This reverts commit e84f11e99e.
This commit was giving a new failure later during the rolling_update
process. Basically, this was modifying the list of devices and started
impacting the ceph-osd itself. The modification to accomodate the
osd_auto_discovery parameter should happen outside of the ceph-osd.
Also we are trying to not play ceph-osd role during the rolling_update
process so we can speed up the upgrade.
Signed-off-by: Sébastien Han <seb@redhat.com>
rolling_update relies on the list of devices when performing the restart
of the OSDs. The task that is builind the devices list out of the
ansible_devices dict only runs when there are no partitions on the
drives. However during an upgrade the OSD are already configured, they
have been prepared and have partitions so this task won't run and thus
the devices list will be empty, skipping the restart during
rolling_update. We now run the same task under different requirements
when rolling_update is true and build a list when:
* osd_auto_discovery is true
* rolling_update is true
* ansible_devices exists
* no dm/lv are part of the discovery
* the device is not removable
* the device has more than 1 sector
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1613626
Signed-off-by: Sébastien Han <seb@redhat.com>
This is used with the lvm osd scenario. When using devices you need the
option to set the crush device class for all of the OSDs that are
created from those devices.
Signed-off-by: Andrew Schoen <aschoen@redhat.com>
This adds the action 'batch' to the ceph-volume module so that we can
run the new 'ceph-volume lvm batch' subcommand. A functional test is
also included.
If devices is defind and osd_scenario is lvm then the 'ceph-volume lvm
batch' command will be used to create the OSDs.
Signed-off-by: Andrew Schoen <aschoen@redhat.com>
The container runs with --rm which means it will be deleted by Docker
when exiting. Also 'docker rm -f' is not idempotent and returns 1 if the
container does not exist.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1609007
Signed-off-by: Sébastien Han <seb@redhat.com>
If we want to be backward compatible with release prior to luminous, we
have to set the rule name accordingly to default values used in jewel.
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
The script ceph-osd-run.sh holds the config options to start the
container, if one of these options are modified we must restart the
container. This was not the case before becauase the 'notify' flag
wasn't present.
Closing: https://bugzilla.redhat.com/show_bug.cgi?id=1596061
Signed-off-by: Sébastien Han <seb@redhat.com>
keyring files in /etc/ceph. Default value is the same as it was (0600),
but this variable allows user to override it (f.e. set it to 0640).
Signed-off-by: George Shuklin <george.shuklin@gmail.com>
As discussed with the cores, the current limits are too low and should
be bumped to higher value.
So now by default monitors get 3GB and OSDs get 5GB.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1591876
Signed-off-by: Sébastien Han <seb@redhat.com>
When configuring openstack, the created keyrings aren't copied over to
all monitors nodes.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1588093
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
Since the openstack_config.yml has been moved to `ceph-osd` we must move
this `set_fact` in ceph-osd otherwise the tasks in
`openstack_config.yml` using `openstack_keys` will actually use the
defaults value from `ceph-defaults`.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1585139
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
This is a follow up on #2628.
Even with the openstack pools creation moved later in the playbook,
there is still an issue because OSDs are not all UP when trying to
create pools.
Adding a task which checks for all OSDs to be UP with a `retries/until`
condition should definitively fix this issue.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1578086
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
in `ceph-osd` there is no need to set `docker_exec_cmd` since the only
place where this fact is used is in `openstack_config.yml` which
delegate all docker command to a monitor node. It means we need the
`docker_exec_cmd` fact that has been set referring to `ceph-mon-*`
containers, this fact is already set earlier in `ceph-defaults`.
By the way, when collocating an OSD with a MON it fails because the container
`ceph-osd-{{ ansible_hostname }}` doesn't exist.
Removing this task will allow to collocate an OSD with a MON.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1584179
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
When deploying a large number of OSD nodes it can be an issue because the
protection check [1] won't pass since it tries to create pools before all
OSDs are active.
The idea here is to move cephfs pools creation in `ceph-mds` role.
[1] e59258943b/src/mon/OSDMonitor.cc (L5673)
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1578086
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
When deploying a large number of OSD nodes it can be an issue because the
protection check [1] won't pass since it tries to create pools before all
OSDs are active.
