Merge pull request #42615 from neha-ojha/wip-51830-pacific

pacific: common/options: Set osd_client_message_cap to 256.

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

Merge pull request #42300 from callithea/wip-51241-pacific

pacific: debian/control: ceph-mgr-modules-core does not Recommend ceph-mgr-roo…

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

Merge pull request #42578 from batrick/i51983

pacific: mon/MDSMonitor: do not pointlessly kill standbys that are incompatible with current CompatSet

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

Merge pull request #42096 from varshar16/wip-nfs-backport-june

pacific: nfs backport June

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

Merge pull request #42909 from gregsfortytwo/wip-51290-pacific

pacific: mon: Sanely set the default CRUSH rule when creating pools in stretch…

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

mon: Sanely set the default CRUSH rule when creating pools in stretch mode

If we get a pool create request while in stretch mode that does not explicitly
specify a crush rule, look at the stretch-mode pools and their rules, and
select the most common one.

Also update set_up_stretch_mode.sh to add a few more rules that let me test
this locally.

Fixes: https://tracker.ceph.com/issues/51270
Signed-off-by: Greg Farnum <gfarnum@redhat.com>
(cherry picked from commit 4f9fdca136bde10479fc173c9bcf4288ea960ace)

Merge pull request #42632 from cfsnyder/wip-51471-pacific

pacific: rgw/sts: correcting the evaluation of session policies

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

Merge pull request #42868 from sebastian-philipp/pacific-backport-42852-42863

pacific: qa/cephadm: centos_8.x_container_tools_3.0.yaml

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

Merge pull request #42850 from ceph/wip-81356-pacific

pacific: run-make-check.sh: Increase failure output log size

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

qa/distro: Add centos_8.2_container_tools_3.0.yaml

Let's avoid latest kubic stable

Fixes: https://tracker.ceph.com/issues/52279
Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit e436483c77adf7f876533c25a11c14d19a2d32a4)

Conflicts:
qa/suites/orch/cephadm/dashboard/0-distro/centos_8.2_kubic_stable.yaml
qa/suites/orch/cephadm/smoke/distro/centos_8.2_kubic_stable.yaml

qa/cephadm: centos_8.3_container_tools_3.0.yaml

Let's avoid latest stable

Fixes: https://tracker.ceph.com/issues/52279
Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit c6bd968eff96666da7e321fb5787c82dc66ddbb7)

Merge pull request #42211 from ideepika/wip-51549-pacific

pacific: monitoring/grafana/build/Makefile: revamp for arm64 builds, pushes to docker and quay, jenkins

Reviewed-by: Alfonso Martínez <almartin@redhat.com>
Reviewed-by: Aashish Sharma <aasharma@redhat.com>

run-make-check.sh: Increase failure output log size

Fixes: https://tracker.ceph.com/issues/43566
Signed-off-by: David Galloway <dgallowa@redhat.com>
(cherry picked from commit e1b55f6878cfe847112959885d2fc44e91f3c44f)

Merge pull request #42736 from sebastian-philipp/pacific-backport-41854-42108-42134-42017-41906-42032-41829-41938-42177-42092-

pacific: cephadm: August batch 1

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

Merge pull request #42753 from guits/wip-52133-pacific

pacific: ceph-volume: work around phantom atari partitions

Merge pull request #42829 from tchaikov/pacific-pr-42811

pacific: mgr/dashboard/api: set a UTF-8 locale when running pip

Reviewed-by: Avan Thakkar <athakkar@redhat.com>
Reviewed-by: Alfonso Martínez <almartin@redhat.com>

mgr/dashboard/api: set a UTF-8 locale when running pip

ansible-core started to include files whose filenames are encoded in
non-ascii characters, so we have to use a more capable encoding for the
locale in order to install this package. otherwise we'd have following
error:

Collecting ansible-core<2.12,>=2.11.3
  Using cached ansible-core-2.11.4.tar.gz (6.8 MB)
ERROR: Exception:

