ceph-ansible.git
2 years agoshrink-osd: fix regression because of a wrong regex 4.2z3 v4.0.62.7
Per Abildgaard Toft [Wed, 20 Oct 2021 07:45:16 +0000 (09:45 +0200)]
shrink-osd: fix regression because of a wrong regex

968891f4498da9625acfdd34bfb01fe445d1eef2 introduced a regression.
The regex is wrong because it doesn't allow to shrink osds with id
greater than 9

Fixes: #6950
Signed-off-by: Per Abildgaard Toft <per@minfejl.dk>
(cherry picked from commit 84118a3063e38ed9d274cca90d115809353819b4)
(cherry picked from commit c5e4851a3ff721ff933c05e9911805cfd7132148)

2 years agoshrink-osd: check osd id format
Guillaume Abrioux [Tue, 12 Oct 2021 15:55:40 +0000 (17:55 +0200)]
shrink-osd: check osd id format

This adds a check early in order to ensure the format of osd ids passed
is correct.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2005734
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 968891f4498da9625acfdd34bfb01fe445d1eef2)
(cherry picked from commit 3f4abb09b400dcf7563d2f662815d556cd4bf5bc)

2 years agoupdate: move a set_fact hotfix-bz2014304 v4.0.62.6
Guillaume Abrioux [Thu, 28 Oct 2021 21:40:18 +0000 (23:40 +0200)]
update: move a set_fact

ceph-facts roles makes decisions based on the fact `rolling_update` so
it must be called before we run this role.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2014304
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit bf365929c526b9c24bb24efa1b38891ee14aaef0)

2 years agomgr: append balancer module to ceph_mgr_modules
Guillaume Abrioux [Thu, 28 Oct 2021 23:28:38 +0000 (01:28 +0200)]
mgr: append balancer module to ceph_mgr_modules

otherwise the osd play in rolling_update can fail when it tries to
disable it before upgrading osd nodes.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
2 years agoceph-mgr: move mgr module list to common
Dimitri Savineau [Thu, 15 Jul 2021 19:38:07 +0000 (15:38 -0400)]
ceph-mgr: move mgr module list to common

Populating the ceph_mgr_modules list in the mgr_modules doesn't make sense
since that file is only executed if the list isn't empty or we're using the
dashboard.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit cd06e7c046b3e56920b1f9bdc1907429382bee5c)
(cherry picked from commit 8d58c50f4571dae8bdd3b4b22a308f42a0965053)

2 years agoupdate: support --limit on monitor nodes v4.0.62.5
Guillaume Abrioux [Thu, 28 Oct 2021 14:17:24 +0000 (16:17 +0200)]
update: support --limit on monitor nodes

Change needed in order to support --limit on mon nodes.
Otherwise, a call to `hostvars[groups[mon_group_name][0]]['_current_monitor_address']`
throws an error:

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

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2014304#c28
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 82eee4303bce3e41b5043bcb03fa3143dcdfd30d)

2 years agorolling_update: fix pre and post osd upgrade play v4.0.62.4
Guillaume Abrioux [Mon, 25 Oct 2021 11:43:25 +0000 (13:43 +0200)]
rolling_update: fix pre and post osd upgrade play

when using --limit osds, the play before and after osd upgrade are
skipped because we use `hosts: "{{ mon_group_name | default('mons') }}[0]"`
using `hosts: "{{ osds_group_name | default('osds') }}" with
`delegate_to` to the first monitor addresses this issue.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit fc9f87c45f6e58a595f59365b58bf1b0e3909bf5)
(cherry picked from commit 3dd96da652633401183e87ca643667df07bac10f)

2 years agoupdate: support upgrading a subset of nodes
Guillaume Abrioux [Wed, 20 Oct 2021 08:01:05 +0000 (10:01 +0200)]
update: support upgrading a subset of nodes

It can be useful in a large cluster deployment to split the upgrade and
only upgrade a group of nodes at a time.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2014304
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit e5cf9db2b04f55196d867f5a7248b455307f4407)
(cherry picked from commit dc1a4c29ea2a1ce77aa56223ed39fb470ef2276f)

2 years agopodman pids.max default value is 2048, docker's one is 4096 which are v4.0.62.3
Teoman ONAY [Tue, 3 Aug 2021 14:06:53 +0000 (16:06 +0200)]
podman pids.max default value is 2048, docker's one is 4096 which are
sufficient for the default value (512) of rgw thread pool size.
But if its value is increased near to the pids-limit value,
it does not leave place for the other processes to spawn and run within
the container and the container crashes.

pids-limit set to unlimited regardless of the container engine.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1987041
Signed-off-by: Teoman ONAY <tonay@redhat.com>
(cherry picked from commit 9b5d97adb95a788bc1fdedbba562a9c71a1808be)
(cherry picked from commit 3d4e15cebf13407a410f06ac16d7a92d418d27a5)

2 years agocommon: do not log keyring secret v4.0.62.2
Guillaume Abrioux [Mon, 9 Aug 2021 12:57:33 +0000 (14:57 +0200)]
common: do not log keyring secret

let's not display any keyring secret by default in ansible log.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1980744
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 7511195738e9d1e8f3d3ec77ad4473fa90d17d22)

2 years agoswitch2container: fix mon quorum check
Dimitri Savineau [Mon, 9 Aug 2021 19:41:40 +0000 (15:41 -0400)]
switch2container: fix mon quorum check

This was reverted by 7ddbe74

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1990733
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 712a9c44033a9224cdeec5467a9d50ffb78e9d1b)

2 years agoceph-dashboard: fix TLS cert openssl generation
Dimitri Savineau [Mon, 9 Aug 2021 14:33:40 +0000 (10:33 -0400)]
ceph-dashboard: fix TLS cert openssl generation

With OpenSSL version prior 1.1.1 (like CentOS 7 with 1.0.2k), the -addext
doesn't exist.
As a solution, this uses the default openssl.cnf configuration file as a
template and add the subjectAltName in the v3_ca section. This temp openssl
configuration file is removed after the TLS certificate creation.
This patch also move the run_once statement at the block level.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1978869
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 5e0ace7e5493f7d8299155e915435691a0f1a007)