The idea here is to move openstack pools creation at the end of `ceph-osd` role.
[1] e59258943b/src/mon/OSDMonitor.cc (L5673)
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1578086
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
The LVM lvcreate fails if the disk already has a GPT header.
We create GPT header regardless of OSD scenario. The fix is to
skip header creation for lvm scenario.
fixes: https://github.com/ceph/ceph-ansible/issues/2592
Signed-off-by: Vishal Kanaujia <vishal.kanaujia@flipkart.com>
The order of fs.aio-max-nr (which is hard-coded to 1048576) means that
if you set fs.aio-max-nr in os_tuning_params it will effectively be
ignored for bluestore scenarios.
To resolve this we should move the setting of fs.aio-max-nr above the
setting of os_tuning_params, in this way the operator can define the
value of fs.aio-max-nr to be something other than 1048576 if they want
to.
Additionally, we can make the sysctl settings happen in 1 task rather
than multiple.
Useful for softwares that do data collection/monitoring like collectd.
They can connect to the socket and then retrieve information.
Even though the sockets are exposed now, I'm keeping the docker exec to
check the socket, this will allow newer version of ceph-ansible to work
with older versions.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1563280
Signed-off-by: Sébastien Han <seb@redhat.com>
We know bindmount with the :z option at the end of the -v command so
this will basically run the exact same command as we used to run. So to
speak:
chcon -Rt svirt_sandbox_file_t /var/lib/ceph
Signed-off-by: Sébastien Han <seb@redhat.com>
This commit does a couple of things:
* use a common.yml file that contains things that can be played on both
container and non-container
* refactor the ability to copy the admin key to the nodes
Signed-off-by: Sébastien Han <seb@redhat.com>
Regardless if the partition is 'ceph' or something else, we don't want
to be as strick as checking for a particular partition.
If the drive has a partition, we just don't do anything.
This solves the case where the server reboots, disks get a different
/dev/sda (node) allocation. In this case, prior to restarting the server
/dev/sda was an OSD, but now it's /dev/sdb and the other way around.
In such scenario, we will try to prepare the OSD and create a new
partition, so let's not mess around with devices that have partitions.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1498303
Signed-off-by: Sébastien Han <seb@redhat.com>
This was causing a lot of pain with the handlers. Also the
implementation was not ideal since we were assembling files. Everything
can now be done with the ceph_crush module so let's remove that.
Signed-off-by: Sébastien Han <seb@redhat.com>
Multipath disks have partitions with a different format than what
ceph-ansible currently supports, this update makes ceph-ansible
aware of that format so multipath disks can be used as OSDs
Signed-off-by: Caleb Boylan <caleb.boylan@ormuco.com>
This commit fixes a bug that occurs especially for dmcrypt scenarios.
There is an issue where the 'disk_list' container can't reach the ceph
cluster because it's not launched with `--net=host`.
If this container can't reach the cluster, it will hang on this step
(when trying to retrieve the dm-crypt key) :
```
+common_functions.sh:448: open_encrypted_part(): ceph --cluster abc12 --name \
client.osd-lockbox.9138767f-7445-49e0-baad-35e19adca8bb --keyring \
/var/lib/ceph/osd-lockbox/9138767f-7445-49e0-baad-35e19adca8bb/keyring \
config-key get dm-crypt/osd/9138767f-7445-49e0-baad-35e19adca8bb/luks
+common_functions.sh:452: open_encrypted_part(): base64 -d
+common_functions.sh:452: open_encrypted_part(): cryptsetup --key-file \
-luksOpen /dev/sdb1 9138767f-7445-49e0-baad-35e19adca8bb
```
It means the `ceph-run-osd.sh` script won't be able to start the
`osd_disk_activate` process in ceph-container because he won't have
filled the `$DOCKER_ENV` environment variable properly.
Adding `--net=host` to the 'disk_list' container fixes this issue.
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1543284
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
Use a nicer syntax for `local_action` tasks.
We used to have oneliner like this:
```
local_action: wait_for port=22 host={{ hostvars[inventory_hostname]['ansible_default_ipv4']['address'] }} state=started delay=10 timeout=500 }}
```
The usual syntax:
```
local_action:
module: wait_for
port: 22
host: "{{ hostvars[inventory_hostname]['ansible_default_ipv4']['address'] }}"
state: started
delay: 10
timeout: 500
```
is nicer and kind of way to keep consistency regarding the whole
playbook.