Traceback (most recent call last):
  File "/tmp/tmp.fX76ASIrch/venv/lib/python3.8/site-packages/pip/_internal/cli/base_command.py", line 173, in _main
    status = self.run(options, args)
...
  File "/tmp/tmp.fX76ASIrch/venv/lib/python3.8/site-packages/pip/_internal/utils/unpacking.py", line 226, in untar_file
    with open(path, "wb") as destfp:
UnicodeEncodeError: 'latin-1' codec can't encode characters in position 137-140: ordinal not in range(256)

Signed-off-by: Kefu Chai <kchai@redhat.com>
(cherry picked from commit 05e4145856bb5ed19ecc879f2e50b5a88cb2045e)

rgw/sts: correcting the evaluation of session policies
passed in with AssumeRoleWithWebIdentity.

Session Policies are used to restrict the permissions
granted by identity-based (Role's permission policy
and resource-policy (bucket policy) in some cases.

Fixes: https://tracker.ceph.com/issues/51019
Signed-off-by: Pritha Srivastava <prsrivas@redhat.com>
(cherry picked from commit bd611d451aaaba65c6c1b7a91ec4486142c90522)

Conflicts:
src/rgw/rgw_op.cc

Cherry-pick notes:
- conflict due to rgw::sal::RGWObject renaming to rgw::sal::Object after Pacific

Merge pull request #42819 from rhcs-dashboard/wip-52293-pacific

pacific: mgr/dashboard: disable create snapshot with subvolumes

Merge pull request #42817 from rhcs-dashboard/wip-52294-pacific

pacific: mgr/dashboard: Refresh button on the iscsi targets page

Merge pull request #42812 from aaSharma14/wip-52135-pacific

pacific: mgr/dashboard: monitoring: replace Grafana JSON with Grafonnet based code

Merge pull request #42702 from joscollin/wip-52084-pacific

pacific: pybind/mgr/stats: validate cmdtag

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

mgr/dashboard: disable create snapshot with subvolumes

Signed-off-by: Pere Diaz Bou <pdiazbou@redhat.com>
(cherry picked from commit b38259b62950ed2b15175935af7de00059926f3f)

mgr/dashboard: Refresh button on the iscsi targets page

Added a refresh button on the iscsi targets page. I've noticed that the
auto reload causes some load on the backend. So I disabled the auto
reload and went for the same approach as we have on rgw. A yellow
warning color on refresh btn to let the user know to manually refresh
that whenever needed.

Fixes: https://tracker.ceph.com/issues/52102
Signed-off-by: Nizamudeen A <nia@redhat.com>
(cherry picked from commit 8f23a587a8ccc686b1e3acbfa587fcab76bee45d)

mgr/dashboard: fix grafonnet build error

This PR tends to fix the issue caused by #42194

Fixes:https://tracker.ceph.com/issues/52238
Signed-off-by: Aashish Sharma <aasharma@redhat.com>
(cherry picked from commit 4907c78bb732b613ae223d3a7d1a046da457d1cc)

Conflicts:
ceph.spec.in (added jsonnet package)
monitoring/grafana/dashboards/CMakeLists.txt (added grafana jsons code)
(cherry picked from commit 570160e291b9029c7998d4dd4fee33327da91312)

Merge pull request #42788 from rhcs-dashboard/wip-52248-pacific

pacific: mgr/dashboard: tox.ini: delete useless env. 'apidocs'

Merge pull request #42789 from rhcs-dashboard/wip-52250-pacific

pacific: mgr/dashboard: cephadm e2e start script: add --expanded option

Merge pull request #42810 from rhcs-dashboard/wip-52291-pacific

pacific: mgr/dashboard: fix rename inventory to disks

mgr/dashboard: monitoring: replace Grafana JSON with Grafonnet based Code

This PR intends to add grafonnet to generate grafana JSON files

Fixes: https://tracker.ceph.com/issues/45184
Signed-off-by: Aashish Sharma <aasharma@redhat.com>
(cherry picked from commit e9bd94515f865b6c8865210b6329c11f3ebe9d23)