2 years agodashboard: subj_alt_names fact refactor
Guillaume Abrioux [Thu, 5 Aug 2021 13:00:49 +0000 (15:00 +0200)]
dashboard: subj_alt_names fact refactor

the current way the variable is built results in:

```
2021-08-03 04:18:23,020 - ceph.ceph - INFO - ok: [ceph-sangadi-4x-indpt6-node1-installer] => changed=false
  ansible_facts:
    subj_alt_names: |-
      subjectAltName=ceph-sangadi-4x-indpt6-node1-installer/subjectAltName=10.0.210.223/subjectAltName=ceph-sangadi-4x-indpt6-node1-installersubjectAltName=ceph-sangadi-4x-indpt6-node2/subjectAltName=10.0.210.252/subjectAltName=ceph-sangadi-4x-indpt6-node2/
```

which is incorrect.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1978869
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 6f1a0634f73ad1f41af613a9452dc9c5f70b2702)

2 years agoinfra: use dedicated variables for balancer status v4.0.62.1
Dimitri Savineau [Tue, 3 Aug 2021 15:58:49 +0000 (11:58 -0400)]
infra: use dedicated variables for balancer status

The balancer status is registered during the cephadm-adopt, rolling_update
and swith2container playbooks. But it is also used in the ceph-handler role
which is included in those playbooks too.
Even if the ceph-handler tasks are skipped for rolling_update and
switch2container, the balancer_status variable is erased with the skip task
result.

play1:
  register: balancer_status
play2:
  register: balancer_status <-- skipped
play3:
  when: (balancer_status.stdout | from_json)['active'] | bool

This leads to issue like:

The conditional check '(balancer_status.stdout | from_json)['active'] | bool'
failed. The error was: Unexpected templating type error occurred on
({% if (balancer_status.stdout | from_json)['active'] | bool %} True
{% else %} False {% endif %}): expected string or buffer.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1982054
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 386661699bcfe05a220de6d58b9d50baa7eb6dc1)

2 years agocommon: remove unnecessary run_once statements v4.0.62
Dimitri Savineau [Tue, 20 Jul 2021 15:38:44 +0000 (11:38 -0400)]
common: remove unnecessary run_once statements

1303611 introduced tasks for disabling the pg_autoscaler on pools and
the balancer but thoses tasks are already executed on the first monitor
node so we don't need to add the run_once statement.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 738fa9428a3c0c9ed0dd1aec566e807ad072daad)

2 years agocommon: fix py2 pool_list from_json when skipped
Dimitri Savineau [Tue, 20 Jul 2021 19:53:48 +0000 (15:53 -0400)]
common: fix py2 pool_list from_json when skipped

When using python 2 and the task with a loop is skipped then it generates
an error.

Unexpected templating type error occurred on
({{ (pool_list.stdout | from_json)['pools'] }}): expected string or buffer

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit cf6e33346e09e680d9b69cafdc24fdd8801cdbf4)

2 years agocommon: disable/enable pg_autoscaler
Guillaume Abrioux [Mon, 14 Jun 2021 16:01:41 +0000 (18:01 +0200)]
common: disable/enable pg_autoscaler

The PG autoscaler can disrupt the PG checks so the idea here is to
disable it and re-enable it back after the restart is done.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 13036115e2862fc8ca9c04e1379fd793e0e7036a)

2 years agoSensitive key data now hidden in output log v4.0.61
Neelaksh Singh [Thu, 20 May 2021 06:04:02 +0000 (02:04 -0400)]
Sensitive key data now hidden in output log

Fixes: #6529
Signed-off-by: Neelaksh Singh <neelaksh48@gmail.com>
(cherry picked from commit d18a9860cde2981bcd71198f152924cc6cf05932)

2 years agoupdate: fail the playbook if straw2 conversion failed
Guillaume Abrioux [Fri, 9 Jul 2021 14:29:09 +0000 (16:29 +0200)]
update: fail the playbook if straw2 conversion failed

It's better to fail the playbook so the user is aware the straw2
migration has failed.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit c396122ad9acda19462c7da363696e174b8162d1)

2 years agoupdate: followup on pr #6689
Guillaume Abrioux [Fri, 9 Jul 2021 07:19:52 +0000 (09:19 +0200)]
update: followup on pr #6689

add mising 'osd' command.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 4eb4268deea262d09898d21fa815912439cecfa5)

2 years agoupdate: convert straw bucket
Guillaume Abrioux [Thu, 8 Jul 2021 19:57:13 +0000 (21:57 +0200)]
update: convert straw bucket

After an upgrade, the presence of straw buckets will produce the
following warning (HEALTH_WARN):

```
crush map has legacy tunables (require firefly, min is hammer)
```

because straw bucket is a firefly feature it needs to be converted to
straw2.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1967964
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit eee576477c73d2585a824814e3b4d85f0c12e8fd)

2 years agoRevert "ceph-validate: check devices from lvm_volumes"
Dimitri Savineau [Tue, 6 Jul 2021 15:52:15 +0000 (11:52 -0400)]
Revert "ceph-validate: check devices from lvm_volumes"

This reverts commit 3557497336338ab2227bcb7c2373047291b4bc73.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
2 years agoRevert "ceph-validate: check block presence first"
Dimitri Savineau [Tue, 6 Jul 2021 15:52:14 +0000 (11:52 -0400)]
Revert "ceph-validate: check block presence first"

This reverts commit 4f89cdcd457debac5fbcd4e4e8ed22f6fb20e771.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
2 years agoRevert "ceph-validate: do not resolve devices"
Dimitri Savineau [Tue, 6 Jul 2021 15:52:14 +0000 (11:52 -0400)]
Revert "ceph-validate: do not resolve devices"

