cephadm-adopt: fix ssl condition in rgw placement task

rgw daemons fail to start as they were looking for
ssl certificates while none were available.

Signed-off-by: Teoman ONAY <tonay@ibm.com>

rolling_update: fails on FSMap sanity check when containerized

Fixes missing environment variables

Signed-off-by: Teoman ONAY <tonay@ibm.com>

ceph_config: fix generate_cmd to generate_ceph_cmd typo

Signed-off-by: Teoman ONAY <tonay@ibm.com>

rolling_update: disable FSMap sanity checks

Disable FSMap sanity checks before upgrade to prevent
mon crash(1)

(1) https://access.redhat.com/solutions/7020523

fixes: https://bugzilla.redhat.com/show_bug.cgi?id=2277756

Signed-off-by: Teoman ONAY <tonay@ibm.com>

Dashboard: upgrade fails

The command ceph dashboard grafana dashboards update fails with Error EINVAL: Traceback
https://access.redhat.com/solutions/6965628

fixes: https://bugzilla.redhat.com/show_bug.cgi?id=2295395

Signed-off-by: Teoman ONAY <tonay@ibm.com>

cephadm-adopt: Fixes binding network for alertmanager

Alertmanager was bind to default * network instead of grafana_server_addr
as it was before. Now on if grafana_server_addr is defined, it will be
bind to that network.

Signed-off-by: Teoman ONAY <tonay@ibm.com>

ceph_orch_apply: fix yaml error when multiple rgw deployed

ceph orch ls rgw --format=yaml returns multiple documents
when multiple rgw are installed, this was not handled
correctly.

Signed-off-by: Teoman ONAY <tonay@ibm.com>

cephadm-adopt: fix "Update the placement of radosgw hosts" task

networks was at the wrong level in the spec file. Failed with
"got an unexpected keyword argument 'networks'"

Signed-off-by: Teoman ONAY <tonay@ibm.com>

cephadm-adopt: fix "Update the placement of radosgw hosts" task

spec file template conditions were incorrect

Signed-off-by: Teoman ONAY <tonay@ibm.com>
(cherry picked from commit e85060cb6713b05757f87f5bda3368416aa6e64e)

ceph_orch_spec: Add ceph orch apply spec feature

Add new module ceph_orch_spec which applies ceph spec files.
This feature was needed to bind extra mount points to the RGW
container (/etc/pki/ca-trust/).

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=2262133
Signed-off-by: Teoman ONAY <tonay@ibm.com>

cephadm-adopt: Fixes hosts addition to be managed by cephadm

The tasks "manage nodes with cephadm - ipv4/6" are skipped when
cephadm_mgmt_network contains more than one ip network which prevent
cephadm from managing the host.

Signed-off-by: Teoman ONAY <tonay@ibm.com>

cephadm-adopt: Add --networks parameter support to ceph orch apply rgw

When radosgw_address_block was defined, it was not taken into account
during rgw adoption process

depends on: https://tracker.ceph.com/issues/62185
fixes: https://bugzilla.redhat.com/show_bug.cgi?id=2224351

Signed-off-by: Teoman ONAY <tonay@ibm.com>
(cherry picked from commit bc54290718b9303c9eadc1f6c4e7c0e79e54e41c)

tests: use quay.io instead of quay.ceph.io

This makes the CI use quay.io instead of quay.ceph.io

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

whitelist->allowlist

https://github.com/tox-dev/tox/blob/master/docs/changelog.rst#v3180-2020-07-23

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

Remove all vagrant boxes

Remove all existing vagrant boxes to prevent vagrant from using
older ones

Signed-off-by: Teoman ONAY <tonay@ibm.com>

Uses a more recent version of the CentOS stream 8 box

Uses the latest centos/streamX image available.

Signed-off-by: Teoman ONAY <tonay@ibm.com>

Fix selinux label issues

Add --security-opt label=disable to all containers
accessing /var/lib/ceph. podman selinux relabeling behavious changed
since version podman-3:4.2.0-1 which prevent some containers to access
files in these subdirectories.