Conflicts:
debian/control (added jsonnet package)
run-make-check.sh (switched ON the WITH_GRAFANA bool for make check)

mgr/dashboard: fix rename inventory to disks

Fix Inventory breadcrumbs

Fixes: https://tracker.ceph.com/issues/50314
Signed-off-by: Navin Barnwal <knbarnwal@gmail.com>
(cherry picked from commit 23f44e894462e5658f8bcb85899df009fe379379)

mgr/dashboard: fix rename inventory to disks

Fix Orchestrator e2e tests

Fixes: https://tracker.ceph.com/issues/50314
Signed-off-by: Navin Barnwal <knbarnwal@gmail.com>
(cherry picked from commit 15b3665f27d7ceb2de74796a365d4bd19d716ff6)

mgr/dashboard: fix rename inventory to disks
Fix failure check in Host detail component

Fixes: https://tracker.ceph.com/issues/50314
Signed-off-by: Navin Barnwal <knbarnwal@gmail.com>
(cherry picked from commit 356c3c3bbf18a1dbce05d5b7b190df5ef69112e0)

mgr/dashboard: fix rename inventory to disks
Inventory -> Physical Disks changed in total 3 locations

Fixes: https://tracker.ceph.com/issues/50314
Signed-off-by: Navin Barnwal <knbarnwal@gmail.com>
(cherry picked from commit f2aa58a5faaa8aa47c4245fc557a584a97496e67)

mgr/dashboard: fix rename inventory to disks

Fixes: https://tracker.ceph.com/issues/50314
Signed-off-by: Navin Barnwal <knbarnwal@gmail.com>
(cherry picked from commit 6c609ace4998a087f21fa55c576b1f3cf1a0e054)

mgr/dashboard: fix rename inventory to disks

Fixes: https://tracker.ceph.com/issues/50314
Signed-off-by: Navin Barnwal <knbarnwal@gmail.com>
(cherry picked from commit fe0bbdaf6a1da13b76a261f8aebe181fd66bee91)

Merge pull request #42659 from cfsnyder/wip-50386-pacific

pacific: SimpleRADOSStriper: use debug_cephsqlite

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

Merge pull request #42693 from tchaikov/pacific-pr-41215

pacific: cmake: Replace boost download url

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

Merge pull request #42692 from tchaikov/pacific-pr-42681

pacific: cmake: link bundled fmt statically

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

Merge pull request #42561 from smithfarm/wip-51969-pacific

pacific: rpm/luarocks: simplify conditional and support Leap 15.3

Reviewed-by: Yuval Lifshitz <yuvalif@yahoo.com>
Reviewed-by: Kyr Shatskyy <kyrylo.shatskyy@suse.com>

Merge pull request #42544 from cybozu/pacific-kv-rocksdbstore-enrich-debug-message

pacific: kv/RocksDBStore: enrich debug message

Reviewed-by: Neha Ojha <nojha@redhat.com>
Reviewed-by: Igor Fedotov <ifedotov@suse.com>

Merge pull request #42542 from smithfarm/wip-51955-pacific

pacific: compression/snappy: use uint32_t to be compatible with 1.1.9

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

Merge pull request #42224 from cfsnyder/wip-51568-pacific

pacific: mon/OSDMonitor: resize oversized Lec::epoch_by_pg, after PG merging, preventing osdmap trimming

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

qa: update mds_pre_upgrade to no longer stop standbys

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit 2cd3494771b41216ad99a3bf462e2462b30254b6)

qa: update mds_pre_upgrade to disable standby-replay

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit 8e0b9bcad67352a5ad0d91993e8fb26f482a08de)

qa: add tests for compat manipulation and upgrade

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit 295971b9c6520389b6780bfa3bddc3606eaba6dd)

Conflicts:
qa/tasks/cephfs/test_admin.py: trivial

doc: remove deprecated compat commands

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit 54b649af9b43b8ff48f39b033cc69f4d4e3b1b30)

doc: update MDS upgrade procedure

Now that CompatSet changes to the FSMap no longer cause old MDS to
suicide.