This reverts commit 2020b1310c1717c807a9baf9ef7be92bdc9c6a11.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
2 years agoRevert "ceph-validate: use root device from ansible_mounts"
Dimitri Savineau [Tue, 6 Jul 2021 15:52:13 +0000 (11:52 -0400)]
Revert "ceph-validate: use root device from ansible_mounts"

This reverts commit b1542fd3403b18eaeba7b336d5225047b1560027.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
2 years agoRevert "ceph-validate: check db/journal/wal devices too"
Dimitri Savineau [Tue, 6 Jul 2021 15:52:13 +0000 (11:52 -0400)]
Revert "ceph-validate: check db/journal/wal devices too"

This reverts commit d6f3e6eac35a604a41fc397fdb1c4aaca27b4749.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
2 years agoRevert "ceph-validate: check logical volumes"
Dimitri Savineau [Tue, 6 Jul 2021 15:52:11 +0000 (11:52 -0400)]
Revert "ceph-validate: check logical volumes"

This reverts commit d7cefe0536f4ac3c1dd9010904afa5f3ebb852d4.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
2 years agoRevert "ceph-facts: move device facts to its own file"
Dimitri Savineau [Tue, 6 Jul 2021 15:52:10 +0000 (11:52 -0400)]
Revert "ceph-facts: move device facts to its own file"

This reverts commit 9f1ec38bbf9fbc341f53c080f6cec7a268cc13b1.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
2 years agodashboard: remove "certificate is valid for" error
Guillaume Abrioux [Tue, 6 Jul 2021 12:18:51 +0000 (14:18 +0200)]
dashboard: remove "certificate is valid for" error

When deploying dashboard with ssl certificates generated by
ceph-ansible, we enforce the CN to 'ceph-dashboard' which can makes
application such alertmanager complain like following:

`err="Post https://mgr0:8443/api/prometheus_receiver: x509: certificate is valid for ceph-dashboard, not mgr0" context_err="context deadline exceeded"`

The idea here is to add alternative names matching all mgr/mon instances
in the certificate so this error won't appear in logs.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1978869
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 72a0336c71cee8bd0a375ac47cb45a292844edc8)

2 years agoceph-crash: add install checkpoint
Dimitri Savineau [Mon, 5 Jul 2021 14:11:57 +0000 (10:11 -0400)]
ceph-crash: add install checkpoint

The ceph crash insatll checkpoint callback was missing in the main
playbooks.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 993d06c4d92f8a0be6441dd7647a27d6b58bcb9b)

2 years agoceph-facts: move device facts to its own file
Dimitri Savineau [Wed, 16 Dec 2020 19:18:08 +0000 (14:18 -0500)]
ceph-facts: move device facts to its own file

Instead of reusing the condition 'inventory_hostname in groups[osds]'
on each device facts tasks then we can move all the tasks into a
dedicated file and set the condition on the import_tasks statement.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit d704b05e52d10910cd68c49033933bd7e6ded268)

2 years agoceph-validate: check logical volumes
Dimitri Savineau [Tue, 15 Dec 2020 22:34:34 +0000 (17:34 -0500)]
ceph-validate: check logical volumes

We currently don't check if the logical volume used in lvm_volumes list
for either bluestore data/db/wal or filestore data/journal exist.
We're only doing this on raw devices for batch scenario.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 55bca07cb612b766bc099e14e0a5661185a7f9a6)

2 years agoceph-validate: check db/journal/wal devices too
Dimitri Savineau [Tue, 15 Dec 2020 20:08:00 +0000 (15:08 -0500)]
ceph-validate: check db/journal/wal devices too

When using dedicated devices for db/journal/wal objecstore with
ceph-volume lvm batch then we should also validate that those devices
exist and don't use a gpt partition table in addition of the devices
and lvm_volume.data variables.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 808e7106dec5f3f7a743fe343ba3023c9390a1ba)

2 years agoceph-validate: use root device from ansible_mounts
Dimitri Savineau [Tue, 15 Dec 2020 20:04:57 +0000 (15:04 -0500)]
ceph-validate: use root device from ansible_mounts

Instead of using findmnt command to find the device associated to the
root mount point then we can use the ansible_mounts fact.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 7e50380f7fc101bf3533fff2d9c3082a59c54399)

2 years agoceph-validate: do not resolve devices
Dimitri Savineau [Tue, 15 Dec 2020 20:02:59 +0000 (15:02 -0500)]
ceph-validate: do not resolve devices

This is already done in the ceph-facts role.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 0df99dda8d8c5e1e9ce4384b480401043eaad07e)

2 years agoceph-validate: check block presence first
Dimitri Savineau [Tue, 15 Dec 2020 20:00:28 +0000 (15:00 -0500)]
ceph-validate: check block presence first

Instead of doing two parted calls we can check first if the device exist
and then test the partition table.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 14d458b3b41ef5fe1ebdb8a70e52a65ff28b42f2)

2 years agoceph-validate: check devices from lvm_volumes
Dimitri Savineau [Tue, 15 Dec 2020 19:49:57 +0000 (14:49 -0500)]
ceph-validate: check devices from lvm_volumes

2888c08 introduced a regression as the check_devices tasks file was
only included based on the devices variable.
But that file also validate some devices from the lvm_volumes variable.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1906022
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit ac0342b72e045b96a31bed23c6baeceeecca7f23)

2 years agoprometheus: fix prometheus target url v4.0.60
Dimitri Savineau [Fri, 2 Jul 2021 13:13:43 +0000 (09:13 -0400)]
prometheus: fix prometheus target url

The prometheus service isn't binding on localhost.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1933560
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 1d56818658ebba36cce5f93fa4c5e2bcac310561)

2 years agopurge: add monitoring group in final cleanup play
Guillaume Abrioux [Fri, 2 Jul 2021 12:57:52 +0000 (14:57 +0200)]
purge: add monitoring group in final cleanup play

This adds the monitoring group in the "final cleanup play" so any cid
files generated are well removed when purging the cluster.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1974536
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 037d8cd05e012dd71a6c0aedbdfcf4aa1c15b992)

2 years agoceph_key: handle error in a better way v4.0.59
Guillaume Abrioux [Mon, 7 Jun 2021 12:51:43 +0000 (14:51 +0200)]
ceph_key: handle error in a better way

When calling the `ceph_key` module with `state: info`, if the ceph
command called fails, the actual error is hidden by the module which
makes it pretty difficult to troubleshoot.

