Scylla monitoring with Grafana and Prometheus

---

** Notice to users **

The repository name was changed to scylla-monitoring make sure you are using the right path.

---

Introduction

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

Prerequisites

• docker
• python module pyyaml (for genconfig.py)
• python module json (for make_dashboards)

CentOS: Prerequisites Installation

On CentOS, you can do:

Install Docker

Please follow the instructions at: https://docs.docker.com/install/linux/docker-ce/centos/ Note that Docker was renamed and it is now docker-ce. If you are running older docker version, remove it first by following the instruction.

Complete Prerequisites Installation

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

Ubuntu 16.04: Prerequisites Installation

On Ubuntu 16.04, you can do:

Install Docker

Please follow the instructions at: https://docs.docker.com/install/linux/docker-ce/ubuntu/ Note that Dokcer was rename and it is now docker-ce. If you are running older docker version, remove it first by following the instruction.

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.

Complete Prerequisites Installation

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

Docker Post Installation

Docker post installation guide can be found here: https://docs.docker.com/install/linux/linux-postinstall/

Note: It is recommended not to run container as root

To avoid running docker as root, you should add the user you are going to use to start the monitoring to the docker group.

1. Create the docker group.

sudo groupadd docker

2. Add your user to the docker group.

sudo usermod -aG docker $USER

3. Start docker by calling:

sudo systemctl enable docker

Install

Installing archived project

Download the latest version from: https://github.com/scylladb/scylla-monitoring/releases

Installing source from git

git clone https://github.com/scylladb/scylla-monitoring.git
cd scylla-monitoring

Start docker service if needed

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

Configuration

In standard installations of Scylla, each node in the cluster provides two sources of metrics: Scylla itself (on port 9180), and a "node exporter" process which provides (on port 9100) standard hardware and OS metrics. We need to tell Prometheus the list of nodes which provides each of these two sources of metrics.

By default, the start-all.sh script (which we will use to run Prometheus and Grafana) gets the configuration of these two sources from the files prometheus/scylla_servers.yml and prometheus/node_exporter_servers.yml. These files should be edited to list the Scylla nodes, a.k.a. targets.

For example, if you have two nodes (172.17.0.2 and 172.17.0.3) in a single dc cluster, update prometheus/scylla_servers.yml to say they provide Scylla metrics on port 9180:

- targets:
      - 172.17.0.2:9180
      - 172.17.0.3:9180
  labels:
       cluster: cluster1
       dc: dc1

specify node_exporter file

By default, the same configuration (with different ports) will be used for the node_exporter. You can specify a specific node_exporter target file with the -n flag.

A typicall node_exporter configuration file would look like:

- targets:
      - 172.17.0.2:9100
      - 172.17.0.3:9100
  labels:
       cluster: cluster1
       dc: dc1

Clusters and Data centers

Note that each "targets" section (there could be more than one) come with its own cluster and dc labels. For multiple DC or multiple cluster create multiple "targets" entries, each with the right cluster or dc.

Using your own target files

You can also use your own target files instead of updating scylla_servers.yml, using the -s for scylla target file.

./start-all.sh -s my_scylla_servers.yml -d data_dir

To automatically generate the target files, one can use the genconfig.py script.

./genconfig.py -n -d myconf 192.168.0.1 192.168.0.2

After that, the monitoring stack can be started pointing to the servers at 192.168.0.1 and 192.168.0.2 with::

./start-all.sh -s myconf/scylla_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.

For node_exporter users that use version 0.16 and higher and use their own Prometheus server, look at the node_exporter configuration section.

Run

./start-all.sh -d data_dir

For full list of options

./start-all.sh -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:

./start-all.sh -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:

./start-all.sh -v 1.6,master

• If you only need the latest version you can use:

./start-all.sh -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.

Kill

./kill-all.sh

Use

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:

./start-all.sh -d data_dir

Comment: data_dir is the local path to original data directory

Data source for Prometheus data:

• Download from Docker Prometheus server, reference: https://github.com/scylladb/scylla/wiki/How-to-report-a-Scylla-problem#prometheus
• Get from Scylla-Cluster-Test log.
• Others

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 load-grafana.sh script.

For example, if you have prometheus running at 192.168.0.1:9090, and grafana at localhost's port 3000, you can do:

./load-grafana.sh -p 192.168.0.1:9090 -g 3000

Using your own Prometheus server

node_exporter specifc configuration

Scylla dashboard are using the metrics from node_exporter 0.14 version. To use version 0.16 and higher an additional configuration is needed. Under the node_exporter job you should configure a mapping between the new names and the old:

Under

- job_name: node_exporter
  metric_relabel_configs:

Add the following part that would rename the metrics to be backward compatible.

    - source_labels: [__name__]
      regex:  'node_disk_read_bytes_total'
      target_label: __name__
      replacement: 'node_disk_bytes_read'
    - source_labels: [__name__]
      regex:  'node_disk_written_bytes_total'
      target_label: __name__
      replacement: 'node_disk_bytes_written'
    - source_labels: [__name__]
      regex:  'node_disk_reads_completed_total'
      target_label: __name__
      replacement: 'node_disk_reads_completed'
    - source_labels: [__name__]
      regex:  'node_disk_writes_completed_total'
      target_label: __name__
      replacement: 'node_disk_writes_completed'
    - source_labels: [__name__]
      regex:  'node_filesystem_avail_bytes'
      target_label: __name__
      replacement: 'node_filesystem_avail'
    - source_labels: [__name__]
      regex:  'node_network_receive_bytes_total'
      target_label: __name__
      replacement: 'node_network_receive_bytes'
    - source_labels: [__name__]
      regex:  'node_network_receive_packets_total'
      target_label: __name__
      replacement: 'node_network_receive_packets'
    - source_labels: [__name__]
      regex:  'node_network_transmit_bytes_total'
      target_label: __name__
      replacement: 'node_network_transmit_bytes'
    - source_labels: [__name__]
      regex:  'node_network_transmit_packets_total'
      target_label: __name__
      replacement: 'node_network_transmit_packets'
    - source_labels: [__name__]
      regex:  'node_filesystem_size_bytes'
      target_label: __name__
      replacement: 'node_filesystem_size'

Alertmanager

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.