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit efb70f2b336b119ca99fe2abb79428ab4d3453a2)

mon,mds: use per-MDS compat to inform replacement

This diff makes the following changes:

- FSMap::compat is now just a "default compat" of currently unknown
  utility. It is used when constructing a new file system but does
  not really have any effect or current use.

- The `mds compat *` CLI commands are deprecated. They manipulate
  the default compat which has no useful effect.

- Each MDS sends its compat to the mons in its beacon. This is from
  MDSMap::get_compat_set_all() at MDS boot. This CompatSet does not
  change for the duration of the MDS lifetime.

- Mons record each MDS compat in the FSMap to inform standby failover.
  An MDS is only promoted if it is compatible with the file system
  compat.

- Mons upgrade (merge) the file system compat when (a) the number of
  *in* MDS is 1 (effected by max_mds=1) and (b) the mons are promoting a
  standby with a new compat. A file system is never upgraded when there
  is more than 1 rank to prevent two MDS with incompatible compat.

- A suite of `fs compat` commands exist to manipulate the file system
  compat. These exist mostly for testing.

The consequence of these changes is that the upgrade procedure for MDS
can be updated to no longer require turning off all MDS but rank 0
before performing any upgrades. A CompatSet change would cause all MDS
receiving the new MDSMap to suicide due to incompatibility (if so).
Instead, the monitors will no longer assign an incompatible MDS to a
file system and enforce an upgrade procedure if incompatibilities exist.

Fixes: https://tracker.ceph.com/issues/49720
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit 58eaa237b0a16d3c934ded77ed4dc53137d9b4a1)

Conflicts:
src/mds/FSMap.h: trivial
src/mon/MDSMonitor.cc: trivial
src/mon/MonCommands.h: work around removed OBSOLETE commands

mon: do not update inline incompat except via mds

The MDS_FEATURE_INCOMPAT_INLINE feature indicates that an MDS knows how
to read/write inline data and that the file system may have it. The
separate setting for inline_data protects this file system feature.

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit 8cdc36c89da224eb8e86ee6ed80cd2eb65354587)

mds: add MDSMap method for creating null MDSMap

It's not necessary to distribute a CompatSet with the null mdsmap. We
only need to communicate that the MDS is not part of any map.

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit 4a10b6016ff76f80e27574ad2ab40ac5b0266264)

mds: only update beacon epoch if newer

This is a defensive programming change. We don't want the beacon epoch
to ever go backwards.

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit 0256ae010f61520a67f95fcd3e8522784ea56f8b)

mds: harden standby_mds lookup

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit b8ad8a8c821ba57d64aa349a9da7117867b0f401)

mon/FSCommands: accept generic ostream rather than stringstream

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit 31c8edd603e7db84529eacde1771aef01dbde089)

Conflicts:
src/mon/FSCommands.cc: trivial

include: add less verbose CompatSet dump

For printing in `fs dump`.

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit 56b36e69fae0fcfb5081ede3934f1d63cde0494a)

include: add dump operator for Feature

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit 5da02036eefe60c89e11f53d4bacb0e41e03e794)

include: add const qualifier to appropriate CompatSet methods

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
(cherry picked from commit d179fa2bc0262854b1e001e0c9db3d2f508bf647)

mon/MDSMonitor.cc: fix join fscid not applied with pending fsmap at boot

Fixes: https://tracker.ceph.com/issues/49157
Signed-off-by: Zhi Zhang <zhangz.david@outlook.com>
(cherry picked from commit 4ab87e2eecbdcb4932775e871c3ae2f793880c20)