Signed-off-by: Teoman ONAY <tonay@ibm.com>

Fixes selinux relabeling issue for nfs container

Signed-off-by: Teoman ONAY <tonay@ibm.com>

Replace ip_version check with ansible test

Instead of checking ip_version variable we should check the input
address for ip version and select code path based on that.

This solves ceph adoption with mixed ipv6 and ipv4 networks.

Resolves: rhbz#2186226
Signed-off-by: Lukas Bezdicka <lbezdick@redhat.com>

cephadm-adopt: fix rbd-mirror adoption

The recent rbdmirror refactor introduced a regression in the
cephadm-adopt playbook.
Given that the rbd-mirror peer addition is now done by using the monitor
config-key store method during the cluster deployment, we can drop this
play from the cephadm-adopt.yml playbook.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2140569
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit c3fae04b8d234c84e4a5c7ca4a390ecfa3d1c1d5)

rbd-mirror: add rx peer only once

in order to avoid the following error:

```
multiple RX peers are not currently supported
```

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2037646
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit aa64747cd111e37988849dbd20600b3b1fc049fa)

rbd-mirror: fix multiple rbd-mirror daemon support

This `run_once: true` breaks multiple rbd-mirror daemons support
as it would make all rbd-mirror daemons use the same keyring.
Each rbd-mirror daemon needs its own keyring in order to start.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2037646
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 665601bf8aae1b98e1e3efdb91c1f1914f0316ae)

Setting fact _radosgw_address fail when RGW is on a different network

Changed the when condition to only execute that fact setting on RGW
nodes while before it was run on all nodes and failed if the node
was not on the same network range as the RGW.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2131150
Signed-off-by: Teoman ONAY <tonay@redhat.com>
(cherry picked from commit 590e57cc1b9a79ced7a9208db8bf8eaf43d3f320)

switch-to-containers: ignore errors when stopping service

There might be cases where it can break idempotency.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 09b8f7b2ea4deb35228adb948b50976ad1854870)

switch-to-containers: fix rbd-mirror migration

`--state=enabled` isn't a valid filter so the unit from the packaging
never gets removed.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2134917
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 7664da58da56d871a9304ea6bec8d0e30ba9b02a)

ansible: require 2.9 for downstream

RHCS 5.3 supports 2.9 only.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2120187
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

facts: set is_rgw_instances_defined from configure_dashboard

When we come from configure_dashboard.yml, this fact should be set if
`rgw_instances` is defined in group_vars/host_vars. Otherwise, the next
task that set the fact `rgw_instances` will be run as it will assume it
wasn't user defined.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2117294
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

config: use osd_memory_target value from ceph_conf_overrides if defined

otherwise it's impossible to override `osd_memory_target`
via `ceph_conf_overrides`.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2056675
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit f19dcb266ae6c6539e3ed613fdaa94fdde636b26)

config: do not always set _osd_memory_target

When 'osd_memory_target' is overridden in ceph_conf_overrides.
The task that sets the fact `osd_memory_target` in the ceph-config role
should be skipped.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2056675#c11
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit e2076e439b95bc695e1bc0764cbb8d7d5fc1fe4a)

purge-dashboard: check for legacy group name 'grafana-server'

When using the legacy group name 'grafana-server', this playbook will run but
won't remove properly all monitoring resources as expected.

