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>
The ceph osd pool ls detail command is a subset of the ceph osd dump
command.
$ ceph osd dump --format json|wc -c
10117
$ ceph osd pool ls detail --format json|wc -c
4740
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
eec3878 introduced a regression for upgrade scenarios where there's no
monitor nodes at all (like ganesha standalone, external clients, etc..)
TASK [get the ceph release being deployed] ************************************
task path: infrastructure-playbooks/rolling_update.yml:121
Thursday 29 July 2021 15:55:29 +0000 (0:00:00.484) 0:00:15.802 *********
fatal: [client0]: FAILED! =>
msg: '''dict object'' has no attribute ''mons'''
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
In the `set osd flags` block, run the Ceph commands that gather information
from the cluster (and don't make any changes to it) even when running in check
mode.
This allows the tasks that depend on the variables set by those tasks to
succeed in check mode.
Signed-off-by: Benoît Knecht <bknecht@protonmail.ch>
Check early which Ceph release is going to be deployed and fail if it
doesn't correspond to the ceph-ansible version being used.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1978643
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
This adds a task that zaps by osd id so we can support the scenario
where osds were deployed with `osd_auto_discovery` is true.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1876860
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
This refactor merges the two playbooks so we only have to maintain 1
playbook.
(Symlink the old purge-container-cluster.yml playbook for backward
compatibility).
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
Those variables are useless given this is not possible to override them.
Let's replace them with the hardcoded name instead.
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
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>
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>
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>
This commit reindents the playbook.
Also improve readability by adding an extra line between plays.
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
If one a the monitor is out of the quorum then nothing prevents the upgrade
playbook to run.
We only check if we have at least three monitor nodes but we should also
check if those monitor nodes are correctly present in the quorum.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1952571
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
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>
The dashboard/monitoring stack can be deployed via the dashboard_enabled
variable. But there's nothing similar if we can to remove that part only
and keep the ceph cluster up and running.
The current purge playbooks remove everything.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1786691
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
Add any_errors_fatal: true in cephadm-adopt playbook.
We should stop the playbook execution when a task throws an error.
Otherwise it can lead to unexpected behavior.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1976179
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
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>
Starting RHCS 5, there's no ISO available anymore.
This removes all ISO variables and the ceph_repository_type variable.
Closes: #6626
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
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>
If multi-realms were deployed with several instances belonging to the same
realm and zone using the same port on different nodes, the service id
expected by cephadm will be the same and therefore only one service will
be deployed. We need to create a service called
`<node>.<realm>.<zone>.<port>` to be sure the service name will be unique
and well deployed on the expected node in order to preserve backward
compatibility with the rgws instances that were deployed with
ceph-ansible.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1967455
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
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>
We need to support rgw multisite deployments.
This commit makes the adoption playbook support this kind of deployment.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1967455
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
When no `[mgrs]` group is defined in the inventory, mgr daemon are
implicitly collocated with monitors.
This task currently relies on the length of the mgr group in order to
tell cephadm to deploy mgr daemons.
If there's no `[mgrs]` group defined in the inventory, it will ask
cephadm to deploy 0 mgr daemon which doesn't make sense and will throw
an error.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1970313
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
0990ae4109 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>
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>
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>
ceph-ansible leaves a ceph-crash container in containerized deployment.
It means we end up with 2 ceph-crash containers running after the
migration playbook is complete.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1954614
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
Due to a recent breaking change in ceph, this command must be modified
to add the <svc_id> parameter.
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
When migrating from a cluster with no MDS nodes deployed,
`{{ cephfs_data_pool.name }}` doesn't exist so we need to create a pool
for storing nfs export objects.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1950403
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
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>
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>
the adoption playbook should use `radosgw_num_instances` in order to
determine how much rgw instance it should set recreate.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1943170
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
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>
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>
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>
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>
This commit makes the playbook fetch the minimal current ceph
configuration and write it later on monitoring nodes so `cephadm` can
proceed with the adoption.
When a monitoring stack was deployed on a dedicated node, it means no
`ceph.conf` file was written, `cephadm` requires a `ceph.conf` in order
to adopt the daemon present on the node.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1939887
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
If the legacy name `grafana-server` is still being used when upgrading
from Nautilus to Pacific, the task that sets the fact `rolling_update`
to `true` doesn't run on the node(s) included in that group. Indeed the
play where we set this fact (`rolling_update`) only runs on the group
`monitoring_group_name | default('monitoring')`.