Merge pull request #42536 from batrick/i51940

pacific: MDSMonitor: monitor crash after upgrade from ceph 15.2.13 to 16.2.4

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

Merge pull request #42773 from neha-ojha/wip-52244-pacific

pacific: os/bluestore: make deferred writes less aggressive for large writes

Reviewed-by: Igor Fedotov <ifedotov@suse.com>

Merge pull request #42135 from dang/wip-dang-41991-backport

pacific: RGW - Bucket Remove Op: Pass in user

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

Merge pull request #40696 from singuliere/wip-50156-pacific

pacific: rgw/notification: fix persistent notification hang when ack-levl=none

Reviewed-by: Casey Bodley <cbodley@redhat.com>
Reviewed-by: Yuval Lifshitz <yuvalif@yahoo.com>

Merge pull request #40695 from singuliere/wip-50157-pacific

pacific: rgw/notification: fixing the "persistent=false" flag

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

Merge pull request #40693 from singuliere/wip-50214-pacific

pacific: rgw: objectlock: improve client error messages

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

mgr/dashboard: run-cephadm-e2e-tests.sh improvements

- Jenkins env.: make sure the cluster is always started.
- PR template: add trigger phrase to the jenkins commands list.
- Cypress: add --no-install flag; clean previous reports.

Fixes: https://tracker.ceph.com/issues/52082
Signed-off-by: Alfonso Martínez <almartin@redhat.com>
(cherry picked from commit cadf8c75ac0266e42c1ea0b4d6733590ea3dac21)

mgr/dashboard: cephadm e2e start script: add --expanded option

- Additional improvements: clean npm cache in jenkins env.; display mgr logs on error.

Fixes: https://tracker.ceph.com/issues/52082
Signed-off-by: Alfonso Martínez <almartin@redhat.com>
(cherry picked from commit 973a6f534383e84bd92b690eaf2862a5a30ba70c)

mgr/dashboard: tox.ini: delete useless env. 'apidocs'

Fixes: https://tracker.ceph.com/issues/52130
Signed-off-by: Alfonso Martínez <almartin@redhat.com>
(cherry picked from commit d7346ec1486d7ca24177c3064f457f9722c0c8cb)

test/store_test: more testing for deferred writes

Signed-off-by: Igor Fedotov <ifedotov@suse.com>
(cherry picked from commit 324ea6b425e5898eca6e86264fa4d34cb868a6ae)

os/bluestore: enforce one more non-inclusive comparision against prefer_deferred_size

Signed-off-by: Igor Fedotov <ifedotov@suse.com>
(cherry picked from commit cb30c9965854e7dbfe851ed0cfaf113b16bcab5e)

os/bluestore: Better handling of deferred write trigger

Now deferred write in _do_alloc_write does not depend on blob size,
but on size of extent allocated on disk.
It is now possible to set bluestore_prefer_deferred_size way larger than
bluestore_max_blob_size and still get desired behavior.
Example: for deferred=256K, blob=64K : when op write is 128K both blobs will be
written as deferred. When op write is 256K then all will go as regular write.

Signed-off-by: Adam Kupczyk <akupczyk@redhat.com>
(cherry picked from commit 06ac561a72a93bd24d05b6c2266d66fee49228b8)

os/bluestore: account for alignment with max_blob_size when splitting
write I/O into chunks.

Without the fix the following write seq:
0~4M
4096~4M

produces tons of deferred writes at the second stage.

Signed-off-by: Igor Fedotov <ifedotov@suse.com>
(cherry picked from commit b06bcf99f50cf7db5856eb1f6cd56100a0d0bbbc)

os/bluestore: introduce l_bluestore_write_deferred_bytes perf counter.