Fixes: #7265
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit a9cb444be1f1b601d96ad8ff8970ac2d487ca17e)

adopt: fix placement update calls for rgw

The commands called here are not built correctly.
This commit fixes it.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2058038#c27
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 30c7e88d81dcc08641323aecee186d1f81ddfb84)

tests: do not use dev repo (rbdmirror)

use stable repositories for rbdmirror scenarios

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

replace 'master' references with 'main'

Because the branch'master' was renamed 'main'

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

flake8: fix 'error E275 missing whitespace after keyword'

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 7ce972e7155fd0e0aaf7352656c0805d9a2afad5)
(cherry picked from commit 82dc265e2f8b8714794be0e517aaa746d670dd6a)

tests: skip rbdmirror tests on non-secondary daemon

the daemon is not running on the 'primary' daemon.
Therefore, these tests are not needed.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit a12f67e1388b35974842f4fa8f7c56c99f56cd0a)
(cherry picked from commit 022199745ecd92d670e6eeff14f5a3fee08e0313)

tests: set no_log_on_ceph_key_tasks=false

In order to not have to always reproduce it when a failure shows up in the CI
having the failure logged can make us save some time.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 89ddc193fc985fbe9a1bb3d1bcd21a29e3bc750d)
(cherry picked from commit 6ef470ef0d2bea713497163c45072987ca3e9fb2)

rbd-mirror: follow up on recent rbd-mirror refactor

- ensure /var/lib/ceph/bootstrap-rbd-mirror exists
- always install ceph-base on rbdmirror nodes (otherwise, ceph-crash
  isn't present)

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 041435e1e371f00b7112efed3306beec91d5a434)
(cherry picked from commit b634fb1cb3cef6d895678501dfcd817a4fdea99c)

Set ceph_rbd_mirror_pool default value

Signed-off-by: Teoman ONAY <tonay@redhat.com>
(cherry picked from commit 0c50bfac98cd174bd785487574af54a468b82ae6)
(cherry picked from commit 8a0b5a9571df55cff3b4d4d5ab8b41d739a6e1cb)

Playbook fails when using --limit to install new MDS

"set_fact container_run_cmd" is not set when using --limit on MDS as facts
were not run on first MON.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2111017
Signed-off-by: Teoman ONAY <tonay@redhat.com>
(cherry picked from commit cef1636f70832c95b463f0d14040e06bebd4fc59)
(cherry picked from commit 80dc46d31f682d9ea54de6e748806cbefd6a70d3)

doc: update readthedoc settings

This is needed to enforce the python version.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 3a8daafbe8c9023c6dcd8034adfcc98893e5c303)

rbd-mirror: major refactor

- Use config-key store to add cluster peer.
- Support multiple pools mirroring.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit b74ff6e22c0d1b95e71384e4d7e2fb2ad556ac39)

replace 'master' references with 'main'

The branch 'master' has been renamed 'main'.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

fixes: https://github.com/ceph/ceph-ansible/issues/7261

Signed-off-by: Sven Kieske <s.kieske@mittwald.de>

config: followup on 8a5628b51

Add missing `--cluster {{ cluster }}` on task
`set osd_memory_target` in the main.yml file of the
ceph-config role.
Also it moves the task after ceph configuration file is actually written.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit cf4a430d0b1d7946fc65babc7841ea1e7c85cbee)

config/osd: various fixes

- sets `osd_memory_target` per osd host.
- ceph.conf refactor (osd)

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2056675
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 8a5628b51616b0b9740680bd14c669d4316941f3)

config: fix indentation in main.yml

For consistency and readability.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 5283fa6e966ae3bcde82559240794b8fd5c0d8f3)

Refresh /etc/ceph/osd json files content before zapping the disks

If the physical disk to device path mapping has changed since the
last ceph-volume simple scan (e.g. addition or removal of disks),
a wrong disk could be deleted.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2071035
Signed-off-by: Teoman ONAY <tonay@redhat.com>
(cherry picked from commit 64e08f2c0bdea6f4c4ad5862dc8f350c6adbe2cd)

backup-and-restore: use archive/unarchive approach

current approach is too complex and causes too many issues permission
issues.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2051640
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit dffe7b47de70b6eeec71a3fa86f8c407adb4dd8e)

update: fix a typo

s/pre-quincy/pre-pacific
s/quincy-only/pacific-only

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

facts: fix set_radosgw_address.yml

use `include_tasks` instead of `import_tasks`.
Given that with `import_tasks` statements are preprocessed
and the tasks that defines it hasn't been run yet, it will fail
and complain like following:

```
The error was: 'ansible.vars.hostvars.HostVarsVars object' has no attribute '_interface'
```