This also fix a potential issue about missing quotation :
```
Traceback (most recent call last):
File "/tmp/ansible_wQtWsi/ansible_module_command.py", line 213, in <module>
main()
File "/tmp/ansible_wQtWsi/ansible_module_command.py", line 185, in main
rc, out, err = module.run_command(args, executable=executable, use_unsafe_shell=shell, encoding=None, data=stdin)
File "/tmp/ansible_wQtWsi/ansible_modlib.zip/ansible/module_utils/basic.py", line 2710, in run_command
File "/usr/lib64/python2.7/shlex.py", line 279, in split
return list(lex) File "/usr/lib64/python2.7/shlex.py", line 269, in next
token = self.get_token()
File "/usr/lib64/python2.7/shlex.py", line 96, in get_token
raw = self.read_token()
File "/usr/lib64/python2.7/shlex.py", line 172, in read_token
raise ValueError, "No closing quotation"
ValueError: No closing quotation
```
writing `local_action: shell echo {{ fsid }} | tee {{ fetch_directory }}/ceph_cluster_uuid.conf`
can cause trouble because it's complaining with missing quotes, this fix solves this issue.
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1510555
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
Description of problem: The 'get osd id' task goes through all the 10 times (and its respective timeouts) to make sure that the number of OSDs in the osd directory match the number of devices.
This happens always, regardless if the setup and deployment is correct.
Version-Release number of selected component (if applicable): Surely the latest. But any ceph-ansible version that contains ceph-volume support is affected.
How reproducible: 100%
Steps to Reproduce:
1. Use ceph-volume (LVM) to deploy OSDs
2. Avoid using anything in the 'devices' section
3. Deploy the cluster
Actual results:
TASK [ceph-osd : get osd id _uses_shell=True, _raw_params=ls /var/lib/ceph/osd/ | sed 's/.*-//'] **********************************************************************************************************************************************
task path: /Users/alfredo/python/upstream/ceph/src/ceph-volume/ceph_volume/tests/functional/lvm/.tox/xenial-filestore-dmcrypt/tmp/ceph-ansible/roles/ceph-osd/tasks/start_osds.yml:6
FAILED - RETRYING: get osd id (10 retries left).
FAILED - RETRYING: get osd id (9 retries left).
FAILED - RETRYING: get osd id (8 retries left).
FAILED - RETRYING: get osd id (7 retries left).
FAILED - RETRYING: get osd id (6 retries left).
FAILED - RETRYING: get osd id (5 retries left).
FAILED - RETRYING: get osd id (4 retries left).
FAILED - RETRYING: get osd id (3 retries left).
FAILED - RETRYING: get osd id (2 retries left).
FAILED - RETRYING: get osd id (1 retries left).
ok: [osd0] => {
"attempts": 10,
"changed": false,
"cmd": "ls /var/lib/ceph/osd/ | sed 's/.*-//'",
"delta": "0:00:00.002717",
"end": "2018-01-21 18:10:31.237933",
"failed": true,
"failed_when_result": false,
"rc": 0,
"start": "2018-01-21 18:10:31.235216"
}
STDOUT:
0
1
2
Expected results:
There aren't any (or just a few) timeouts while the OSDs are found
Additional info:
This is happening because the check is mapping the number of "devices" defined for ceph-disk (in this case it would be 0) to match the number of OSDs found.
Basically this line:
until: osd_id.stdout_lines|length == devices|unique|length
Means in this 2 OSD case it is trying to ensure the following incorrect condition:
until: 2 == 0
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1537103
This is to keep backward compatibility with stable-2.2 and satisfy the
check "verify dedicated devices have been provided" in
`check_mandatory_vars.yml`. This check is looking for
`dedicated_devices` so we need to default it's value to
`raw_journal_devices` when `raw_multi_journal` is set to `True`.
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1536098
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
On a non-collocated scenario, if a drive is faulty we can't really
remove it from the list of 'devices' without messing up or having to
re-arrange the order of the 'dedicated_devices'. We want to keep this
device list ordered. This will prevent the activation failing on a
device that we know is failing but we can't remove it yet to not mess up
the dedicated_devices mapping with devices.