As a workaround, we can run earlier the task which converts the
`grafana-server` group name to `monitoring`.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1935554
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
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>
This commit makes sure purge playbooks remove those file if for any reason they
have been left.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1920900
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
There's no need to slow down the playbook execution time by migrating
all the `ceph-crash` instances in a serial way. Let's remove the
`serial: 1` so the migration is achieved in a parallel way.
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
When running the rolling_update.yml playbook and adding the dashboard
component in the same time then the requirement (like container packages)
aren't installed.
This could lead to a failure in case of using authentication on the
container registry because the playbook will try to login on the registry
but podman/docker aren't yet installed.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1903504
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1918650
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
Since b105549 we don't install node-exporter on client nodes so we should
also exclude the client node from the node-exporter upgrade.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
since master is now deploying quincy, we must update this.
Otherwise, it will fail like following:
```
Error EPERM: require_osd_release cannot be lowered once it has been set
```
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
There's no reason to not use the ceph_osd_flag module to set/unset osd
flags.
Also if there's no OSD nodes in the inventory then we don't need to
execute the set/unset play.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
Instead of doing some scripting via the shell module, we can use the
parted ansible module to check the boot flag on partitions.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
Those devices (db/wal) are never zapped in lvm batch deployment.
Iterating over `dedicated_devices` and `bluestore_wal_devices` fixes
this issue.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1922926
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
When rerunning the cephadm-adopt.yml playbook the radosgw realm,
zonegroup and zone tasks will fail because the task isn't
idempotent.
Using the radosgw ansible modules solves that problem.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
If the cephadm-adopt.yml fails during the first execution and some
daemons have already been adopted by cephadm then we can't rerun
the playbook because the old container won't exist anymore.
Error: no container with name or ID ceph-mon-xxx found: no such container
If the daemons are adopted then the old systemd unit doesn't exist anymore
so any call to that unit with systemd will fail.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1918424
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
Since the default of `osd_objectstore` has changed as of 3.2, some
deployments might have a mix of filestore and bluestore OSDs on a same
node. In some specific cases, there's a possibility that a filestore OSD
shares a journal/db device with a bluestore OSD. We shouldn't try to
redeploy in this context because ceph-volume will complain. (either
because in lvm batch you can't pass partition or about gpt header).
The safest option is to skip the migration on the node when such a mix
is detected or force all osds including those already using bluestore
(option `force_filestore_to_bluestore=True` has to be passed as an extra var).
If all OSDs are using filestore, then they will be migrated to
bluestore.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1875777
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
The current check makes no sense because it checks any of other monitor
than the one being played (either a previous one already converted or a
next that isn't yet converted) is present on the quorum.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1909011
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
Instead of iterate over the host list for adding the node/label to the
host orchestrator configuration then we can do it parallelly.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
This adds cephadm_adopt ansible module for replacing the command module
usage with the cephadm adopt command.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
Let's discard the ansible lint error 306 and add a "# noqa 306" on tasks
where we don't need `set -o pipefail`
Fixes: #6090
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
We should always use the ceph_volume ansible module when possible.
This patch replace the ceph-volume inventory and lvm {list,zap} commands
called via the command/shell modules by the corresponding call with the
ceph_volume module.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
This adds cephadm_bootstrap ansible module for replacing the command module
usage with the cephadm bootstrap command.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
This adds ceph_osd_flag ansible module for replacing the command module
usage with the ceph osd set/unset commands.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
This adds ceph_osd ansible module for replacing the command module
usage with the ceph osd destroy/down/in/out/purge/rm commands.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
This adds ceph_mgr_module ansible module for replacing the command module
usage with the ceph mgr module enable/disable commands.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
`ceph.target` should be disabled only. Otherwise, in collocation
scenario you stop other collocated services in the OSD play which isn't
what we want to do. Each daemon has its corresponding play for managing
the transition to container.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1901865
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
This adds ceph_volume_simple_{activate,scan} ansible modules for replacing
the command module usage with the ceph-volume simple activate/scan commands.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
adding monitor is no longer possible because we generate a new mon
keyring each time the playbook is run.
Fixes: #5864
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
ignore 302,303 and 505 errors
[302] Using command rather than an argument to e.g. file
[303] Using command rather than module
[505] referenced files must exist
they aren't relevant on these tasks.