Using `include_tasks` instead fixes this.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 434793e2feec64fdbbb8a7bdb959b08dacbb2006)

facts: fix deployments with different net interface names

Deployments when radosgws don't have the same names for
network interface.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2095605
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit f6b49f78a9f14c37b2ca81fa6172beba8f43adc4)

backup-and-restore: various fixes

- preserve mode and ownership on main directories
- make sure the directories are well present prior to restoring files.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2051640
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 047af3a3f66b6135cf7ab6a7c28b9c30dea3c836)

tests: add yes_i_know=true in tox-shrink_osd.ini

main branch requires it. Otherwise the playbook won't run.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit e223630cf07f71a1763f39be621950dfc723fed0)

tests: drop shrink_osd from tox.ini

shrink_osd has its own tox config file (tox-shrink_osd.ini)

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 6623f34679f68f21d0129111ecef1042d3359bf4)

Revert "upgrade: block upgrade when rgw multisite is active"

This reverts commit 51bc8cb63642b5b1ca460e79adf4649d92773aa4.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 7d848fa19e80389e3c53a94303acfb117d558400)

common: fix a typo

s/of/or ..

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=2099828#c25
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 2e823b117ef5e6c7396517aa0eb2c8553bd33887)

backup-and-restore: fix check on 'target_node' variable

If the user doesn't pass a valid name (present in the inventory)
the playbook will fail like following:

```
fatal: [localhost -> {{ target_node }}]: FAILED! =>
  msg: |-
    The task includes an option with an undefined variable. The error was: "hostvars['10.70.46.40']" is undefined
```

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2051640
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit b18a1aa3cafa235dfa4f564d6574318ee539f961)

backup-and-restore: fix check on 'mode' variable

Typical failure:

```
fatal: [localhost]: FAILED! =>
  msg: |-
    The conditional check 'mode not in ['backup', 'restore']' failed. The error was: error while evaluating conditional (mode not in ['backup', 'restore']): 'mode' is undefined
```

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2051640
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 848dd03fa66c0febaf1ed081d4986d5de8472a09)

purge: reset-failed ceph-crash

This ensures we always reset-failed the ceph-crash service.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2055992
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 5ab46f836dd0903b6e009178c3ef950d7524bf32)

backup-and-restore: fix a typo

Typo introduced during initial implementation.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2051640
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit e28c486e528517439997583bcb0faa3df16ab848)

fix(ceph-grafana): make dashboard download work again

This fixes the dashboard download for pacific and later.

Signed-off-by: Michael Wagner <mitch.wagna@gmail.com>
(cherry picked from commit 4edaab5f4c5445cb1fafc5d8824c49717e9f96c8)

Add ceph_infra tag to rolling_update

When the upgrade from Ceph 4 to 5 is performed in the OpenStack context,
ceph-ansible triggers the rolling_update playbook, which is supposed to
rollout new Ceph containers.  The ceph-infra role tries to take care
about firewall, ntp config and logrotate; however, TripleO manages them
through tripleo-heat-templates.  This patch just add an additional tag
to skip the ceph-infra role in the OpenStack context.

