]> git.apps.os.sepia.ceph.com Git - ceph-ansible.git/log
ceph-ansible.git
7 years agoiscsi-gw: move repository configuration to ceph-iscsi-gw
Guillaume Abrioux [Tue, 10 Oct 2017 08:39:02 +0000 (10:39 +0200)]
iscsi-gw: move repository configuration to ceph-iscsi-gw

This is something that has nothing to do in `ceph-common`, this
is too specific to `ceph-iscsi-gw` role.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoDocs: fix error `Unknown target name`
Guillaume Abrioux [Mon, 9 Oct 2017 19:38:31 +0000 (21:38 +0200)]
Docs: fix error `Unknown target name`

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agonfs: move repository configuration in ceph-nfs role
Guillaume Abrioux [Mon, 9 Oct 2017 19:34:17 +0000 (21:34 +0200)]
nfs: move repository configuration in ceph-nfs role

This is something that has nothing to do in `ceph-common`, this
is too specific to `ceph-nfs` role.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoMerge pull request #2014 from ceph/fixes-2 v3.0.0rc19
Guillaume Abrioux [Mon, 9 Oct 2017 18:14:06 +0000 (20:14 +0200)]
Merge pull request #2014 from ceph/fixes-2

infra: use the pg check in the right place

7 years agonfs: move packages installation to own role 2014/head
Guillaume Abrioux [Mon, 9 Oct 2017 15:10:39 +0000 (17:10 +0200)]
nfs: move packages installation to own role

Make role `ceph-nfs` handling itself the installation of nfs
packages.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agomds: move installation packages in role itself
Guillaume Abrioux [Mon, 9 Oct 2017 13:37:51 +0000 (15:37 +0200)]
mds: move installation packages in role itself

Make role `ceph-mds` handling itself the installation of `ceph-mds`
package.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoiscsi: move package install to ceph-iscsi-role
Sébastien Han [Mon, 9 Oct 2017 13:30:54 +0000 (15:30 +0200)]
iscsi: move package install to ceph-iscsi-role

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agomgr: move installation packages in role itself
Guillaume Abrioux [Mon, 9 Oct 2017 13:26:15 +0000 (15:26 +0200)]
mgr: move installation packages in role itself

Make role `ceph-mgr` handling itself the installation of `ceph-mgr`
package because it's complicated to manage it regarding we are going to
install `jewel vs. luminous`

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agotests: adapt current testing for collocation scenario
Guillaume Abrioux [Mon, 9 Oct 2017 11:48:03 +0000 (13:48 +0200)]
tests: adapt current testing for collocation scenario

Since we introduced collocation testing scenario, we need to adapt
current tests to this new scenario.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoci: disable vnc console
Sébastien Han [Mon, 9 Oct 2017 09:56:46 +0000 (11:56 +0200)]
ci: disable vnc console

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoosd: bindmount /run/udev
Sébastien Han [Mon, 9 Oct 2017 08:41:14 +0000 (10:41 +0200)]
osd: bindmount /run/udev

Ensures that "udevadm" is able to check the status of udev's event queue.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoci: re-add osd_pool_default_size to 1 with the override
Sébastien Han [Sun, 8 Oct 2017 15:29:32 +0000 (17:29 +0200)]
ci: re-add osd_pool_default_size to 1 with the override

If we don't do this the client will create pools with a replica 3 since
osd_pool_default_size was gone in ceph-override.json. This was making
switch_to_containers failing.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agotox: decrease sleep to 2min instead of 5.
Sébastien Han [Sun, 8 Oct 2017 13:56:39 +0000 (15:56 +0200)]
tox: decrease sleep to 2min instead of 5.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoinfra: add independant purge-iscsi-gateways.yml
Sébastien Han [Sun, 8 Oct 2017 13:54:36 +0000 (15:54 +0200)]
infra: add independant purge-iscsi-gateways.yml

The current inclusion of purge-iscsi-gateways.yml in purge-cluster.yml
is not working well and blocking the CI too. So removing it from
purge-cluster.yml and re-add the original purge-iscsi-gateways.yml.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoconfig: do not duplicate sections when doing collocation
Sébastien Han [Sun, 8 Oct 2017 13:16:40 +0000 (15:16 +0200)]
config: do not duplicate sections when doing collocation

