Sébastien Han [Mon, 10 Mar 2014 15:52:31 +0000 (16:52 +0100)]
Separate OSD scenarios to multiple files
Currently everything lives in main.yml, the file has become difficult to
read at some point and can be a real mess since we keep adding new
scenarios.
I think we should separate the scenarios into dedicated files and just
do includes in the main.yml file.
Closes: #16 Signed-off-by: Sébastien Han <sebastien.han@enovance.com>
Sébastien Han [Mon, 10 Mar 2014 23:47:24 +0000 (00:47 +0100)]
Autogenerate initial mon key
As mentionned in the issue 24 it's not really safe to store a default
fsid nor a monitor key. Thus the commit brings the auto-generation of
the initial monitor key. However it is quite complex to do the same for
the fsid, so I leave this to the person in charge of the deployment to
generate one and edit group_vars/all accordingly. The default fsid has
been removed as well.
Close: #24
Signed-off-by: Sébastien Han <sebastien.han@enovance.com>
Sébastien Han [Mon, 10 Mar 2014 16:14:11 +0000 (17:14 +0100)]
Allow to disable MDS
Even if MDS are not configured in site.yml the playbook has a
dependancy on the ceph.conf template.
This disables the mds section from the ceph.conf file.
Closes: #21 Signed-off-by: Sébastien Han <sebastien.han@enovance.com>
Sébastien Han [Fri, 7 Mar 2014 13:04:25 +0000 (14:04 +0100)]
Add multi-journal support and fix collocation scenario
This commits brings the support of multiple journals where each journal
points to a specific OSD and vice-versa. The commit also clarifies the
usage of multi scenarios for both journal and osd_data.
In the meantime, it fixes the collocation scenario.
Signed-off-by: Sébastien Han <sebastien.han@enovance.com>
Sébastien Han [Thu, 6 Mar 2014 10:12:25 +0000 (11:12 +0100)]
Support diverse setup disks
This brings the support of heterogeneous hardware. Not all the servers
are identical, some have more or less disks than the others. Prior this
commit, the 'parted' command was hanging, now the command simply exits 1
if the device doesn't exist, same for the 'egrep' piped command after.
Then we skip these errors and continue to run. So now, you can specify
multiple devices in group_vars/osds that don't exist on all the
servers.
Signed-off-by: Sébastien Han <sebastien.han@enovance.com>
projects / ceph-ansible.git / log