The current code always states that if rc is not equal to 0 the keyring
doesn't exist.

`state: info` should always return the actual rc, stdout and stderr.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1964889
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit d58500ade01c7c4cc20b6bbaf8f5789d825b8d1b)

2 years agocontainer: set tcmalloc value by default
Dimitri Savineau [Tue, 29 Jun 2021 17:24:29 +0000 (13:24 -0400)]
container: set tcmalloc value by default

All ceph daemons need to have the TCMALLOC_MAX_TOTAL_THREAD_CACHE_BYTES
environment variable set to 128MB by default in container setup.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1970913
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit 9758e3c51303faf89d5c489f04539e278c49477e)

2 years agodashboard: Add new prometheus alert
Boris Ranto [Tue, 8 Jun 2021 07:43:23 +0000 (09:43 +0200)]
dashboard: Add new prometheus alert

It was requested for us to update our alerting definitions to include a
slow OSD Ops health check.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1951664
Signed-off-by: Boris Ranto <branto@redhat.com>
(cherry picked from commit 2491d4e004c7b162216bc17e2288f05d0b049a87)

2 years agoupdate: do not gather facts on each play
Guillaume Abrioux [Fri, 14 May 2021 14:24:55 +0000 (16:24 +0200)]
update: do not gather facts on each play

There's no benefit to gather facts again on each play in
rolling_update.yml

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 2c77d0094c9bab02dee32a144faf5168262049df)

2 years agotests: remove legacy file
Guillaume Abrioux [Mon, 28 Jun 2021 16:05:26 +0000 (18:05 +0200)]
tests: remove legacy file

This inventory isn't used anywhere.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 304d1cbb97e52f1546f98171ba7b7b6a658c85c1)

2 years agoshrink-mgr: modify existing mgr check
Guillaume Abrioux [Mon, 28 Jun 2021 18:16:03 +0000 (20:16 +0200)]
shrink-mgr: modify existing mgr check

Do not rely on the inventory aliases in order to check if the selected
manager to be removed is present.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1967897
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 26a7256c4cec17fc9bc20a239c112cbf8584be09)

2 years agoworkflows: add signed-off check
Guillaume Abrioux [Tue, 29 Jun 2021 22:24:01 +0000 (00:24 +0200)]
workflows: add signed-off check

This adds a github workflow for checking the signed off line in commit
messages.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 8c09497567308bac75212be77cdb7b1928d2d9a2)

2 years agoworkflow: add group_vars/defaults checks
Guillaume Abrioux [Tue, 29 Jun 2021 19:06:37 +0000 (21:06 +0200)]
workflow: add group_vars/defaults checks

let's use github workflow for checking defaults values.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit d71db816c6d393d6a42b934912f4f47a290c91fe)

2 years agoworkflow: add syntax check
Guillaume Abrioux [Tue, 29 Jun 2021 18:47:33 +0000 (20:47 +0200)]
workflow: add syntax check

This adds the ansible --syntax-check test in the ansible-lint workflow

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 5ed423ad8886a8985a4aa6424930f812864845bc)

2 years agoworkflow/lint: pin ruamel.yaml.clib to 0.2.2
Dimitri Savineau [Tue, 29 Jun 2021 15:33:23 +0000 (11:33 -0400)]
workflow/lint: pin ruamel.yaml.clib to 0.2.2

Recent release is broken with py27.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
2 years agotox: add ceph_stable_release to switch2container
Dimitri Savineau [Tue, 29 Jun 2021 15:28:39 +0000 (11:28 -0400)]
tox: add ceph_stable_release to switch2container

We need to set the ceph_stable_release variable during the switch2container
playbook.

Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
2 years agoswitch2container: run ceph-validate role
Dimitri Savineau [Mon, 28 Jun 2021 14:46:40 +0000 (10:46 -0400)]
switch2container: run ceph-validate role

This adds the ceph-validate role before starting the switch to a containerized
deployment.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1968177
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
(cherry picked from commit fc160b3be18edd1615ebb2b813a8468748f6d7d3)

2 years agonfs: do no copy client.bootstrap-rgw when using mds v4.0.58
Guillaume Abrioux [Tue, 15 Jun 2021 09:02:05 +0000 (11:02 +0200)]
nfs: do no copy client.bootstrap-rgw when using mds

There's no need to copy this keyring when using nfs with mds

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 8dbee99882968005e744d3c8199dbfdea17495cc)

2 years agoContainer: Fixing service name lvm2-lvmetad
VasishtaShastry [Mon, 14 Jun 2021 12:08:51 +0000 (17:38 +0530)]
Container: Fixing service name lvm2-lvmetad

Playbook failing saying:

msg: 'Could not find the requested service lvmetad: host'

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1955040
Signed-off-by: VasishtaShastry <vipin.indiasmg@gmail.com>
(cherry picked from commit e49c38f8b7fc7c6c8ef06173874ece802d58102d)

2 years agomultisite: fix bug during switch2containers
Guillaume Abrioux [Wed, 16 Jun 2021 07:39:18 +0000 (09:39 +0200)]
multisite: fix bug during switch2containers

When running the switch-to-containers playbook with multisite enabled,
the fact "rgw_instances" is only set for the node being processed
(serial: 1), the consequence of that is that the set_fact of
'rgw_instances_all' can't iterate over all rgw node in order to look up
each 'rgw_instances_host'.