Prior to this commit, when collocating a RGW and NFS on the same box the
ceph.conf layout was the following:

[client.rgw.rgw0]
host = mds0
host = rgw0
rgw frontends = civetweb port=192.168.15.50:8080
num_threads=100[client.rgw.mds0]
rgw frontends = civetweb port=192.168.15.70:8080 num_threads=100
rgw frontends = civetweb port=192.168.15.50:8080 num_threads=100
keyring = /var/lib/ceph/radosgw/test-rgw.mds0/keyring
keyring = /var/lib/ceph/radosgw/test-rgw.rgw0/keyring
rgw data = /var/lib/ceph/radosgw/test-rgw.rgw0
log file = /var/log/ceph/test-rgw-mds0.log
log file = /var/log/ceph/test-rgw-rgw0.log

[mds.mds0]
host = mds0

[global]
rgw override bucket index max shards = 16
fsid = 70e1d368-57b3-4978-b746-cbffce6e56b5
rgw bucket default quota max objects = 1638400
osd_pool_default_size = 1
public network = 192.168.15.0/24
mon host = 192.168.15.10,192.168.15.11,192.168.15.12
osd_pool_default_pg_num = 8
cluster network = 192.168.16.0/24

[mds.rgw0]
host = rgw0

[client.rgw.mds0]
host = mds0
rgw data = /var/lib/ceph/radosgw/test-rgw.mds0
keyring = /var/lib/ceph/radosgw/test-rgw.mds0/keyring
rgw frontends = civetweb port=192.168.15.70:8080 num_threads=100
log file = /var/log/ceph/test-rgw-mds0.log

Basically appending all the sections. This commits solves that.
Now the sections appear like this:

-bash-4.2# cat /etc/ceph/test.conf
[client.rgw.rgw0]
log file = /var/log/ceph/test-rgw-rgw0.log
host = rgw0
keyring = /var/lib/ceph/radosgw/test-rgw.rgw0/keyring
rgw frontends = civetweb port=192.168.15.50:8080 num_threads=100

[client.rgw.mds0]
log file = /var/log/ceph/test-rgw-mds0.log
host = mds0
keyring = /var/lib/ceph/radosgw/test-rgw.mds0/keyring
rgw frontends = civetweb port=192.168.15.70:8080 num_threads=100

[global]
cluster network = 192.168.16.0/24
mon host = 192.168.15.10,192.168.15.11,192.168.15.12
osd_pool_default_size = 1
public network = 192.168.15.0/24
rgw bucket default quota max objects = 1638400
osd_pool_default_pg_num = 8
rgw override bucket index max shards = 16
fsid = 77a21980-3033-4174-9264-1abc7185bcb3

[mds.rgw0]
host = rgw0

[mds.mds0]
host = mds0

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoinfra: use the pg check in the right place
Sébastien Han [Sun, 8 Oct 2017 12:45:48 +0000 (14:45 +0200)]
infra: use the pg check in the right place

Use the pg check before doing the pg check, not on the quorum check.
Also never quote int when doing comparaison.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoMerge pull request #2013 from ceph/wip-purge-cluster
Sébastien Han [Mon, 9 Oct 2017 15:18:30 +0000 (17:18 +0200)]
Merge pull request #2013 from ceph/wip-purge-cluster

A couple of purge cluster fixes

7 years agoMerge pull request #2011 from ceph/switch-pg v3.0.0rc18
Sébastien Han [Sat, 7 Oct 2017 01:42:44 +0000 (03:42 +0200)]
Merge pull request #2011 from ceph/switch-pg

[skip ci] switch: check pgs only when num_pgs > 0

7 years agoswitch: check pgs only when num_pgs > 0 2011/head
Sébastien Han [Sat, 7 Oct 2017 01:42:09 +0000 (03:42 +0200)]
switch: check pgs only when num_pgs > 0

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoMerge pull request #2009 from ceph/fix-clean-pg
Sébastien Han [Sat, 7 Oct 2017 01:39:26 +0000 (03:39 +0200)]
Merge pull request #2009 from ceph/fix-clean-pg

[skip ci] handler: do not test if pgs_num = 0

7 years agohandler: add serial restart back 2009/head
Sébastien Han [Fri, 6 Oct 2017 23:23:30 +0000 (01:23 +0200)]
handler: add serial restart back

We now restart daemons on each machine in a serialized fashion.

