The goal is to manage multiple Ceph clusters with a single salt master. At this time, only a single Ceph cluster can be managed.
This diagram should explain the intended flow for the orchestration runners and related salt states.
DeepSea currently supports the following functionality:
To learn more about DeepSea, take a look at the Wiki.
There is also a dedicated mailing list deepsea-users. If you have any questions, suggestions for improvements or any other feedback, please join us there! We look forward to your contribution.
If you think you've found a bug or would like to suggest an enhancement, please submit it via the bug tracker on GitHub.
If you would like to contribute to DeepSea, refer to the contribution guidelines.
For those interested in learning about some of the uses of Salt in DeepSea, see here for explanations and examples.
You need at least a minimum of 4 nodes to be able to test and use DeepSea properly.
To be able to use less than 4 nodes during the deployment stages (e.g. in a
development/testing environment), you could set the option
DEV_ENV=true as an
environment variable or globally as a pillar variable in
DEV_ENV allows you to:
DeepSea needs to be installed on the Salt master node.
On other distributions or when using a development version, install Salt via the distribution's preferred installation method and check out the git repository:
$ git clone https://github.com/SUSE/DeepSea.git $ cd DeepSea $ sudo make install
The cluster deployment process has several phases. First, you need to prepare all nodes of the cluster by configuring Salt and then deploy and configure Ceph.
The following procedure describes the cluster preparation in detail.
Install a minimum of four machines (or define
DEV_ENV=true either as an environment variable or a pillar variable during the deployment process) with SUSE Leap 42.3 or Tumbleweed and add the DeepSea repo to your defined "admin" node:
# zypper ar http://download.opensuse.org/repositories/filesystems:/ceph:/luminous/openSUSE_Leap_42.3/filesystems:ceph:luminous.repo # zypper refresh
Make sure that each node can resolve the host names of all other nodes.
The Salt master needs to resolve all its Salt minions by their host names, as well as all Salt minions need to resolve the Salt master by its host name.
If you don't have a DNS Server you could also add all hosts of your cluster to the ``/etc/hosts``` file.
Configure, enable, and start the NTP time synchronization server on all nodes:
# systemctl enable ntpd.service # systemctl start ntpd.service
Note that the Cluster deployment with DeepSea will not work with Firewall/AppArmor enabled.
Check whether the AppArmor service is running and disable it on each cluster node to avoid problems:
# systemctl disable apparmor.service
Check whether the Firewall service is running and disable it on each cluster node to avoid problems:
# systemctl disable SuSEfirewall2.service
Install DeepSea on the node which will be the Salt master:
# zypper install deepsea
The command installs the
salt-minion packages as a dependency.
Check that the
salt-master service is enabled and started - enable and start it if needed:
# systemctl enable salt-master.service # systemctl start salt-master.service
Install the package
salt-minion on all minion nodes:
# zypper in salt-minion
Configure all minions (including the master minion) to connect to the master. If
your Salt master is not reachable by the host name "salt", edit the file
/etc/salt/minion or create a new file
the following content:
If you performed any changes to the configuration files mentioned above, restart the Salt service on all Salt minions:
# systemctl restart salt-minion.service
Check that the file
/srv/pillar/ceph/master_minion.sls on the Salt master
points to your Salt master.
If your Salt master is reachable via more host names, use the one suitable for the storage cluster. If you used the default host name for your Salt master in the example domain, then the file looks as follows:
Check that the
salt-minion service is enabled and started on all nodes (including
the master node). Enable and start it if needed:
# systemctl enable salt-minion.service` # systemctl start salt-minion.service
Accept all salt keys on the Salt master:
# salt-key --accept-all
Verify that the keys have been accepted:
# salt-key --list-all
In order to avoid conflicts with other minions managed by the Salt master, DeepSea
needs to know which Salt minions should be considered part of the Ceph cluster to
be deployed. This can be configured in file
by defining a naming pattern. By default, DeepSea targets all minions that have a
deepsea applied to them. This can be accomplished by running the following
Salt command on all Minions that should be part of your Ceph cluster:
# salt -L <list of minions> grains.append deepsea default
Alternatively, you can change
deepsea_minions.sls to any valid
Salt target definition. See
man deepsea-minions for details.
