~~~~~~
                                    boost::movelib::unique_ptr
/opt/ceph/include/boost/move/unique_ptr.hpp:354:7: note: 'boost::movelib::unique_ptr' declared here
class unique_ptr
      ^

Signed-off-by: Kefu Chai <tchaikov@gmail.com>

Merge pull request #45047 from ceph/wip-yuriw-release-15.2.16-master

doc: 15.2.16 Release Notes

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

doc/start: alphabetize hardware-recs links

This PR alphabetizes the internal and external
references at the bottom of the Restructured
Text.

(This is really just a PR made so that I can
show how to make a documentation pull request.)

(In a video.)

(For people who prefer step-by-step instructions
in video form.)

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

Merge pull request #44050 from ronen-fr/wip-rf-digest-update

osd/scrub: ignoring unsolicited DigestUpdate events

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

crimson/seastore: fix bad_alloc

Signed-off-by: chunmei-liu <chunmei.liu@intel.com>
Signed-off-by: Samuel Just <sjust@redhat.com>

Merge pull request #45139 from zdover23/wip-doc-2022-02-24-hardware-recs-adatri-recs

doc/start: include A. D'Atri's hardware-recs recs

Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Anthony D'Atri <anthony.datri@gmail.com>

Merge pull request #45191 from neesingh-rh/cephfs-top_doc_fix

doc/cephfs: Typographical error.

doc/releases: couple of notable changes for core

also fix release date for v15.2.16

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

Merge pull request #45136 from adk3798/extra-all

mgr/cephadm: extend extra_container_args to other service types

Reviewed-by: Michael Fritch <mfritch@suse.com>

Merge pull request #45166 from pritha-srivastava/wip-rgw-bucket-sync-run

rgw: add OPT_BUCKET_SYNC_RUN to gc_ops_list, so that

Reviewed-by: Adam C. Emerson <aemerson@redhat.com>
Reviewed-by: Casey Bodley <cbodley@redhat.com>

doc/cephfs: Typographical error.

Signed-off-by: Neeraj Pratap Singh <neesingh@redhat.com>

doc/start: include A. D'Atri's hardware-recs recs

This PR restores material about partition alignment
and material about separating OS and OSD data that
was removed in an earlier rewrite. The restoration
of this information was requested by Anthony D'Atri in
https://github.com/ceph/ceph/pull/45123/

This PR also includes several refinements to the language
that could not be made to this text until now, owing to my
(Zac's) ignorance and illiteracy.

I call upon Mark Nelson (and anyone else with sufficient
command of the current state of storage technology) to advise
me on whether the Ceph Foundation feels comfortable in the year
2022 referring to QLC as an emerging technology.

Signed-off-by: Zac Dover <zac.dover@gmail.com>
(squash) more notes and revisions

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

cephadm/box: default add hosts

Signed-off-by: Pere Diaz Bou <pdiazbou@redhat.com>

rgw: add OPT_BUCKET_SYNC_RUN to gc_ops_list, so that
gc is initialised and send_chain does not crash.

Also deleting objects inline in case gc is uninitialised.

Fixes: https://tracker.ceph.com/issues/54417
Signed-off-by: Pritha Srivastava <prsrivas@redhat.com>

osd/scrub: cluster-logging unsolicited DigestUpdate events

... and not just to the OSD's log

Signed-off-by: Ronen Friedman <rfriedma@redhat.com>

Merge pull request #44941 from ronen-fr/wip-rf-scrubBEv1_2

osd/scrub: limiting scrubber-backend external interfaces

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

Merge pull request #43957 from petrutlucian94/windows_ipv6

msg: Fix Windows IPv6 support

Merge pull request #44870 from phlogistonjohn/jjm-nfs-cleanups2

mgr/nfs: support managing exports without orchestration enabled

Reviewed-by: Ramana Raja <rraja@redhat.com>

Merge pull request #44845 from phlogistonjohn/jjm-nfs-cleanups

mgr/nfs: various minor cleanups

Reviewed-by: Adam King <adking@redhat.com>
Reviewed-by: Michael Fritch <mfritch@suse.com>
Reviewed-by: Ramana Raja <rraja@redhat.com>

Merge pull request #45113 from adk3798/minimize-image

cephadm: still set container_image when --no-assimilate-config is provided in bootstrap

Reviewed-by: John Mulligan <jmulligan@redhat.com>
Reviewed-by: Michael Fritch <mfritch@suse.com>

Merge pull request #45127 from rkachach/fix_issue_54026_part1

mgr/cephadm: Adding python natsort module

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

Merge pull request #45132 from melissa-kun-li/asyncssh_error_messages

mgr/cephadm: reduce log level for asyncssh error messages

Reviewed-by: Michael Fritch <mfritch@suse.com>

mgr/cephadm: extend extra_container_args to other service types

Otherwise, without this change, this can only be used for mgr,
mon and crash (daemons without their own service spec class)

Fixes: https://tracker.ceph.com/issues/54390
Signed-off-by: Adam King <adking@redhat.com>

Merge pull request #45041 from cyx1231st/wip-crimson-cleanup-epm-segment-writer

crimson/os/seastore: introduce SegmentAllocator for Journal and ExtentPlacementManager

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

Merge pull request #45109 from cbodley/wip-rgw-dbstore-strings

rgw/dbstore: use string_view for string constants

Reviewed-by: Soumya Koduri <skoduri@redhat.com>

Merge pull request #45126 from guits/fix-bug-cv-bz2056719

ceph-volume: abort when passed devices have partitions

osd/scrub: fix some scrubber-backend formatting

no code changes.