Closes: https://github.com/ceph/ceph-ansible/issues/1989
Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoci: use by-id instead of by-path
Sébastien Han [Fri, 6 Oct 2017 22:58:54 +0000 (00:58 +0200)]
ci: use by-id instead of by-path

by-id relies on the disk WWID which is more reliable then by-path
(pointing to the PCI info)

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agorolling_update: perform pg check when pgs_num > 0
Sébastien Han [Fri, 6 Oct 2017 22:46:46 +0000 (00:46 +0200)]
rolling_update: perform pg check when pgs_num > 0

If num_pgs = 0 the check will never return 0.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agocommon: fix debian rhcs installation
Sébastien Han [Fri, 6 Oct 2017 20:17:50 +0000 (22:17 +0200)]
common: fix debian rhcs installation

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoswitch: rework and fix clean pg wait
Sébastien Han [Fri, 6 Oct 2017 17:52:05 +0000 (19:52 +0200)]
switch: rework and fix clean pg wait

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agopurge-docker: rework device detection
Sébastien Han [Fri, 6 Oct 2017 17:50:15 +0000 (19:50 +0200)]
purge-docker: rework device detection

we don't need "devices" and other device variable anymore, the playbook
detects that for us.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoshrink-osd: admin key not needed for container shrink
Sébastien Han [Fri, 6 Oct 2017 17:15:26 +0000 (19:15 +0200)]
shrink-osd: admin key not needed for container shrink

Also do some clean

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agopurge-cluster: Do not use shell for rm 2013/head
Boris Ranto [Fri, 6 Oct 2017 20:54:34 +0000 (22:54 +0200)]
purge-cluster: Do not use shell for rm

The shell wildcard expansion of non-existing paths fails on zsh making
the whole script fail. We can use file module with with_fileglob to
alleviate the problem instead.

Signed-off-by: Boris Ranto <branto@redhat.com>
7 years agopurge-cluster: Do not fail on systemd commands
Boris Ranto [Fri, 6 Oct 2017 20:52:53 +0000 (22:52 +0200)]
purge-cluster: Do not fail on systemd commands

The systemd can't stop services if the unit files were removed before
the cluster was purged. We should just ignore these.

Signed-off-by: Boris Ranto <branto@redhat.com>
7 years agoMerge pull request #1983 from jprovaznik/suffix
Sébastien Han [Fri, 6 Oct 2017 20:40:57 +0000 (22:40 +0200)]
Merge pull request #1983 from jprovaznik/suffix

Allow to override systemd service instance id

7 years agoMerge pull request #1984 from jprovaznik/exportdir
Sébastien Han [Fri, 6 Oct 2017 20:38:13 +0000 (22:38 +0200)]
Merge pull request #1984 from jprovaznik/exportdir

Include exports dir in ceph-nfs config file

7 years agoMerge pull request #2010 from ceph/wip-nfs-set-rgw-key-permissions
Sébastien Han [Fri, 6 Oct 2017 19:47:49 +0000 (21:47 +0200)]
Merge pull request #2010 from ceph/wip-nfs-set-rgw-key-permissions

nfs: missing conditional for setting rgw key permissions

7 years agonfs: missing conditional for setting rgw key permissions 2010/head
Ali Maredia [Fri, 6 Oct 2017 19:21:35 +0000 (15:21 -0400)]
nfs: missing conditional for setting rgw key permissions

Signed-off-by: Ali Maredia <amaredia@redhat.com>
7 years agoMerge pull request #2005 from ceph/wip-nfs-export-id
Sébastien Han [Fri, 6 Oct 2017 15:05:21 +0000 (17:05 +0200)]
Merge pull request #2005 from ceph/wip-nfs-export-id

nfs: config var changes

7 years agouse get to check stdout_lines
Sébastien Han [Fri, 6 Oct 2017 14:49:46 +0000 (16:49 +0200)]
use get to check stdout_lines