Adding a condition checking whether hostvars[item]["rgw_instances_host"]
is defined fixes this issue.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1967926
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 8279d14d328ad4bacafc01847f3ad9f385b4a3d5)

2 years agorolling_update: fix mon+rgw/multisite collocation
Guillaume Abrioux [Thu, 10 Jun 2021 18:09:55 +0000 (20:09 +0200)]
rolling_update: fix mon+rgw/multisite collocation

When monitors and rgw are collocated with multisite enabled, the
rolling_update playbook fails because during the workflow, we run some
radosgw-admin commands very early on the first mon even though this is
the monitor being upgraded, it means the container doesn't exist since
it was stopped.

This block is relevant only for scaling out rgw daemons or initial
deployment. In rolling_update workflow, it is not needed so let's skip
it.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1970232
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit f7166cccbfa206f07a1ecca8d3eba0fe9d396c3a)

2 years agotests: disable test_mgr_dashboard_is_listening v4.0.57
Guillaume Abrioux [Fri, 4 Jun 2021 10:45:47 +0000 (12:45 +0200)]
tests: disable test_mgr_dashboard_is_listening

Due to a recent commit that has introduced a regression in ceph, this
test is failing.
Temporarily disabling it to unblock the CI.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 2e19d1705efa8ef247eecb540357195d5142da77)

2 years agodashboard: set cookie_secure in grafana
Guillaume Abrioux [Wed, 2 Jun 2021 09:05:07 +0000 (11:05 +0200)]
dashboard: set cookie_secure in grafana

When using grafana behind https `cookie_secure` should be set to `true`.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1966880
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 4daed1f137a4c61940abd9ae2db179fafd8de17a)

2 years agodashboard: fix rgw user creation
Guillaume Abrioux [Thu, 3 Jun 2021 08:11:30 +0000 (10:11 +0200)]
dashboard: fix rgw user creation

When deploying dashboard in a cluster with rgw multisite deployed.
Due to the last rgw multisite refactor, we now expect the variable
`rgw_zonemaster` to be defined in the dict `rgw_instances`.
The idea here is to create that user on the cluster as soon as we have 1
`rgw_zonemaster` set to `true` in `rgw_instances`.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1964995
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
2 years agocrash: fix --limit deployments (containers) v4.0.56
Guillaume Abrioux [Tue, 25 May 2021 13:06:41 +0000 (15:06 +0200)]
crash: fix --limit deployments (containers)

ceph-crash deployments is broken when ceph-ansible playbook is called
with --limit in containerized contexts since we don't set
`container_exec_cmd` on the first monitor.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1964835
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 10ed26f14dc38211d4c77cbeeeb681d215a351b7)

2 years agodashboard: fix typo introduced during backport
Guillaume Abrioux [Wed, 26 May 2021 10:04:20 +0000 (12:04 +0200)]
dashboard: fix typo introduced during backport

during backport of c8b92deba10c0b6e0ebcb0e31315b1e6174fdc0c the pattern
should have been s/monitoring_group_name/grafana_server_group_name/

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1964907
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit ac0a5c1e682e0a6e6c67d7c3c630b1da110785bb)

2 years agoprometheus: enforce osd nodes in templates
Guillaume Abrioux [Tue, 25 May 2021 07:26:28 +0000 (09:26 +0200)]
prometheus: enforce osd nodes in templates

When osd nodes are collocated in the clients group (HCI context for
instance), the current logic will exclude osd nodes since they are
present in the client group.

The best fix would be to exclude clients node only when they are not
member of another group but for now, as a workaround, we can enforce
the addition of osd nodes to fix this specific case.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1947695
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 664dae0564d54d6df873b96264e59fa38c8280a4)

2 years agofs2bs: use match filter in selectattr()
Guillaume Abrioux [Tue, 25 May 2021 19:40:45 +0000 (21:40 +0200)]
fs2bs: use match filter in selectattr()

0990ae41099d8b9d678f546fc8790033f421711f changed the filter in
selectattr() from 'match' to 'equalto' but due to an incompatibility with
the Jinja2 version for python 2.7 on el7 we must stick to using 'match'
filter.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit d6745e9cd9c2b5f0e20280e4ae0fe1b125287b0c)

2 years agofs2bs: fix wrong filter when setting osd_ids
Guillaume Abrioux [Tue, 25 May 2021 14:01:27 +0000 (16:01 +0200)]
fs2bs: fix wrong filter when setting osd_ids

using 'match' filter in that task will lead to bad behavior if I have
the following node names for instance:

- node1
- node11
- node111

with `selectattr('name', 'match', inventory_hostname)` it will match
'node1' along with 'node11' and 'node111'.

using 'equalto' filter will make sure we only match the target node.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1963066
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 0990ae41099d8b9d678f546fc8790033f421711f)

2 years agocontainer: conditionnally disable lvmetad
Guillaume Abrioux [Fri, 21 May 2021 11:25:25 +0000 (13:25 +0200)]
container: conditionnally disable lvmetad

Enabling lvmetad in containerized deployments on el7 based OS might
cause issues.
This commit make it possible to disable this service if needed.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1955040
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
2 years agoMake sure the repo url contains the correct arch
Brad Hubbard [Fri, 19 Mar 2021 04:11:42 +0000 (14:11 +1000)]
Make sure the repo url contains the correct arch

We can end up with an arm only repo unless we are specific about the
architecture we require. Brings the deb code in line with the rpm
equivalent.

Signed-off-by: Brad Hubbard <bhubbard@redhat.com>
(cherry picked from commit 267cce9e8360fc8cb9c192fde2406e5dca724610)

2 years agovalidate: check virtual_ips variable v4.0.55
Guillaume Abrioux [Mon, 11 Jan 2021 09:06:08 +0000 (10:06 +0100)]
validate: check virtual_ips variable

This commit checks the length of `virtual_ips` doesn't exceed the length
of `groups[rgwloadbalancer_group_name]`.
It also ensure this variable is defined when
`groups[rgwloadbalancer_group_name]` contains at least one node.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 3b63e0649cdaf322af1ad5a5b5b276ec9085d2d5)

2 years agoceph-rgw-loadbalancer: Fix keepalived master selection
Benoît Knecht [Mon, 1 Jun 2020 15:09:18 +0000 (17:09 +0200)]
ceph-rgw-loadbalancer: Fix keepalived master selection

While 2ca33641 fixed a bug in the way the `keepalived.conf.j2` template matched
hostnames to set the VRRP `MASTER`/`BACKUP` states, it also introduced a
regression in the case where `virtual_ips` is a list of more than one IP
address.