Closes: https://bugzilla.redhat.com/2090456
Signed-off-by: Francesco Pantano <fpantano@redhat.com>
(cherry picked from commit 0e9b3902b074f3526e2fc9e392a32113cf4c6d99)

ansible: use ansible.utils.ipwrap instead of ansible.netcommon.ipwrap

ansible.netcommon.ipwrap is deprecated and is not being redirected with ansible 2.9.*

Signed-off-by: Seena Fallah <seenafallah@gmail.com>
Co-authored-by: Guillaume Abrioux <gabrioux@redhat.com>

doc: update ansible version requirement

This updates the documentation regarding the Ansible version
requirement for `stable-6.0` branch.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 5d9001b8e9b36c9671c93fd503677de61ca38dd2)

switch to ansible.netcommon.ipwrap

As of 2.10, Ansible moved ipwrap to netcommon collection.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

tests: uninstall ansible prior to install 6.0 req

in order to avoid conflict with previous version installed

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

ceph_pool: set target size ratio on both 'on' and 'warn' mode

when we set target_size_ratio to warn it means that the administrator wants to get suggestion from the mgr module but apply it manually when he/she wants. So it's in the same approach as 'on' mode just triggered by hand.
So there is no need to set pg_num when target_size_ratio is 'warn' and the mgr module will calculate the correct pg_num and the administrator will adjust it whenever he/she wants.

It is the same approach that was in #6471

Signed-off-by: Seena Fallah <seenafallah@gmail.com>
(cherry picked from commit bb849a55861e3900362ec46e68a02754b2c892ec)

cephadm_adopt: set autotune_memory_target_ratio

This adds a task that sets `autotune_memory_target_ratio` depending on the
value of `is_hci`.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2028693
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 41d62596fcbe01c83c97bfa28ecbf7ee00b9d072)

tests: use latest version for pytest

with the bump of py version, let's use newer version for pytest.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 5a136889795ee46f8fbee7c95ba7fd8ce0b12bdf)

core: bump ansible version

We should consider bumping ansible version for future releases, so let's
start testing against ansible 2.10

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 839fac8f9418ce54af66a4ff0a3a9b8db6814f26)

common: config rhcs tools repo on all nodes

Otherwise `cephadm` can't be installed during cephadm-adopt.yml
playbook execution.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2073480
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 1e11f879f6f535bcfc1ccf9788a990b0886ea72d)

cephadm-adopt: remove legacy directory after adoption

When this directory is left after the osd adoption, it leads to the following error:

```
[WRN] CEPHADM_REFRESH_FAILED: failed to probe daemons or devices
    host axdesec2ocs1n002.ecommerce.inditex.grp `cephadm ceph-volume` failed: cephadm exited with an error code: 1, stderr:Inferring config /var/lib/ceph/41555360-e96b-4b16-a37c-873e0c940091/mon.axdesec2ocs1n002/config
ERROR: [Errno 2] No such file or directory: '/var/lib/ceph/41555360-e96b-4b16-a37c-873e0c940091/mon.axdesec2ocs1n002/config'.
```

this is because of an unexpected behavior regarding 'config inferring' when a legacy directory is present in /var/lib/ceph.

Note: this doesn't fix the root cause, this is a workaround.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2075510
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 6e2ebe857d0f8c2b9a7ebbd5c63fa54c9a603141)

contrib: add a playbook

this playbook can backup or restore some ceph files.
(/etc/ceph, /var/lib/ceph, ...)

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2051640
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit ed0bba4d770d7b886de83efe25d482e08d558f97)

facts: follow up on aa0cc93

when these variables are defined in the inventory host file,
all tasks are skipped then because the node being played isn't
aware about the values from the rgw nodes.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2063029
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

facts: fix mon/mgr collocation

`service dump` hangs when no active mgr is available.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

dashboard: fix regression

introduced by ceph/ceph-ansible/pull/7150

when no rgw is present, it fails.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2076192
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

added AlmaLinux and Rocky for iscsi deploy

Signed-off-by: Ingo Ebel <ingo.ebel@desy.de>
(cherry picked from commit c5bb450f8796530495315cd9b81539e059661c3d)

dashboard: support --limit execution with rgw

When the following conditions are met:

- rgw is deployed,
- dashboard is deployed,
- playbook is called with --limit,
- a node being processed is collocated on either a mon or mgr.