During the initial play, the docker command doesn't not exist and then
there is no stdout_lines to the command. So get allows us to fix this by
declaring an array if the command fails.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agohandler: do not test if pgs_num = 0
Sébastien Han [Fri, 6 Oct 2017 12:59:23 +0000 (14:59 +0200)]
handler: do not test if pgs_num = 0

We don't need to wait if they are no PGS.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoMerge pull request #2006 from ceph/fix_jewel_nightly
Guillaume Abrioux [Fri, 6 Oct 2017 14:06:23 +0000 (16:06 +0200)]
Merge pull request #2006 from ceph/fix_jewel_nightly

Fix jewel nightly

7 years agoosd: fix `set_fact build dedicated_devices` 2006/head
Guillaume Abrioux [Fri, 6 Oct 2017 12:35:42 +0000 (14:35 +0200)]
osd: fix `set_fact build dedicated_devices`

Use an intermediate variable to build the final `dedicated_devices` list
to avoid duplicate entry in that array. (We need a 1:1 relation between
`dedicated_devices` and `devices` since we are using a `with_together`
later.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agotests: skip tests for nfs nodes when release is jewel
Guillaume Abrioux [Fri, 6 Oct 2017 10:49:39 +0000 (12:49 +0200)]
tests: skip tests for nfs nodes when release is jewel

nfs nodes are not deployed on jewel so we should skip the tests on them.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agorbd: fix bug when trying to fetch key
Guillaume Abrioux [Thu, 5 Oct 2017 20:25:34 +0000 (22:25 +0200)]
rbd: fix bug when trying to fetch key

With jewel, `bootstrap_rbd_keyring` is not set because of this condition:

```
when:
  - ceph_release_num.{{ ceph_release }} >= ceph_release_num.luminous
```

Therefore, the task `try to fetch ceph config and keys` will fail.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoAllow to override systemd service instance id 1983/head
Jan Provaznik [Wed, 4 Oct 2017 11:09:43 +0000 (13:09 +0200)]
Allow to override systemd service instance id

It's useful to have constant service instance id when ceph-nfs
is managed by pacemaker.

7 years agonfs: config var changes 2005/head
Ali Maredia [Thu, 5 Oct 2017 20:51:23 +0000 (16:51 -0400)]
nfs: config var changes

- remove unused ganesha config vars,

- set different default Export_ids for each FSAL

Signed-off-by: Ali Maredia <amaredia@redhat.com>
7 years agoMerge pull request #2003 from ceph/debian-iso v3.0.0rc17
Sébastien Han [Thu, 5 Oct 2017 16:57:47 +0000 (18:57 +0200)]
Merge pull request #2003 from ceph/debian-iso

[skip ci] common: iso install on Debian is supported by rhcs

7 years agocommon: iso install on Debian is supported by rhcs 2003/head
Sébastien Han [Thu, 5 Oct 2017 16:33:17 +0000 (18:33 +0200)]
common: iso install on Debian is supported by rhcs

Also adds support for RCSH installation on Debian.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoMerge pull request #2004 from ceph/debian-url
Sébastien Han [Thu, 5 Oct 2017 16:46:13 +0000 (18:46 +0200)]
Merge pull request #2004 from ceph/debian-url

common: fix ga verison for debian rhcs

7 years agocommon: fix ga verison for debian rhcs 2004/head
Sébastien Han [Thu, 5 Oct 2017 16:45:30 +0000 (18:45 +0200)]
common: fix ga verison for debian rhcs

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoMerge pull request #1985 from ceph/debian-rhcs
Sébastien Han [Thu, 5 Oct 2017 16:42:46 +0000 (18:42 +0200)]
Merge pull request #1985 from ceph/debian-rhcs

[skip ci] common: fix rhcs installation on debian

7 years agoMerge pull request #2001 from ceph/iscsi
Sébastien Han [Thu, 5 Oct 2017 16:29:06 +0000 (18:29 +0200)]
Merge pull request #2001 from ceph/iscsi

iscsi: fix wrong group name for iscsi

7 years agoMerge pull request #2002 from ceph/non-col
Sébastien Han [Thu, 5 Oct 2017 16:25:16 +0000 (18:25 +0200)]
Merge pull request #2002 from ceph/non-col