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
fa2bb3a only fix the symlink owner/group issue in the OSD play. If the
OSDs are collocated with other services like MONs and MGRs then the
chown command will fail.
$ find /var/lib/ceph/osd/ceph-0 -not -user 167 -execdir chown 167:167 {} +
chown: cannot dereference './block': Permission denied
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1896448
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
When deploying the ceph OSD via the packages then the ceph-osd@.service
unit is configured as enabled-runtime.
This means that each ceph-osd service will inherit from that state.
The enabled-runtime systemd state doesn't survive after a reboot.
For non containerized deployment the OSD are still starting after a
reboot because there's the ceph-volume@.service and/or ceph-osd.target
units that are doing the job.
$ systemctl list-unit-files|egrep '^ceph-(volume|osd)'|column -t
ceph-osd@.service enabled-runtime
ceph-volume@.service enabled
ceph-osd.target enabled
When switching to containerized deployment we are stopping/disabling
ceph-osd@XX.servive, ceph-volume and ceph.target and then removing the
systemd unit files.
But the new systemd units for containerized ceph-osd service will still
inherit from ceph-osd@.service unit file.
As a consequence, if an OSD host is rebooting after the playbook execution
then the ceph-osd service won't come back because they aren't enabled at
boot.
This patch also adds a reboot and testinfra run after running the switch
to container playbook.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1881288
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
cec994b introduced a regression when a mgr is collocated with a mon.
During the mon upgrade, the mgr service is masked to avoid to be
restarted on packages update.
Then the start mgr task is failing because the service is still masked.
Instead we should unmask it.
Fixes: #5983
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
bd611a7 introduced the new ceph_fs module but missed some tasks in
rolling_update and shrink-mds playbooks.
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
The ceph status command returns a lot of information stored in variables
and/or facts which could consume resources for nothing.
When checking the cluster health, we're using the health structure in the
ceph status output.
To optimize this, we could use the ceph health command which contains
the same needed information.
$ ceph status -f json | wc -c
2001
$ ceph health -f json | wc -c
46
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
The ceph status command returns a lot of information stored in variables
and/or facts which could consume resources for nothing.
When checking the rgw/rbdmirror services status, we're only using the
servicmap structure in the ceph status output.
To optimize this, we could use the ceph service dump command which contains
the same needed information.
This command returns less information and is slightly faster than the ceph
status command.
$ ceph status -f json | wc -c
2001
$ ceph service dump -f json | wc -c
1105
$ time ceph status -f json > /dev/null
real 0m0.557s
user 0m0.516s
sys 0m0.040s
$ time ceph service dump -f json > /dev/null
real 0m0.454s
user 0m0.434s
sys 0m0.020s
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
The ceph status command returns a lot of information stored in variables
and/or facts which could consume resources for nothing.
When checking the quorum status, we're only using the quorum_names
structure in the ceph status output.
To optimize this, we could use the ceph quorum_status command which contains
the same needed information.
This command returns less information.
$ ceph status -f json | wc -c
2001
$ ceph quorum_status -f json | wc -c
957
$ time ceph status -f json > /dev/null
real 0m0.577s
user 0m0.538s
sys 0m0.029s
$ time ceph quorum_status -f json > /dev/null
real 0m0.544s
user 0m0.527s
sys 0m0.016s
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
The ceph status command returns a lot of information stored in variables
and/or facts which could consume resources for nothing.
When checking the pgs state, we're using the pgmap structure in the ceph
status output.
To optimize this, we could use the ceph pg stat command which contains
the same needed information.
This command returns less information (only about pgs) and is slightly
faster than the ceph status command.
$ ceph status -f json | wc -c
2000
$ ceph pg stat -f json | wc -c
240
$ time ceph status -f json > /dev/null
real 0m0.529s
user 0m0.503s
sys 0m0.024s
$ time ceph pg stat -f json > /dev/null
real 0m0.426s
user 0m0.409s
sys 0m0.016s
The data returned by the ceph status is even bigger when using the
nautilus release.
$ ceph status -f json | wc -c
35005
$ ceph pg stat -f json | wc -c
240
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>
Seena Fallah
Guillaume Abrioux
VasishtaShastry
Dimitri Savineau
Benoît Knecht
Alex Schultz