The playbook fails because `rgw_instances` is undefined.
The idea here is to make sure this variable is always defined.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2063029
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

dashboard: allow collecting stats from the host

This commit makes podman bindmount `/:/rootfs:ro` so the container can
collect data from the host.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2028775
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 0f34cd16d808dbb800efd0f558b29bcfb218bbfa)

ceph-facts: fix ansible templating error for auto osd discovery

This commit fixes templating error that occurs when using auto osd discovery. Getting the len before converting the result to a list causes "object of type generator has no len()" error.

Signed-off-by: pinotelio <ahmadreza.mollapour@gmail.com>
(cherry picked from commit f288364c5c268e61d49165704900e8e01ca643c8)

validate: drop a check

Since the ISO install method removal, ceph-ansible isn't able
to detect wheter the user is deploying in a 'disconnected environment'.
By the way, given that ceph-ansible is available only for upgrading to RHCS 5,
this check doesn't make sense anymore, let's drop it.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2062147
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 1cd1fa05601384c5cd8d907b0581fdfaf125c098)

do not update Debian cache when package-install is disabled

When deploying with --skip-tags=package-install (when there is no access to a repository), the playbook is still trying to update the package cache, which makes the playbook fail.
This change prevents the playbook to try to update the cache when the package-install tag is skipped.

Signed-off-by: Florent CARLI <florent.carli@rte-france.com>
(cherry picked from commit 58fdc03e63f5ae726090a19a2412ea2eec5a1f4e)

library/ceph_pool: Fix potential null value when creating pools

Before, creating a pool by providing *only* `pg_num` would result in an
error as it would produce `--pgp-num null`.

This commit fix this behavior by defaulting `pgp_num` value to `pg_num`.

Signed-off-by: Mathias Chapelain <mathias.chapelain@proton.ch>
(cherry picked from commit f0f1dd986aab8b32fad55a8885faf5ed7ef66e06)

dashboard: always set `dashboard_server_addr`

When running the playbook with `--limit`, if the play targeted doesn't match
hosts present in the mgr group the playbook can fail.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2063029
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 72e4654aaea9638627c3e44d917d5af2a6ecdf0c)

Using another user than root for cephadm ssh connections fails

Fixes commit da42f3d139e595d09edfb30334fbc7ce17ffa3fe

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2048734
Signed-off-by: Teoman ONAY <tonay@redhat.com>
(cherry picked from commit f851d3232c6237c46044fb68fdb70ccc0d0deb3c)

tests: update the system before deploying

Having a system up-to-date is usually a good idea.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 3e87df5e8fefb963da86a4ab87a5f555b89ea3e1)

upgrade: block upgrade when rgw multisite is active

With this commit, upgrading a cluster from Nautilus to Pacific with
active rgw multisite replication will be blocked.
This is because a lot of bugs are currently present in Pacific regarding
RGW multisite.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2063702
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 51bc8cb63642b5b1ca460e79adf4649d92773aa4)

Turn off SELinux separation for containers MON and RGW

Initially MONs and RGW binded /etc/pki/ca-trust/extracted using the :z flag
(introduced to solve an OSP TripleO issue on RHEL - #3638) but using
this flag prevents local services (like sssd) running on the host from accessing
the certificates/files in that folder.

Signed-off-by: Teoman ONAY <tonay@redhat.com>
(cherry picked from commit 7e8ce2567ec7f163c763547252a7a5bcc983fd98)

purge: ceph-crash purge fixes

This fixes the service file removal and makes the playbook
call `systemctl reset-failed` on the service because in Ceph
Nautilus, ceph-crash doesn't handle `SIGTERM` signal.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2055992
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 2f11982590e298e92f7d3c4a06026fb00bea40cd)

adopt: fix node labelling

When using group of group, the playbook will apply undesired
labels on nodes.
This commit fixes it by applying only the expected labels.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2057528
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 266b6e739c26a931f1c975a990f22906223015ef)

Add cluster custom name support

When using cluster custom names, cephadm commands are executed using
the default admin keyring name which fails.