Signed-off-by: Igor Fedotov <ifedotov@suse.com>
(cherry picked from commit f38f18c149b9347a26d845539eaeef29979e9006)

os/bluestore: use non-inclusive comparision against prefer_deferred_size

Fixes: https://tracker.ceph.com/issues/52089
Signed-off-by: Igor Fedotov <ifedotov@suse.com>
(cherry picked from commit f1c4448fc549ed58bb58f920dfa16886e4910315)

os/bluestore: improve logging around deferred writes

Signed-off-by: Igor Fedotov <ifedotov@suse.com>
(cherry picked from commit 89a20d63ec7e255d36363bd8977c3fca1d277877)

os/bluestore: fix missing 'l_bluestore_write_deferred' perf counter increment

Signed-off-by: Igor Fedotov <ifedotov@suse.com>
(cherry picked from commit dc367c4bbc701b04a5bde562e30ab5b9c1705657)

os/bluestore: cosmetic cleanups

Signed-off-by: Igor Fedotov <ifedotov@suse.com>
(cherry picked from commit 080531d6813b868eda562cf27babe18d8a974413)

Merge pull request #42679 from callithea/wip-51992-pacific

pacific: monitoring/grafana/cluster: use per-unit max and limit values

Merge pull request #42755 from guits/wip-50896-pacific

pacific: ceph-volume: Consider /dev/root as mounted

ceph-volume: work around phantom atari partitions

See Rook issue https://github.com/rook/rook/issues/7940 for full
information.

Ceph bluestore disks can sometimes appear as though they have "phantom"
Atari (AHDI) partitions created on them when they don't in reality. This
is due to a series of bugs in the Linux kernel when it is built with
Atari support enabled. This behavior does not appear for raw mode OSDs on
partitions, only on disks.

Changing the on-disk format of Bluestore OSDs comes with
backwards-compatibility challenges, and fixing the issue in the Kernel
could be years before users get a fix. Working around the Kernel issue
in ceph-volume is therefore the best place to fix the issue for Ceph.

To work around the issue in Ceph volume, there are two behaviors that need
adjusted:
1. `ceph-volume inventory` should not report that a partition is
   available if the parent device is a BlueStore OSD.
2. `ceph-volume raw list` should report parent disks if the disk is a
   BlueStore OSD and not report the disk's children, BUT it should still
   report children if the parent disk is not a BlueStore OSD.

Signed-off-by: Blaine Gardner <blaine.gardner@redhat.com>
(cherry picked from commit 9212420bbc88e0caaf0b2302dd1f379f563f7d26)

ceph-volume: use safer check for bluestore label

Using only the exit status of `ceph-bluestore-tool show-label` to
determine if a device is a bluestore OSD could report a false negative
if there is a system error when `ceph-bluestore-tool` opens the device.

A better check is to open the device and read the bluestore device
label (the first 22 bytes of the device) to look for the bluestore
device signature ("bluestore block device"). If ceph-volume fails to
open the device due to a system error, it is safest to assume the device
is BlueStore so that an existing OSD isn't overwritten.

Signed-off-by: Blaine Gardner <blaine.gardner@redhat.com>
(cherry picked from commit 651b28f2e3cb39dbe9c7038cd677a01523f08821)

ceph-volume: consider /dev/root mounted

Fixes: https://tracker.ceph.com/issues/50604
Signed-off-by: David Caro <david@dcaro.es>
(cherry picked from commit c7b959c35ff463cfbffa2abf0f56a1c1a5f594c8)

Merge pull request #42493 from dsavineau/wip-51852-pacific

pacific: ceph-volume: lvm batch: fast_allocations(): avoid ZeroDivisionError

Merge pull request #42546 from dsavineau/wip-51958-pacific

pacific: ceph-volume/tests: retry when destroying osd

cephadm: fix container name detection

'enter' was broken because we weren't correctly identifying the container
name.  Strip the newline from the inspect result so that we can reliably
match against the 'running' state.

Signed-off-by: Sage Weil <sage@newdream.net>
(cherry picked from commit fb5e4b3740241d5c41b348e0fc5f59ebde5ee558)

cephadm: add AlmaLinux to supported distro

AlmaLinux [1] is an open-source 1:1 binary compatible fork of RHEL.

