cephmetrics

Cephmetrics is a tool that allows a user to visually monitor various metrics in a running Ceph cluster.

Prerequisites

RHEL 7 should be running on all hosts
A functional ceph cluster running version ceph-osd-10.2.7-27.el7cp.x86_64 or later is already up and running.
Another host machine independent of the ceph machines must be available. This host will be used to receive data pushed by the hosts in the Ceph cluster, and will run the dashboard to display that data.
A host machine on which to execute ansible-playbook to orchestrate the deployment must be available.
Passwordless SSH access from the deploy host to the ceph hosts. The username should be the same for all hosts.
Passwordless sudo access on the ceph and dashboard hosts
All hosts must share the same DNS domain

Resulting configuration

After running this procedure, you will have the following configuration.

The ceph nodes will have collectd installed, along with collector plugins from cephmetrics-collectd
The dashboard host will have grafana installed and configured to display various dashboards by querying data received from Ceph nodes via a graphite-web, python-carbon, and python-whisper stack.

Installation

Install cephmetrics-ansible

First, decide which machine you want to use to run ansible-playbook. If you used ceph-ansible to set up your cluster, you may want to reuse that same host to take advantage of the inventory file that was created as part of that process.

Once the host is selected, perform the following steps there. This will install a repo which includes the cephmetrics installation code and ansible (version 2.2.3 or later):

sudo su -
mkdir ~/cephmetrics
subscription-manager repos --enable rhel-7-server-optional-rpms --enable rhel-7-server-rhscon-2-installer-rpms
curl -L -o /etc/yum.repos.d/cephmetrics.repo http://download.ceph.com/cephmetrics/rpm-master/el7/cephmetrics.repo
yum install cephmetrics-ansible

Create or edit the inventory file

Next, we need an inventory file. If you are running ansible-playbook on a host that previously ran ceph-ansible, you may simply modify /etc/ansible/hosts; otherwise you may copy /usr/share/cephmetrics-ansible/inventory.sample and modify it if you wish.

The inventory file format looks like:

[ceph-grafana]
grafana_host.example.com

[osds]
osd0.example.com
osd1.example.com
osd2.example.com

[mons]
mon0.example.com
mon1.example.com
mon2.example.com

[mdss]
mds0.example.com

[rgws]
rgw0.example.com

If you are running ansible-playbook on a host mentioned in the inventory file, you will need to append ansible_connection=local to each line in the inventory file that mentions that host. An example:

    my_host.example.com ansible_connection=local

Omit the mdss section if no ceph mds nodes are installed. Omit the rgws section if no rgw nodes are installed.

Ansible variables can be set in a vars.yml file if necessary. If it is required, make sure to add -e '@/path/to/vars.yml to your ansible-playbook invocation below. Click here for more information.

Deploy via ansible-playbook

If you are using a ceph-ansible host, run these commands:

cd /usr/share/cephmetrics-ansible
ansible-playbook -v playbook.yml

Otherwise, run these commands:

cd /usr/share/cephmetrics-ansible
ansible-playbook -v -i /path/to/inventory playbook.yml

Note: The reason it is necessary to change directories is so that ansible-playbook will use the bundled ansible.cfg; there is currently no command-line argument allowing the specification of an arbitrary .cfg file.

ceph / cephmetrics