Signed-off-by: Teoman ONAY <tonay@redhat.com>
(cherry picked from commit f8c6bba65782a861e5be51432e072efb0c4b0110)

Enable user to change the account used for ssh connection

By default cephadm uses root account to connect remotely
to other nodes in the cluster. This change allows to choose
another account.
This commit also allows to use a dedicated subnet for cephadm mgmt.

Signed-off-by: Teoman ONAY <tonay@redhat.com>
(cherry picked from commit da42f3d139e595d09edfb30334fbc7ce17ffa3fe)

ceph-facts: ignore mounted disks on osd auto discovery

Ignore disks with active mountpoint when osd_auto_discovery is true

Signed-off-by: Seena Fallah <seenafallah@gmail.com>
(cherry picked from commit 9d87fd87cb1af68d8af21bf20bb3acadc80f587b)

switch2containers: fail if less than 3 monitors

This playbook doesn't support less than 3 monitors present in the inventory.
Just like the rolling_update playbook, let's fail if less than
3 monitors are present.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2049132
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit f08129edf2efb8bad21043b7e90af59e6f3e1a8d)

tests/library: Add radosgw_caps tests

Provide basic tests for `radosgw_caps` module.

Signed-off-by: Mathias Chapelain <mathias.chapelain@proton.ch>
(cherry picked from commit 6c0c53fa2d0da34fde63c69067faebb01d2087df)

library: Add radosgw_caps to manage capabilities

This commit add `radosgw_caps` module to be able to manage RadosGW users
capabilities.

Usage from module's documentation:
```YAML
- name: add users read write and all buckets capabilities
  radosgw_caps:
    name: foo
    state: present
    caps:
      - users=read,write
      - buckets=*
- name: remove usage write capabilities
  radosgw_caps:
    name: foo
    state: absent
    caps:
      - usage=write
```

This module support check mode by simulating the original `radosgw-admin`
behavior when adding capabilities.

Signed-off-by: Mathias Chapelain <mathias.chapelain@proton.ch>
(cherry picked from commit 65003f4ff90413094d46b682b78f6d749fcc2747)

module_utils/ca_common: Add optional diff argument to exit_module

Add optional `diff` argument defaulting to `{ "before": "", "after": ""}`.

Signed-off-by: Mathias Chapelain <mathias.chapelain@proton.ch>
(cherry picked from commit 055326be7cfe7221517bd382a220a2434bcc604e)

cephadm: set allow_overwrite at bootstrap step

Signed-off-by: yasinlachiny <yasin.lachiny@gmail.com>
(cherry picked from commit 2a2875847235ac0d1a6e61ecf4a21b45553f38ac)

ceph-facts: Fix get_def_crush_rule_name.yml in check mode

This construct doesn't work as intended since ansible/ansible#74212:

```
item.stdout | default('{}') | from_json
```

That PR made the `command` module return `stdout` even in check mode (setting
it to the empty string), so `default()` has no effect in that case and
`from_json()` fails to parse an empty string.

Instead, `default()` needs to be invoked with its second argument set to
`True`, so that it replaces any `False` value (such as an empty string) with
its first argument:

```
item.stdout | default('{}', True) | from_json
```

Signed-off-by: Benoît Knecht <bknecht@protonmail.ch>
(cherry picked from commit 7684d892c083ea1e07ada163e78da9f0359b116f)

ceph-osd: Fix crush_rules.yml in check mode

Set a default value for `item.stdout` before passing it to `from_json()`. The
`when` condition doesn't prevent this template from being evaluated in check
mode, so it fails if `item.stdout` doesn't contain a valid JSON string.

Signed-off-by: Benoît Knecht <bknecht@protonmail.ch>
(cherry picked from commit ef05e9a313ad232508440131b59f1ab832aaa46d)