Signed-off-by: Ronen Friedman <rfriedma@redhat.com>

osd/scrub: cleanup scrubber-be's usage of the size-of-disk service

Limit the amount of scrubber code that directly calls the
PG-Backend services - in this case, the logical-size to
on-disk-size translation.

Signed-off-by: Ronen Friedman <rfriedma@redhat.com>

osd/scrub: clean-up scrubber-be interfaces

Esp - the shallow/deep-error counters.

Signed-off-by: Ronen Friedman <rfriedma@redhat.com>

osd/scrub: make the scrubber-be the sole owner of the authoritative set

simplifying the interfaces used by the scrubber backend.

Signed-off-by: Ronen Friedman <rfriedma@redhat.com>

osd/scrub: remove I/O (store) direct access from the Scrubber Backend

Target: improving Scrubber BE testability by reducing the number and
complexity of the external APIs it uses.

(note - first step of many)

Signed-off-by: Ronen Friedman <rfriedma@redhat.com>

crimson/os/seastore/epm: integrate SegmentAllocator with Writer

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

crimson/os/seastore: introduce SegmentAllocator and integrate with Journal

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

crimson/os/seastore/journal: cleanup, move commit_to to RecordSubmitter

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

crimson/os/seastore/epm: replace condition_variable by a shared_promise

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

crimson/os/seastore/epm: simplify gating writes for Writer

Dropped open_segment_wrapper_t.

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

crimson/os/seastore/segment_cleaner: drop unnecessary future from get_segment()

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

osd: require osd_pg_max_concurrent_snap_trims > 0

If osd_pg_max_concurrent_snap_trims is zero, we mistakenly clear
the snaptrim queue. Require it to be > 0.

Fixes: https://tracker.ceph.com/issues/54396
Signed-off-by: Dan van der Ster <daniel.vanderster@cern.ch>

Merge pull request #45121 from yaarith/fix-wear-level-empty-page

mgr/devicehealth: skip null pages when extracting wear level

Reviewed-by: Laura Flores <lflores@redhat.com>

Merge pull request #45118 from sseshasa/wip-fix-osd-bench-fill-non-zeros

osd: Write non-zero data as part of osd benchmark test.

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

Merge pull request #45010 from cyx1231st/wip-seastore-types-cleanup

crimson/os/seastore/seastore_types: cleanups

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

test: fix TierFlushDuringFlush to wait until dedup_tier is set on base pool

When start_dedup() is called while the base pool is not set the dedup_tier,
it is not possible to know the target pool of the chunk object.

1. User set the dedup_tier on a base pool by mon_command().
2. User issues tier_flush on the object which has a manifest (base pool)
  before the dedup_tier is applied on the base pool.
3. OSD calls start_dedup() to flush the chunk objects to chunk pool.
4. OSD calls get_dedup_tier() to get the chunk pool of the base pool,
  but it is not possible to know the chunk pool.
5. get_dedup_tier() returns 0 because it is not applied on the base pool yet.
6. This makes refcount_manifest() lost it's way to chunk pool.

To prevent this issue, start_dedup() has to be called after dedup_tier is set
on the base pool. To do so, this commit prohibits getting chunk pool id if
dedup_tier is not set.

Fixes: http://tracker.ceph.com/issues/53855
Signed-off-by: Sungmin Lee <sung_min.lee@samsung.com>

doc/mgr/nfs: document that nfs exports related mgr call requirements

A recent change in the mgr/nfs module should enable the functioning
of export management commands/API calls as long as the rados namespaces
and objects have been already established. Document this fact, noting
that now only the `ceph nfs cluster ...` calls *require* an
orchestration module.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/nfs: support managing exports without orchestration enabled

This change allows the `ceph nfs export ...` commands to function
without the entire mgr/nfs subsystem requiring orchestration to be
enabled.  When there's no orchestration available, the code falls back
to examining the namespaces in the ".nfs" rados pool to determine what
cluster_id values are valid.

This change does not add support for creating the rados objects and
namespace needed to manage a nfs cluster. As discussed with the
orchestration group on 2022-01-22, rook does not need the mgr module to
establish the namespace. So, for now, we'll defer the work needed to
create the namespace/objects when orchestration is disabled.

Fixes: https://tracker.ceph.com/issues/54043
Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/nfs: fix typo in error message

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/nfs: add unit test for normalize_path

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/nfs: change method format_path to function normalize_path

This function was not using self and thus has no need to be a method.
While we're at it, rename it to normalize_path because that's what
it is doing.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/nfs: clean up rados object naming code

The naming of rados objects used to store the nfs config was spread
all over the code, including inline f-strings, not-static methods,
etc.
This change unifies the naming by putting constant string prefixes
and name generating functions into the utils.py file.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/nfs: make _check_rados_notify a function

This was previously a staticmethod. This static method was only used by
NFSRados object. Staticmethods are nearly always better implemented as
functions, which is done so here.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/nfs: limit dependency of NFSRados object

Previously, the NFSRados object accepted the "Module" as the
first argument but only used the rados attribute (type rados.Rados).
It's better to limit the scope of types when reasonably possible
so we can see what the true dependencies are. So we restrict
NFSRados to accepting a rados.Rados as the argument.

Signed-off-by: John Mulligan <jmulligan@redhat.com>