Prior to deploying a cluster with DeepSea make sure that all disks that were used as OSD by previous clusters are empty without partitions. To ensure this, you have to manually zap all the disks. Remember to replace 'X' with the correct disk letter.
Attention - This command will remove all your data from that disk, please be careful!
Wipe the beginning of each partition:
# for partition in /dev/sdX[0-9] ; do \ dd if=/dev/zero of=$partition bs=4096 count=1 oflag=direct ; done
Wipe the partition table:
# sgdisk -Z --clear -g /dev/sdX
Wipe the backup partition tables:
# size=`blockdev --getsz /dev/sdX` # position=$((size/4096 - 33)) # dd if=/dev/zero of=/dev/sdX bs=4096 count=33 seek=$position oflag=direct
Now you deploy and configure Ceph. Unless specified otherwise, all steps are mandatory.
Note: Salt Command Conventions
There are two possibilities how to run
salt-run state.orch - one is with
stage.<stage number>, the other is with a name of the stage. Both notations
have the same impact and it is fully up to your preference which command you
want to use. Both notations are used in the following deployment steps. So
please choose what you prefer.
During this stage all required updates are applied and your system may be rebooted.
salt-run state.orch ceph.stage.0
salt-run state.orch ceph.stage.prep
Note: If during Stage 0 the Salt master reboots to load new kernel version, you need to run Stage 0 again, otherwise minions will not be targeted.
The discovery stage collects all hardware in your cluster and also collects
necessary information for the Ceph configuration. The configuration fragments
are stored in the directory
The data is stored in YAML format in
salt-run state.orch ceph.stage.1
salt-run state.orch ceph.stage.discovery
After the previous command finishes successfully, create a
policy.cfg file in
You can find an example in our
folder. Please change the example file to fit to your needs, e.g. by changing
If you need more detailed information please refer to the Policy wiki page.
If you need to change the cluster's network setting, edit
/srv/pillar/ceph/proposals/config/stack/default/ceph/cluster.yml and adjust
the lines starting with
The configuration stage parses the
policy.cfg file and merges the included files
into their final form. Cluster and role related contents are placed in
/srv/pillar/ceph/cluster, while Ceph specific content is placed in
Run the following command to trigger the configuration stage:
# salt-run state.orch ceph.stage.2
# salt-run state.orch ceph.stage.configure
The configuration step may take several seconds. After the command finishes, you
can view the pillar data for the specified minions (for example named
ceph_minion2 ...) by running:
# salt 'ceph_minion*' pillar.items
Now you run the deployment stage. In this stage, the pillar is validated and monitors and ODS daemons are started on the storage nodes. Run the following to start the stage:
# salt-run state.orch ceph.stage.3
# salt-run state.orch ceph.stage.deploy
The command may take several minutes. If it fails, you have to fix the issue and run the previous stages again. After the command succeeds, run the following to check the status:
# ceph -s
If you only want to deploy a Ceph cluster without any additional services, congratulations - you're done. Otherwise you have to continue with Stage 4.
The last step of the Ceph cluster deployment is the services stage. Here you instantiate any of the currently supported services: iSCSI Gateway, CephFS, RADOS Gateway, and NFS Ganesha. In this stage, the necessary pools, authorizing keyrings and starting services are created.
To start the stage, run the following:
# salt-run state.orch ceph.stage.4
salt-run state.orch ceph.stage.services
Once a cluster is deployed one might want to verify functionality or run benchmarks to verify the cluster works as expected.
In order to gain some confidence in your cluster after the inital deployment (stage 3) run:
# salt-run state.orch ceph.benchmarks.baseline
This runs an osd benchmark on each OSD and aggregates the results. It reports your average OSD performance and points out OSDs that deviate from the average.
Please note that for now the baseline benchmark assumes all uniform OSDs.
To load test CephFS run:
# salt-run state.orch ceph.benchmarks.cephfs
This requires a running MDS (deploy in stage 4) and at least on minion with the
client-cephfs role. The
cephfs_benchmark stage will then mount the CephFS
instance on the mds-client and run a bunch of
fio tests. See the benchmark
readme for further details.
# salt-run state.orch ceph.benchmarks.rbd
This runs fio using the RBD backend against the cluster as a whole. This
requires at least one minion with the
benchmark-rbd role. See the
benchmark readme for further details.