[skip ci] osd: do not do unique on dedicated_devices

7 years agoosd: do not do unique on dedicated_devices 2002/head
Sébastien Han [Thu, 5 Oct 2017 16:20:18 +0000 (18:20 +0200)]
osd: do not do unique on dedicated_devices

This is needed later, if we do unique, only the first OSD will get a
journal.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agocommon: fix rhcs installation on debian 1985/head
Sébastien Han [Wed, 4 Oct 2017 11:36:01 +0000 (13:36 +0200)]
common: fix rhcs installation on debian

* Change version from 2 to 3.
* use ceph_rhcs_cdn_debian_repo_version to use other repositories along
* with ceph_rhcs_cdn_debian_repo

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoiscsi: fix wrong group name for iscsi 2001/head
Sébastien Han [Thu, 5 Oct 2017 14:40:07 +0000 (16:40 +0200)]
iscsi: fix wrong group name for iscsi

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1498490
Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoMerge pull request #1957 from ceph/multi_env_test
Sébastien Han [Thu, 5 Oct 2017 14:00:47 +0000 (16:00 +0200)]
Merge pull request #1957 from ceph/multi_env_test

tests: make all subnet uniq per scenario

7 years agoMerge pull request #1997 from rrmichel/osd_fragment
Sébastien Han [Thu, 5 Oct 2017 13:58:49 +0000 (15:58 +0200)]
Merge pull request #1997 from rrmichel/osd_fragment

Fixing path to osd_fragment.yml

7 years agoMerge pull request #1995 from ceph/remove-rbd-check
Sébastien Han [Thu, 5 Oct 2017 13:31:48 +0000 (15:31 +0200)]
Merge pull request #1995 from ceph/remove-rbd-check

jewel: remove rbd check

7 years agotests: set `volume_cache` 'unsafe' 1957/head
Guillaume Abrioux [Wed, 4 Oct 2017 15:09:34 +0000 (17:09 +0200)]
tests: set `volume_cache` 'unsafe'

Set volume_cache to unsafe for CI VMs.
We might be using tmpfs for volume disks soon, therefore 'unsafe' is a
prerequisite for volume_cache

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agotests: disable shared folder
Guillaume Abrioux [Wed, 4 Oct 2017 12:32:34 +0000 (14:32 +0200)]
tests: disable shared folder

Shared folder is not required for tests.
We should avoid hitting the error :
```
uninitialized constant VagrantPlugins::ProviderLibvirt::Action::ShareFolders
```
Also, disabling it might reduce the needed time in certains cases for the VMs
to be started.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agotests: make all subnet uniq per scenario
Guillaume Abrioux [Thu, 28 Sep 2017 12:14:38 +0000 (14:14 +0200)]
tests: make all subnet uniq per scenario

If two environments are using the same subnet, we will get trouble
because of ips addresses conflicts.
This commit ensures each scenario has a uniq subnet for both public and cluster
network so we can setup several test environment at a time on a same hypervisor.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoMerge pull request #1999 from ceph/bz-1486062
Sébastien Han [Thu, 5 Oct 2017 13:01:00 +0000 (15:01 +0200)]
Merge pull request #1999 from ceph/bz-1486062

common: remove unusuable conditions

7 years agocommon: remove unusuable conditions 1999/head
Guillaume Abrioux [Thu, 5 Oct 2017 12:56:31 +0000 (14:56 +0200)]
common: remove unusuable conditions

`ceph_release` isn't available at this step of the playbook because it
is set later based on the installed binaries.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1486062
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoMerge pull request #1993 from jprovaznik/log
Sébastien Han [Thu, 5 Oct 2017 12:43:26 +0000 (14:43 +0200)]
Merge pull request #1993 from jprovaznik/log

Fix bind mount for /var/lib/nfs/ganesha directory

7 years agoFixing path to osd_fragment.yml 1997/head
Michel Rode [Thu, 5 Oct 2017 12:42:10 +0000 (14:42 +0200)]
Fixing path to osd_fragment.yml

7 years agoInclude exports dir in ceph-nfs config file 1984/head
Jan Provaznik [Wed, 4 Oct 2017 11:25:53 +0000 (13:25 +0200)]
Include exports dir in ceph-nfs config file

Exports dir is used when dynamic exports creation is enabled.