The previous behavior would result in each host in the `rgwloadbalancers` group
to be `MASTER` for one of the `virtual_ips`, but the new behavior caused the
first host to be `MASTER` for all the IP address in `virtual_ips`.

This commit restores the original behavior.

Signed-off-by: Benoît Knecht <bknecht@protonmail.ch>
(cherry picked from commit 2bede4762efaf566a7da689dd1021ee66c0c94eb)

2 years agoupdate: fix ceph-crash stop task
Guillaume Abrioux [Tue, 4 May 2021 09:42:15 +0000 (11:42 +0200)]
update: fix ceph-crash stop task

This is a workaround for an issue in ansible.
When trying to stop/mask/disable this service in one task, the stop
didn't actually happen, the task doesn't fail but for some reason the
container is still present and running.
Then the task starting the service in the role ceph-crash fails because
it can't start the container since it's already running with the same
name.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1955393
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 3db1ea7ec41475243559beff50b7011773c021e5)

2 years agoceph-osd: allow to use ceph_tcmalloc_max_total_thread_cache for bluestore v4.0.54
Seena Fallah [Wed, 14 Oct 2020 13:00:57 +0000 (16:30 +0330)]
ceph-osd: allow to use ceph_tcmalloc_max_total_thread_cache for bluestore

TCMALLOC_MAX_TOTAL_THREAD_CACHE_BYTES is for both bluestore and filestore

Signed-off-by: Seena Fallah <seenafallah@gmail.com>
(cherry picked from commit 41295f0ef661511953be607d44d82a80f5656d2e)

2 years agoceph-mon: Fix check mode for deploy monitor tasks
Benoît Knecht [Mon, 19 Oct 2020 09:23:59 +0000 (11:23 +0200)]
ceph-mon: Fix check mode for deploy monitor tasks

Skip the `get initial keyring when it already exists` task when both commands
whose `stdout` output it requires have been skipped (e.g. when running in check
mode).

Signed-off-by: Benoît Knecht <bknecht@protonmail.ch>
(cherry picked from commit e98d9b70bd2d272385519efdefe776c4abc00f0a)

2 years agofs2bs: add a final play
Guillaume Abrioux [Tue, 13 Apr 2021 12:40:48 +0000 (14:40 +0200)]
fs2bs: add a final play

This removes the fact `skipped_nodes` which is useless when we run with
`--limit` since it gets reset when a new iteration is made.

Instead, let's print within a final play which node has been skipped
reusing the `skip_this_node` fact.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 3d4267051f03c053dcd7d3f6f381988be0b2939d)

2 years agoConfig the monitoring stack components api urls using a VIP
Francesco Pantano [Fri, 9 Apr 2021 15:16:03 +0000 (17:16 +0200)]
Config the monitoring stack components api urls using a VIP

When dashboard_frontend_vip is provided, all the services should be
configured using the related VIP. A new VIP variable is added for
both prometheus and alertmanager: we're already able to properly
config the grafana vip using dashboard_frontend_vip variable.
This change adds the same variable for both prometheus and
alertmanager.

Signed-off-by: Francesco Pantano <fpantano@redhat.com>
(cherry picked from commit 441651638d883b73c960363becae54096c8cee85)

3 years agoosd: always allow setting target_size_ratio
Guillaume Abrioux [Thu, 15 Apr 2021 11:16:31 +0000 (13:16 +0200)]
osd: always allow setting target_size_ratio

We shouldn't prevent from setting target_size_ratio when the autoscaler
is set to 'warn'.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1906305
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
3 years agoceph-rgw-loadbalancer: Fix rgw_ports fact
Benoît Knecht [Tue, 13 Apr 2021 11:57:44 +0000 (13:57 +0200)]
ceph-rgw-loadbalancer: Fix rgw_ports fact

The `set_fact rgw_ports` task was failing due to a templating error, because
`hostvars[item].rgw_instances` is a list, but it was treated as if it was a
dictionary.

Another issue was the fact that the `unique` filter only applied to the list
being appended to `rgw_ports` instead of the entire list, which means it was
possible to have duplicate items.

Lastly, `rgw_ports` would have been a list of integers, but the `seport` module
expects a list of strings.

This commit fixes all of the issues above, allowing the `ceph-rgw-loadbalancer`
role to work on systems with SELinux enabled.

Signed-off-by: Benoît Knecht <bknecht@protonmail.ch>
(cherry picked from commit c07851347521b09c39b556818378767fc23f05e5)

3 years agoswitch-to-containers: only chown corresponding files v4.0.53
Guillaume Abrioux [Fri, 9 Apr 2021 07:24:05 +0000 (09:24 +0200)]
switch-to-containers: only chown corresponding files

When collocating daemons, if we chown all files under `/var/lib/ceph` it
can cause issues for the collocated daemons that wouldn't have been
migrated yet.