[1] https://almalinux.org/

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit f450f25bcf2ddd3322efca8f0c21eba90178840d)

doc/cephadm: update documentation

This commit fixes a couple of typos in the cephadm documentation.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 32b2c57bf66ccb030e24f78e3a5902babc1bbb38)

cephadm: fix unit tests.

Fixes AttributeError: 'FakePipeWrapper' object has no attribute 'readable'

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit 61cfaec7002caadb3b4e09d672919ed159853311)

doc/cephadm: Offline hosts will pause upgrades

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit 1eb48e3876d0b1fab7d1fac2574639112d39a8bf)

cephadm: Introduce unit.stop

Reason is we now have to stop two containers named differently. This is possible
with `bash -c ... echo %i | tr . -`. But we're not gainig anything readability
compared to putting it into a unit.stop script.

As not all daemons have this stop script, we still have to call podman for old daemons.

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit 0f66a5198e38c59aba1be11d586d49ad01c9afce)

cephadm: don't log on exit.

At this point, the garbage collector already colleceted the `open` function

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit 71162c6fea7aba2e30cf50d8a0f1a5a0cf3b2462)

cephadm: `stats` might return `--` for containers

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit d183661be552acf7f30fce2e038173c5bdfa8421)

cephadm: exec: also search for old cname

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit a57ab6f50968921b26b87695e704f84a25101597)

cephadm: dashes: use both cnames for `inspect`

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit a1f9959df474e671e979849450f8a00d0e24ea72)

cephadm: Add CephContainer.for_daemon()

To avoid calling get_container()

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit bb0587b8e0bde24d9bbfed7c65e0d351c5f7446b)

cephadm: use dashes for container names

podman adds the current container name to the /etc/hosts
file. Turns out, python's `socket.getfqdn()` differs from
`hostname -f`, when we have the container names containing
dots in it.:

[root@sebastians-laptop /]# cat /etc/hosts
127.0.0.1   localhost
::1         localhost
127.0.1.1   sebastians-laptop foo.bar.baz.com
[root@sebastians-laptop /]# hostname -f
sebastians-laptop
[root@sebastians-laptop /]# python3 -c 'import socket; print(socket.getfqdn())'
foo.bar.baz.com

Fascinatingly, this doesn't happen when using dashes.

Fixes: https://tracker.ceph.com/issues/51590
Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit 67abea15b12c1a60a9da2db35e3470d8d8a128f7)

doc/cephadm: MON IP change

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit 9df20490b210cf0e5fcf17920e91d390989cc4bf)

doc/cephadm: Move some sections from mon.rst to serivce-management.rst

Avoid duplication and instead only reference the corresponding sections.

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit 24b753f2c8111305cf73de9b40a393c0c1a9bae7)

cephadm: Clarify no container engine message

Multiple people didn't understand the message. Let's try to impove it.

(Also make mypy happy)

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit 240b2435bad903b811a3279b0e7a5d0806ee0b93)

cephadm: Add missing type annotations

Turnd out `command_maintenance` missing an if-branch that returned
None instead of a str

Signed-off-by: Sebastian Wagner <sewagner@redhat.com>
(cherry picked from commit 0c78d239a1933338fea14e21fde2638a418bc660)

Conflicts:
src/cephadm/cephadm

cephadm: set global default container for ingress

This was partially done in b94c8de but only for haproxy in the cephadm
mgr module not in the cephadm binary.
This adds the same change for keepalived container image.
Now both haproxy and keepalived container images are fully qualified
(registry + namespace + image).

Fixes: https://tracker.ceph.com/issues/51973
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit f9eb7d826563ab9b995b4123c03f6d1b89423e01)

doc/cephadm: update about disabling logging to journald for quincy

Signed-off-by: Varsha Rao <varao@redhat.com>
(cherry picked from commit bdb2241ca5a9758e8c52d47320d8b5ea0766aea9)