Signed-off-by: Sébastien Han <seb@redhat.com>
the gpt label creation doesn't work even with parted module.
This commit fixes the gpt label creation by using parted command
instead.
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
We have a scenario when we switch from non-container to containers. This
means we don't know anything about the ceph partitions associated to an
OSD. Normally in a containerized context we have files containing the
preparation sequence. From these files we can get the capabilities of
each OSD. As a last resort we use a ceph-disk call inside a dummy bash
container to discover the ceph journal on the current osd.
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1525612
Signed-off-by: Sébastien Han <seb@redhat.com>
The name docker_version is very generic and is also used by other
roles. As a result, there may be name conflicts. To avoid this a
ceph_ prefix should be used for this fact. Since it is an internal
fact renaming is not a problem.
Add the variables ceph_osd_docker_cpuset_cpus and
ceph_osd_docker_cpuset_mems, so that a user may specify
the CPUs and memory nodes of NUMA systems on which OSD
containers are run.
Provides a example in osds.yaml.sample to guide user
based on sample `lscpu` output since cpuset-mems refers
to the memory by NUMA node only while cpuset-cpus can
refer to individual vCPUs within a NUMA node.
This is particularly useful in CI environments where you dont have
the option of adding extra devices or volumes to the host. It is also
a simple change to support loopback devices
This task was originally needed to fix a docker installation issue
(see: #1030). This has been fixed, therefore it can be removed.
Fixes: #2199
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
We were activating dmcrypt devices with the wrong command. Basically the
first task execute the wrong activate command. The task fails but
continues because of the 'failed_when: false'. Then the right activation
sequence is being done by the next task.
Signed-off-by: Sébastien Han <seb@redhat.com>
there is no need to have a condition on this task, this test should be
always run since the result will be interpreted later.
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
This will prevent ceph-ansible from using a loop device while it
shouldn't in auto_discovery mode.
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
Use `devices` variable instead of `ansible_devices`, otherwise it means
we are not using the devices which have been 'auto discovered'
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
The current code will also return lvm devices such as /dev/dm-2, this
kind of device type is not supported by ceph-disk at the moment. Now we
just ignore them.
Signed-off-by: Sébastien Han <seb@redhat.com>
During the initial implementation of this 'old' thing we were falling
into this issue without noticing
https://github.com/moby/moby/issues/30341 and where blindly using --rm,
now this is fixed the prepare container disappears and thus activation
fail.
I'm fixing this for old jewel images.
Also this fixes the machine reboot case where the docker logs are
purgend. In the old scenario, we now store the log locally in the same
directory as the ceph-osd-run.sh script.
Signed-off-by: Sébastien Han <seb@redhat.com>
Use "ceph_tcmalloc_max_total_thread_cache" to set the
TCMALLOC_MAX_TOTAL_THREAD_CACHE_BYTES value inside /etc/default/ceph for
Debian installs, or /etc/sysconfig/ceph for Red Hat/CentOS installs.
By default this is set to 0, so the default package value will be used,
if specified this value will be changed to match the variable, and ceph
osd services will be restarted.
There was a huge resync from luminous to jewel in ceph-docker:
https://github.com/ceph/ceph-docker/pull/797
This change brought a new handy function to discover partitions tight to
an OSD. This function doesn't exist in the old image so the
ceph-osd-run.sh script breaks when trying to deploy Jewel OSD with that
old Jewel image version.
Signed-off-by: Sébastien Han <seb@redhat.com>
stable-3.0 brought numerous changes in ceph-ansible variables, this PR
aims to maintain backward compatibility for someone running stable-2.2
upgrading to stable-3.0 but keeps its groups_vars untouched.
We will then determine the right options to make sure the upgrade works
but we are expecting that new variables should be used.
We will drop this in a near future, maybe 3.1 or 3.2.
Signed-off-by: Sébastien Han <seb@redhat.com>
This commit add new osd scenarios, it aims to simplify the CI setup and
brings a better coverage on the OSD scenarios.
We decided to differentiate between filestore and bluestore, thinking
ahead when filestore won't be supported anymore.