This commit makes the playbook chown only the files corresponding to the
daemon being migrated.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit ddbc11c4a9c815ccc46fa52fd900f452c082612f)

3 years agocontainer/systemd: ensure /var/log/ceph exists
Guillaume Abrioux [Wed, 14 Apr 2021 11:41:59 +0000 (13:41 +0200)]
container/systemd: ensure /var/log/ceph exists

This adds a `ExecStartPre=-/usr/bin/mkdir -p /var/log/ceph` in all
systemd service templates for all ceph daemon.
This is specific to RHCS after a Leapp upgrade is done. Indeed, the
`/var/log/ceph` seems to be removed after the upgrade.
In order to work around this issue let's ensure the directory is present
before trying to start the containers with podman.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1949489
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit bab403b603feb8b7534dde65924699b370f4bc0d)

3 years agorbdmirror: add retries/until when configuring mirroring
Guillaume Abrioux [Mon, 12 Apr 2021 20:29:50 +0000 (22:29 +0200)]
rbdmirror: add retries/until when configuring mirroring

`configure_mirroring.yml` is called right after the daemon is started.
Sometimes, it can happen the first task in `configure_mirroring.yml` is
run while the daemon isn't yet ready, adding a retries/until on that
task should help to avoid causing the playbook to fail.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1944996
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit b1e7e1ad0f517c345ade159961d8bb2cb13b1c22)

3 years agodocker2podman: skip some role imports from handler v4.0.52
Guillaume Abrioux [Fri, 9 Apr 2021 09:02:12 +0000 (11:02 +0200)]
docker2podman: skip some role imports from handler

when running docker-to-podman playbook, there's no need to call
`ceph-config` and `ceph-rgw` from the role `ceph-handler`.
It can even have side effects when coming from a baremetal cluster that
was previously migrated using the switch-to-containers playbook. Indeed
it might complain about missing .target systemd unit since they are
removed during that migration.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1944999
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 70f19be367390aeca43340711d380ce4825e02b5)

3 years agodocker2podman: add documentation/header
Guillaume Abrioux [Mon, 12 Apr 2021 07:25:25 +0000 (09:25 +0200)]
docker2podman: add documentation/header

this adds a small documentation in the header of the playbook in order
to explain what is the goal of this playbook.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 36b4227dcd79c2525102f8c3102771cfab7fe693)

3 years agoswitch_to_containers: support iscsigws migration
Guillaume Abrioux [Wed, 7 Apr 2021 12:36:28 +0000 (14:36 +0200)]
switch_to_containers: support iscsigws migration

This adds the iscsigws migration to containers.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=<bz-number>
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 2c74c2732113ca2ef3e98361c64e1748790828b0)

3 years agocommon: selinux tasks related refactor
Guillaume Abrioux [Thu, 1 Apr 2021 12:37:45 +0000 (14:37 +0200)]
common: selinux tasks related refactor

This moves some task from the `ceph-nfs` role in `ceph-common` since
some of them are needed in `ceph-rgwloadbalancer` role.
This avoids duplicated tasks.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit d0442d81b9e8087ae8ec086798c68382680a7d73)

3 years agorgw-loadbalancers: add all rgw_ports to http_port_t type
Guillaume Abrioux [Thu, 1 Apr 2021 07:50:54 +0000 (09:50 +0200)]
rgw-loadbalancers: add all rgw_ports to http_port_t type

This adds all rgw ports to the http_port_t selinux type so it
allows haproxy to connect to those ports in order to avoid AVC.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1923890
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 6bbb90198b600b460a34faf322a9e36646b3689e)

3 years agorgw-loadbalancer: Update haproxy.cfg.j2
kalebskeithley [Tue, 9 Mar 2021 21:10:35 +0000 (16:10 -0500)]
rgw-loadbalancer: Update haproxy.cfg.j2

haproxy gets an AVC when configured to connect to port 8081

This commit adds a snippet regarding haproxy in a selinux environment

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1923890
Signed-off-by: Kaleb S KEITHLEY <kkeithle@redhat.com>
(cherry picked from commit 9e7f22a071c0f7aecd98baeb73a1f90e290ef039)

3 years agonfs: set idmap config for Ceph-NFS
Guillaume Abrioux [Fri, 12 Mar 2021 15:27:12 +0000 (16:27 +0100)]
nfs: set idmap config for Ceph-NFS

Currently NFS Ganesha (ceph-nfs) consumes /etc/idmapd.conf, which
controls mapping of user/owner identities under NFSv4+. With
containerized service deployment, this file is an immutable part of the
container image and cannot be modified.

Here we provide group variables, and a taskk and templates for the
ceph-nfs role, to set the path of the idmap configuration file and
to make the most common adjustment to the contents of that file --
namely to set the 'Domain'. We default the path to /etc/ganesha/idmap.conf
so that we will not conflict with /etc/idmapd.conf on the controller nodes
where ganesha runs. NFSv4 clients, as used for example by the Cinder NFS
driver, consume /etc/idmapd.conf and may require different settings than
what is wanted for NFS Ganesha. Additionally, because we already bind
/etc/ganesha from the host into the ceph-nfs container, the file NFS
Ganesha consumes will no longer be an immutable part of the container.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1925646
Signed-off-by: Tom Barron tpb@dyncloud.net
Co-Authored-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 2db2208e406df83806c264207e7df90623add154)

3 years agodashboard: support prometheus storage.tsdb.retention.time parameter
Guillaume Abrioux [Tue, 30 Mar 2021 07:49:10 +0000 (09:49 +0200)]
dashboard: support prometheus storage.tsdb.retention.time parameter

This commit adds the parameter `--storage.tsdb.retention.time` to the
prometheus systemd unit template.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1928000
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit b60c61ce450db5b755744c491b4dfd1ce70fb124)

3 years agoupdate: followup on 07029e1 v4.0.51
Guillaume Abrioux [Fri, 26 Mar 2021 15:12:42 +0000 (16:12 +0100)]
update: followup on 07029e1

Playbook must fail anyway, the `rescue` block has been introduced for
unmasking the unit after the playbook has failed.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit e9ddb972fef5ab73e3a6f4f95f451ae223dcf64c)

3 years agorolling_update: unmask monitor service after a failure
Guillaume Abrioux [Thu, 18 Mar 2021 08:08:51 +0000 (09:08 +0100)]
rolling_update: unmask monitor service after a failure

if for some reason the playbook fails after the service was
stopped, disabled and masked and before it got restarted, enabled and
unmasked, the playbook leaves the service masked and which can make users
confused and forces them to unmask the unit manually.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1917680
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 07029e1bf1880dedd5007ad09328ef7e2c1a85f7)

3 years agodefaults: add a comment about `igw_network`
Guillaume Abrioux [Fri, 26 Mar 2021 17:05:48 +0000 (18:05 +0100)]
defaults: add a comment about `igw_network`