7 years agoMerge pull request #1988 from ceph/fix_keyrings
Sébastien Han [Thu, 5 Oct 2017 12:30:09 +0000 (14:30 +0200)]
Merge pull request #1988 from ceph/fix_keyrings

docker: fix keyrings copied on all nodes

7 years agojewel: remove rbd check 1995/head
Sébastien Han [Thu, 5 Oct 2017 12:21:37 +0000 (14:21 +0200)]
jewel: remove rbd check

The value of doing this is fairly low compare to the added value.
So we remove these tasks, if rbd pool on Jewel doesn't have the right PG
value you can always increase it.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoFix bind mount for /var/lib/nfs/ganesha directory 1993/head
Jan Provaznik [Thu, 5 Oct 2017 11:44:43 +0000 (13:44 +0200)]
Fix bind mount for /var/lib/nfs/ganesha directory

7 years agotests: add ceph-override.json for ubuntu/cluster 1988/head
Guillaume Abrioux [Thu, 5 Oct 2017 10:59:29 +0000 (12:59 +0200)]
tests: add ceph-override.json for ubuntu/cluster

in addition to 18e2ab4d this commit adds the same file for ubuntu
testing scenario.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoMerge pull request #1992 from jprovaznik/pool
Sébastien Han [Thu, 5 Oct 2017 10:13:47 +0000 (12:13 +0200)]
Merge pull request #1992 from jprovaznik/pool

Evaluate cephfs pool variables

7 years agotests: fix broken osd test for xenial_cluster
Guillaume Abrioux [Thu, 5 Oct 2017 09:03:41 +0000 (11:03 +0200)]
tests: fix broken osd test for xenial_cluster

the path `/dev/disk/by-path/pci-0000:00:01.1-ata-1.0` doesn't exist.
it has to be changed to `/dev/disk/by-path/pci-0000:00:01.1-ata-1`

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoEvaluate cephfs pool variables 1992/head
Jan Provaznik [Thu, 5 Oct 2017 08:00:20 +0000 (10:00 +0200)]
Evaluate cephfs pool variables

Otherwise pools with names 'cephfs_data' and 'cephfs_metadata'
are created.

7 years agodocker: fix keyrings copied on all nodes
Guillaume Abrioux [Wed, 4 Oct 2017 21:10:10 +0000 (23:10 +0200)]
docker: fix keyrings copied on all nodes

All keyring are getting copied to all nodes.
This commit fixes a leftover from a previous code refactor.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1498583
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoMerge pull request #1976 from ceph/rbd_pool v3.0.0rc16
Sébastien Han [Wed, 4 Oct 2017 16:52:49 +0000 (18:52 +0200)]
Merge pull request #1976 from ceph/rbd_pool

Only perform actions on the rbd pool after it has been created

7 years agotests: increase number of threads for testinfra 1976/head
Guillaume Abrioux [Wed, 4 Oct 2017 14:51:32 +0000 (16:51 +0200)]
tests: increase number of threads for testinfra

from 4 to 8 to make testing faster.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agotests: fix brokens tests for mds
Guillaume Abrioux [Wed, 4 Oct 2017 14:47:09 +0000 (16:47 +0200)]
tests: fix brokens tests for mds

5968cf0 broke the test on mds because of leftover.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoiscsi-gw: Create a rbd pool if it doesn't exist
Guillaume Abrioux [Wed, 4 Oct 2017 06:58:03 +0000 (08:58 +0200)]
iscsi-gw: Create a rbd pool if it doesn't exist

iscsi-gw needs a 'rbd' pool to configure iscsi target.
Note: I could have used the facts already set in `ceph-mon` but I voluntarily
didn't do it to not create a dependancy between these two roles.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoRefact code for set_osd_pool_default_*
Guillaume Abrioux [Tue, 3 Oct 2017 11:55:27 +0000 (13:55 +0200)]
Refact code for set_osd_pool_default_*

