Guided text based installation UI, that runs ceph-ansible. The goal of the project is to provide a UI over the top of the complexity of ceph-ansible. Users new to Ceph may then interact with the UI to build their first Ceph cluster instead of first building up their Ansible knowledge.
Does it work? Well in my testing, the short answer is YES but it needs a heap more testing on more varied configurations!
The UI elements are all in place, as is the Ansible playbook integration and updates of the Ansible yml files.
What does it look like?
Here's an example run that illustrates the workflow for a small cluster of 3 nodes.
A more detailed demo can be found on my blog and there are some screenshots showing the welcome page across the major distros in the screenshots folder.
- Added ssh credentials workflow
- hosts selected are checked for passwordless ssh access (root only at the moment)
- for hosts that fail the ssh check, the admin may enter a root password common to all hosts or specify a password per host
- copilot uses the passwords(s) provided to setup passwordless ssh
- once all hosts have passed the ssh check, the UI continues to the next page (host validation)
- the UI's color palette is based on the Linux distribution
- perform cluster & host sanity checks (primarily useful for prod deployments)
- support two deployment modes - production and development
- the mode is selected using a --mode switch when copilot is started
- in dev mode, pretty much anything is allowed (the default!)
- in prod mode;
- confirm the number of hosts are appropriate for the intended role
- confirm host specs are appropriate (e.g. cpu core count matches #disks for OSD hosts)
- prevent ceph daemon collocation
- supports the following ceph roles; MONs, OSDs, radosgw and MDS
- defaults to a cluster name of 'ceph' (use --cluster-name= to override)
- allows selection of OSD type, encryption, hosts and installation source
- validates deployment user exists (locally)
- hosts may be specified by name or mask
- hosts are checked for DNS (2s dns lookup timeout per host - default)
- hosts are checked for passwordless ssh
- candidate hosts are probed (using Ansible)
- host specs are shown
- hosts are validated against the intended role
- public and cluster networks are detected from the host probe
- admin may choose which networks are used based on the ones detected
- rgw interface defaults to the NIC on the public network in this release
- uses plugins to create ceph-ansible group_vars files
- performs host sanity checks to confirm the host spec is appropriate for the intended role
- deployment playbook may be passed at run time
- if the playbook file does not exist, the program aborts before starting the deployment process is tracked 'live' in the UI
- if the deployment fails,
- the playbook may be rerun from the UI
- playbook failures are shown in the UI (hostname and stderr)
- deployment playbook output is logged to a file for diagnostics (in addition to the normal /var/log/ansible.log file)
How do I install and run it?
- ansible-2.4.x (tested against 2.4.1 and 2.4.2)
- ceph-ansible - tested against Master (Dec 2017)
For 2.4.x of Ansible you may need to enable additional repositories depending on your distribution.
- RHEL - enable extras with
subscription-manager repos --enable=rhel-7-server-extras-rpms
- Ubuntu - enable the ansible ppa with
- All testing has been done against CentOS7
- 'copilot' is tied to the Ansible 2.4.x python API. Attempting to use an older version of Ansible will fail.
- download the archive to your ansible server
- extract the archive and 'cd' to it
- run the setup program
> python setup.py install --record files.txt
- to run copilot
> cd /usr/share/ceph-ansible > copilot
- You need to cd to the ceph-ansible directory, since the playbook needs to reference ceph-ansibles roles, actions etc
- If you're not using the root account, you'll need to use sudo for steps 3 and 4.
Here's some ideas on how copilot could evolve;
- support non-root user Installation
- support iSCSI gateways role
- post install configuration..what about enabling add-on pages through plugins for radosgw config for example.
- add more debug information to improve support (e.g. ansible messages, host information)
Crystal ball gazing - why not take this workflow to a cockpit plugin to provide a web frontend to the ceph-ansible installation process? Now that would be cool!