This add a quick documentation in ceph-defaults about `igw_network`

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit c5728bdc6351fe985e17b095cbd5871ef673dc29)

3 years agodashboard: support igw nodes with dedicated subnet
Guillaume Abrioux [Tue, 2 Mar 2021 08:51:26 +0000 (09:51 +0100)]
dashboard: support igw nodes with dedicated subnet

This adds the possibility to deploy the dashboard with igw nodes using
a dedicated subnet.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1926170
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit c33de174f1d8ec2d4b94a4322cdcdcf25262f1f6)

3 years agoPeer addition won't be skipped if remote is not in peer
VasishtaShastry [Wed, 24 Mar 2021 15:47:56 +0000 (21:17 +0530)]
Peer addition won't be skipped if remote is not in peer
rbd-mirroring is not configured as adding peer is getting skipped.
Peer addition should not get skipped if its not added already

Closes - https://bugzilla.redhat.com/show_bug.cgi?id=1942444

Signed-off-by: VasishtaShastry <vipin.indiasmg@gmail.com>
(cherry picked from commit 006998e804f803f3d3a6767403803a9cad1ed9b0)

3 years agoupdate: stop ceph-crash service before upgrading
Guillaume Abrioux [Fri, 26 Mar 2021 09:31:06 +0000 (10:31 +0100)]
update: stop ceph-crash service before upgrading

This adds the missing service stop task for ceph-crash upgrade workflow.

It should have been added through commit
`15872e3db1e342238636bc9c8e1aef6bd1d3dcd8` in stable-4.0 but at the time
we backported this patch ceph-crash wasn't implemented yet so the
ceph-crash related content in this patch was removed. Then, ceph-crash
has been implemented later so we are still missing this part of the patch in
stable-4.0.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1943471
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
3 years agotests: pin ruamel.yaml version
Guillaume Abrioux [Fri, 26 Mar 2021 13:25:08 +0000 (14:25 +0100)]
tests: pin ruamel.yaml version

0.17.0 which was released today (03/26/2021) breaks ansible-lint execution with
py2.7.

From https://pypi.org/project/ruamel.yaml we can read:

> The 0.16.13 release was the last that will tested to be working on Python 2.7.

Let's enforce the version on 0.16.13 when running with py2.7

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
3 years agodocs: rgw multisite docs with new rgw_instances config
Ali Maredia [Mon, 25 Jan 2021 06:33:54 +0000 (01:33 -0500)]
docs: rgw multisite docs with new rgw_instances config

Docs reflect that each instance of `rgw_instances`
can now take rgw_zonemaster, rgw_zonesecondary,
rgw_zonegroupmaster, rgw_multisite_proto.

Signed-off-by: Ali Maredia <amaredia@redhat.com>
(cherry picked from commit a59bc2da3b82fcc5d6ce413350c74654fecba381)

3 years agoconvert some missed `ansible_*`` calls to `ansible_facts['*']`
Guillaume Abrioux [Thu, 25 Mar 2021 02:47:40 +0000 (03:47 +0100)]
convert some missed `ansible_*`` calls to `ansible_facts['*']`

This converts some missed calls to `ansible_*` that were missed in
initial PR #6312

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 0163ecc924ccec3e5da9a9c4cfb8ac4558ec33ef)

3 years agoDisable facts by default in ansible.cfg
Alex Schultz [Thu, 11 Mar 2021 15:43:53 +0000 (08:43 -0700)]
Disable facts by default in ansible.cfg

As a continuation of a7f2fa73e63e69dba2e41aaac9732397eec437c9, this
change switches fact injection to off by default in the provided
ansible.cfg.

Signed-off-by: Alex Schultz <aschultz@redhat.com>
(cherry picked from commit db031a4993480ca987011630fa716f7e2ccda976)

3 years agoUse ansible_facts
Alex Schultz [Wed, 3 Mar 2021 14:43:50 +0000 (07:43 -0700)]
Use ansible_facts

It has come to our attention that using ansible_* vars that are
populated with INJECT_FACTS_AS_VARS=True is not very performant.  In
order to be able to support setting that to off, we need to update the
references to use ansible_facts[<thing>] instead of ansible_<thing>.

Related: ansible#73654
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1935406
Signed-off-by: Alex Schultz <aschultz@redhat.com>
(cherry picked from commit a7f2fa73e63e69dba2e41aaac9732397eec437c9)

3 years agolibrary: drop ceph_facts
Guillaume Abrioux [Thu, 25 Mar 2021 02:49:11 +0000 (03:49 +0100)]
library: drop ceph_facts

This is never called in the playbook and seems unmaintained.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit b01f16e8353ac261d8de34ef8b04e618c455492d)

3 years agoREADME-MULTISITE: fix typos
Ken Dreyer [Wed, 24 Mar 2021 20:44:52 +0000 (14:44 -0600)]
README-MULTISITE: fix typos

This commit fixes some typos in MULTISITE documentation.

Signed-off-by: Ken Dreyer <ktdreyer@redhat.com>
(cherry picked from commit 63a246db41195f842537473e1784f642ca6292cf)

3 years agotests: switch to quay.ceph.io for dashboard images
Guillaume Abrioux [Tue, 23 Mar 2021 13:49:30 +0000 (14:49 +0100)]
tests: switch to quay.ceph.io for dashboard images

for some reason, `quay.io/app-sre/grafana` no longer exist.
as a workaround, all dashboard related images have been mirrored on
quay.ceph.io.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit c90b0985e50cd9a4a0160d0707af544f319a7fe8)

3 years agoiscsi: fetch right repo from shaman
Guillaume Abrioux [Tue, 23 Mar 2021 10:25:51 +0000 (11:25 +0100)]
iscsi: fetch right repo from shaman

due to recent changes in shaman, we must fetch the right repo by
filtering on the desired architecture.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit 5801171b37a3acd6d7d26b874c7d541d135b9f15)