So we now have two classes of tests:
* Filestore
* Bluestore
In each of those classes we have container and non-container.
Then for each we test the following:
* collocated
* collocated dmcrypt
* non-collocated
* non-collocated dmcrypt
* auto discovery collocated
* auto discovery collocated dmcrypt
This gives us a nice coverage and also reduces the footprint on the CI.
We are now up to 4 scenarios, each containing 6 OSD VMs.
Signed-off-by: Sébastien Han <seb@redhat.com>
This will solve the following issue when starting docker containers on ubuntu:
invalid argument "1\u00a0" for --cpus=1 : failed to parse 1 as a rational number
Closes-bug: #2056
This patch changes the `when:` keys so that they have no jinja2
delimiters. This avoids Ansible warnings which could turn into
errors in a future Ansible release.
The `always_run` key is deprecated and being removed in Ansible 2.4.
Using it causes a warning to be displayed:
[DEPRECATION WARNING]: always_run is deprecated.
This patch changes all instances of `always_run` to use the `always`
tag, which causes the task to run each time the playbook runs.
This is causing unknown issues when trying to start a dmcrypt container.
Basically the container is stuck at mount opening the LUKS device. This
is still unknown why this is causing trouble but we need to move
forward. Also, this doesn't seem to help in any ways to fix the race
condition we've seen.
Here is the log for dmcrypt:
cryptsetup 1.7.4 processing "cryptsetup --debug --verbose --key-file
key luksClose fbf8887d-8694-46ca-b9ff-be79a668e2a9"
Running command close.
Locking memory.
Installing SIGINT/SIGTERM handler.
Unblocking interruption on signal.
Allocating crypt device context by device
fbf8887d-8694-46ca-b9ff-be79a668e2a9.
Initialising device-mapper backend library.
dm version [ opencount flush ] [16384] (*1)
dm versions [ opencount flush ] [16384] (*1)
Detected dm-crypt version 1.14.1, dm-ioctl version 4.35.0.
Device-mapper backend running with UDEV support enabled.
dm status fbf8887d-8694-46ca-b9ff-be79a668e2a9 [ opencount flush ]
[16384] (*1)
Releasing device-mapper backend.
Trying to open and read device /dev/sdc1 with direct-io.
Allocating crypt device /dev/sdc1 context.
Trying to open and read device /dev/sdc1 with direct-io.
Initialising device-mapper backend library.
dm table fbf8887d-8694-46ca-b9ff-be79a668e2a9 [ opencount flush
securedata ] [16384] (*1)
Trying to open and read device /dev/sdc1 with direct-io.
Crypto backend (gcrypt 1.5.3) initialized in cryptsetup library
version 1.7.4.
Detected kernel Linux 3.10.0-693.el7.x86_64 x86_64.
Reading LUKS header of size 1024 from device /dev/sdc1
Key length 32, device size 1943016847 sectors, header size 2050
sectors.
Deactivating volume fbf8887d-8694-46ca-b9ff-be79a668e2a9.
dm status fbf8887d-8694-46ca-b9ff-be79a668e2a9 [ opencount flush ]
[16384] (*1)
Udev cookie 0xd4d14e4 (semid 32769) created
Udev cookie 0xd4d14e4 (semid 32769) incremented to 1
Udev cookie 0xd4d14e4 (semid 32769) incremented to 2
Udev cookie 0xd4d14e4 (semid 32769) assigned to REMOVE task(2) with
flags (0x0)
dm remove fbf8887d-8694-46ca-b9ff-be79a668e2a9 [ opencount flush
retryremove ] [16384] (*1)
fbf8887d-8694-46ca-b9ff-be79a668e2a9: Stacking NODE_DEL [verify_udev]
Udev cookie 0xd4d14e4 (semid 32769) decremented to 1
Udev cookie 0xd4d14e4 (semid 32769) waiting for zero
Signed-off-by: Sébastien Han <seb@redhat.com>
Sébastien Han
Andrew Schoen
Guillaume Abrioux
George Shuklin
Konstantin Shalygin
Vishal Kanaujia
Andy McCrae
vasishta p shastry
Alfredo Deza
Ning Yao
Caleb Boylan
Christian Berendt
John Fulton
Markos Chandras
Major Hayden
Yixing Yan