This commit refacts the code regarding all `set_osd_pool_default_*`
related tasks by avoiding usage of useless `set_fact` to determine
whether a key is present in `ceph_conf_overrides`.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoOnly perform actions on the rbd pool after it has been created
Al Lau [Fri, 29 Sep 2017 17:19:05 +0000 (10:19 -0700)]
Only perform actions on the rbd pool after it has been created

The rbd pool is the default pool that gets created during ceph cluster
initializaiton.  If we act on the rbd related operations too early, the
rbd pool does not exist yet.  Move the call to perform rbd operations
to a later stage after other pools have been created.

The rbd_pool.yml playbook has all the operations related to the rbd pool.

Replace the always_run (deprecated) directive with check_mode.

Most of the ceph related tasks only need to run once.  The run_once directive
executes the task on the first host.

The ceph sub-command to delete a pool is delete (not rm).

The changes submitted here were tested with this ceph version.
ceph version 0.94.9-9.el7cp (b83334e01379f267fb2f9ce729d74a0a8fa1e92c)

This upload includes these changes:
  - Use the fail module (instead of assert).
  - From luminous release, the rbd pool is no longer created by default.
    Delete the code to create the rbd pool for luminous release
  - Conform the .yml files to use the suggested syntax.

The commands are executed on the mcp nodes and I think shell ansible module
is the right one to use.  The command module is used to execute commands on
remote nodes.  I can make the change to use command module if that is
prefrerred.

7 years agoMerge pull request #1986 from ceph/fixes
Sébastien Han [Wed, 4 Oct 2017 13:38:33 +0000 (15:38 +0200)]
Merge pull request #1986 from ceph/fixes

[skip ci] defaults: fix check socket non-container handler

7 years agodefaults: fix check socket non-container handler 1986/head
Sébastien Han [Wed, 4 Oct 2017 13:33:52 +0000 (15:33 +0200)]
defaults: fix check socket non-container handler

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoMerge pull request #1953 from ceph/colocation
Sébastien Han [Wed, 4 Oct 2017 13:02:57 +0000 (15:02 +0200)]
Merge pull request #1953 from ceph/colocation

[skip ci] Colocation

7 years agonfs: add run once to user creation 1953/head
Sébastien Han [Wed, 4 Oct 2017 11:16:04 +0000 (13:16 +0200)]
nfs: add run once to user creation

The create user call is idempotent but it's also blocking for some
reasons.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoset docker_exec_cmd fact early in each role
Guillaume Abrioux [Wed, 4 Oct 2017 08:39:17 +0000 (10:39 +0200)]
set docker_exec_cmd fact early in each role

This is to ensure `docker_exec_cmd` fact is set with the correct value
in case of daemons collocation

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoci: add collocation scenario
Sébastien Han [Wed, 27 Sep 2017 16:22:15 +0000 (18:22 +0200)]
ci: add collocation scenario

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoMerge pull request #1977 from ktdreyer/doc-master-ansible-version v3.0.0rc15
Sébastien Han [Tue, 3 Oct 2017 20:55:22 +0000 (22:55 +0200)]
Merge pull request #1977 from ktdreyer/doc-master-ansible-version

docs: master requires ansible 2.3.1

7 years agoMerge pull request #1978 from ktdreyer/doc-installing-ansible
Sébastien Han [Tue, 3 Oct 2017 19:36:27 +0000 (21:36 +0200)]
Merge pull request #1978 from ktdreyer/doc-installing-ansible

[skip ci] docs: add instructions for installing Ansible

7 years agoMerge pull request #1974 from ceph/mgr-upgrade-luminous
Sébastien Han [Tue, 3 Oct 2017 17:57:31 +0000 (19:57 +0200)]
Merge pull request #1974 from ceph/mgr-upgrade-luminous

upgrade: a support for mgrs

7 years agodocs: add instructions for installing Ansible 1978/head
Ken Dreyer [Tue, 3 Oct 2017 17:38:58 +0000 (11:38 -0600)]
docs: add instructions for installing Ansible

Add RPM and DEB instructions for installing Ansible.

