Simple monitoring of Scylla with Grafana
Shell Python
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.

Scylla monitoring with Grafana and Prometheus

*** Notice to users using git ***

The monitoring stack was changed to use versions and releases.

Users should move to a stable release. Current release is 1.0, if you are using master do git checkout origin/branch-1.0 to switch to 1.0 version and make sure that you are using the latest stable version.


An out-of-the-box configuration will have a server dedicated to running both Prometheus and Grafana. Some teams may already have existing Promotheus and/or Grafana infrastructure, in which case you are able to use your existing architecture.

The monitoring infrastructure consists of several components, wrapped in Docker containers:

  • prometheus - collects and stores metrics
  • grafana - dashboard server
  • alertmanager - The alert manager collect Prometheus alerts


  • git
  • docker
  • python module pyyaml (for
  • python module json (for make_dashboards)

CentOS: Prerequisites Installation

On CentOS, you can do:

sudo yum install -y git docker python-pip
sudo pip install --upgrade pip
sudo pip install pyyaml

Ubuntu 16.04: Prerequisites Installation

On Ubuntu 16.04, you can do:

You'll need to add the Docker repo to your /etc/apt/sources.list (and accept the key, see Docker website for full instructions).

deb [arch=amd64] xenial stable

On Ubuntu, the latest package name is docker-ce for "Community Edition". You may want/need to adjust other Docker specific settings to meet your requirements. These instructions will get you a basic working Docker host.

sudo apt-get update && apt-get install -y python-pip docker-ce git
sudo pip install --upgrade pip
sudo pip install pyyaml


Installing archived project

Download the latest version from:

Installing source from git

git clone
cd scylla-grafana-monitoring

Start docker service if needed

ubuntu $ sudo systemctl restart docker
centos $ sudo service docker start

Update prometheus/scylla_servers.yml and prometheus/node_exporter_servers.yml with the targets (server you wish to monitor).

For every server, there are two targets, one under scylla job which is used for the scylla metrics. Use port 9180.

For example, update targets in prometheus/scylla_servers.yml :

- targets:

Second, for general node information (disk, network, etc.) add the server under node_exporter job. Use port 9100. For example, update targets in prometheus/node_exporter_servers.yml :

- targets:

You can also use your own target files instead of updating scylla_servers.yml and node_exporter_servers.yml, using the -s for scylla target file and -n for node taget file. For example:

./ -s my_scylla_servers.yml -n my_node_exporter_servers.yml -d data_dir

In many deployments the contents of those files are very similar, with the same servers being listed differing only in the ports scylla and node_exporter listen to. To automatically generate the target files, one can use the script, using the -n and -s flags to control which files get created:

./ -ns -d myconf

After that, the monitoring stack can be started pointing to the servers at and with::

./ -s myconf/scylla_servers.yml -n myconf/node_exporter_servers.yml

node_exporter Installation

node_exporter is an exporter of hardware and OS metrics such as disk space.

For a fully functional dashboard you need to have the node_exporter running on each of the nodes and configure the prometheus accordingly.

As part of Scylla installation, the scylla_setup script will prompt to install node_exporter. If you skipped that step, you could always install node_exporter later with the node_exporter_install script.

node_exporter_install will download and install the node_exporter as a service.


./ -d data_dir

For full list of options

./ -h

Multiple versions support

As counters change their names between versions, we create a new dashboard for each new version. We use tags to distinguish between the different versions, to keep the dashboard menu, relatively short, by default, only the last two releases are loaded. You can load specific versions by using the -v flag.

  • You can supply multiple comma delimited versions, for example to load only 1.5 and 1.6 version:
./ -v 1.5,1.6
  • Use the all to load all available versions.

  • The master branch is called master, so to load 1.6 and master you would use:

./ -v 1.6,master
  • If you only need the latest version you can use:
./ -v latest

Note: The -d data_dir is optional, but without it, Prometheus will erase all data between runs.

For systems in production it is recommended to use an external directory.

Prometheus Command Line Options

-b storage.local.retention=1000h -b query.staleness-delta=1m

connecting Scylla and the Monitoring locally - the local flag

When running the Prometheus and Grafana on the same host as scylla, use the local -l flag, so processes inside the containers will share the host network stack and would have access to the localhost.




Direct your browser to your-server-ip:3000 By default, Grafana authentication is disabled. To enable it and set a password for user admin use the -a option

Choose Disk and network interface

The dashboard holds a drop down menu at its upper left corner for disk and network interface. You should choose relevant disk and interface for the dashboard to show the graphs.

Update Scylla servers to send metrics

See here

Load original data to Prometheus server

Additional parameters: -d data_dir

Full commandline:

./ -d data_dir

Comment: data_dir is the local path to original data directory

Data source for Prometheus data:

Using your own Grafana installation

Some users who already have grafana installed can just upload the Scylla dashboards into your existing grafana environment. This is possible using the script.

For example, if you have prometheus running at, and grafana at localhost's port 3000, you can do:

./ -p -g 3000


Prometheus Alertmanager handles alerts that are generated by the Prometheus server.

Alerts are generated according to the Alerting rules.

The Alertmanager listen on port 9093 and you can use a web-browser to connect to it.