7 years agodocs: master requires ansible 2.3.1 1977/head
Ken Dreyer [Tue, 3 Oct 2017 17:26:43 +0000 (11:26 -0600)]
docs: master requires ansible 2.3.1

We require Ansible 2.3.1 or greater to run ceph-ansible's master branch.

7 years agoMerge pull request #1956 from ceph/osd-container-id
Sébastien Han [Tue, 3 Oct 2017 16:52:24 +0000 (18:52 +0200)]
Merge pull request #1956 from ceph/osd-container-id

Osd container

7 years agoMerge pull request #1975 from ceph/nfs-site
Sébastien Han [Tue, 3 Oct 2017 16:45:10 +0000 (18:45 +0200)]
Merge pull request #1975 from ceph/nfs-site

site: always play ceph-defaults and ceph-docker-common

7 years agoupgrade: a support for mgrs 1974/head
Sébastien Han [Tue, 3 Oct 2017 10:25:12 +0000 (12:25 +0200)]
upgrade: a support for mgrs

Also we now play ceph-config to have everything being generated for new
daemons bootstrap during upgrade.

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1497959
Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoiscsi: fix when condition 1956/head
Sébastien Han [Tue, 3 Oct 2017 14:48:17 +0000 (16:48 +0200)]
iscsi: fix when condition

generate_crt|bool|default(false) won't apply the default value, this
generate_crt|default(false)|bool will

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agosite: always play ceph-defaults and ceph-docker-common 1975/head
Sébastien Han [Tue, 3 Oct 2017 13:19:07 +0000 (15:19 +0200)]
site: always play ceph-defaults and ceph-docker-common

So we can later evaluate the conditions.
Also fix the variable, we are comparing ceph_release not
ceph_stable_release

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1486062
Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoosd: container use id instead of dev name
Sébastien Han [Wed, 27 Sep 2017 22:17:12 +0000 (00:17 +0200)]
osd: container use id instead of dev name

Closes: https://bugzilla.redhat.com/show_bug.cgi?id=1494127
Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agotest: add handler support
Sébastien Han [Mon, 2 Oct 2017 20:46:28 +0000 (22:46 +0200)]
test: add handler support

Add idempotency and handler test.

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoosd: do not copy admin key on collocated scenario
Sébastien Han [Mon, 2 Oct 2017 09:05:14 +0000 (11:05 +0200)]
osd: do not copy admin key on collocated scenario

ceph-disk used to have a bug requiring the admin key to store the
encrypted key in the mon kv store. This was reported in:
http://tracker.ceph.com/issues/17849

Fixed and backported here: https://github.com/ceph/ceph/pull/11996

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agotest: add test for device with 'by-path'
Sébastien Han [Fri, 29 Sep 2017 11:32:19 +0000 (13:32 +0200)]
test: add test for device with 'by-path'

We now test devices to be passed like:
/dev/disk/by-path/pci-0000:00:01.1-ata-1.0

Signed-off-by: Sébastien Han <seb@redhat.com>
7 years agoMerge pull request #1973 from ceph/refact_ceph-defaults
Sébastien Han [Tue, 3 Oct 2017 09:58:35 +0000 (11:58 +0200)]
Merge pull request #1973 from ceph/refact_ceph-defaults

defaults: change running order in main.yml

7 years agodefaults: change running order in main.yml 1973/head
Guillaume Abrioux [Mon, 2 Oct 2017 13:49:05 +0000 (15:49 +0200)]
defaults: change running order in main.yml

The task which sets `ceph_current_fsid` in `facts.yml` in case of containerized
deployment, will definitely fail because `docker_exec_cmd` is not set
yet.
This commits simply makes `facts.yml` played after `check_socket.yml` so
`docker_exec_cmd` is set properly.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
7 years agoMerge pull request #1968 from ceph/bz-1488999 v3.0.0rc14
Sébastien Han [Mon, 2 Oct 2017 12:42:08 +0000 (14:42 +0200)]
Merge pull request #1968 from ceph/bz-1488999

refact MDS role