Aura_infrastructure

Ansible/Docker project for operating Aura's servers

How to contribute

Please refer to How To Contribute

tl;dr

Create the group_vars/all/secrets.yml file with the following variables:

jupyter_client_id: <GitHub OAuth application Client ID for JupyterHub>
jupyter_client_secret: <GitHub OAuth application Client Secret for JupyterHub>

ansible-playbook -vv --diff -i inventories/dev.yml --key-file "/PATH/TO/YOUR/KEY/FILE" install.yml [-t timescaledb -t jupyter -t reverse_proxy]

Prerequisites

• Ansible v2.4.2

Development environment

Prerequisites

To deploy this environment on Amazone ec2 instance :

• create a config file in your local machin .ssh directory and add Host aura_aws_timescale Hostname YOUR IP ADDRESS User centos IdentityFile /PATH/TO/YOUR/KEY/FILE

• in your local machin /etc/hosts add following lines :

  • YOUR EC2 IP ADDRESS aura_aws_timescale
  • YOUR EC2 IP ADDRESS elasticsearch.aura.healthcare.local.aws.timescale kibana.aura.healthcare.local.aws.timescale monitor.aura.healthcare.local.aws.timescale

• in /TimescaleDB/group_vars/all/main.yml set aura_local_ip variable to aura_local_ip = "YOUR EC2 IP ADDRESS"

Everytime you stop and start the ec2, you need to change IP address in previous files.

ELK stack

Five docker containers compone a monitoring plateform

Filebeat

Filebeat is a lightweight logs shipper. It is only used to recover logs from all docker containers.

Metricbeat

Metricbeat is a lightweight metric shipper. It is only used to recover metrics from all docker containers.

Logstash

Two logs files are created to gather informations about data loading in influxdb database. Logstash harvests, transforms and analyzes these logs and sends them to elasticsearch.

Metrics and logs from above-mentionned beat logs shippers are also retrieved by logstash but not processed. They are also sent to elasticsearch.

Elasticsearch

Elasticsearch stores and index data sent by logstash. Using logstash as output for beat logs shippers forces to load manually indexes templates. - filebeat-6.6.2-* is used for docker containers logs from filebeat - metricbeat-6.6.2-* is used for docker containers metrics from metricbeat - logstash-* is used for personal logs harvested by logstash itself

Kibana

Kibana is used to display relevant dashboard. Two dashboard are uploaded in kibana.

The container_monitoring_dashboard allows to follow docker containers system metrics. It displays CPU, memory, number of container etc ...

The TimescaleDB_monitoring_dashboard monitors the influxdb container. This dashboard displays influxdb's metrics. It adds two visualizations : - Nb input / time displays the total amount of data about to be stored during an injection along the time.

To load your own dashboard, follow the steps: - create your visualizations with the kibana UI - create your dashboard with the kibana UI using the visualizations you have created before - download your dashboard's json format using the REST API: curl -X GET "kibana.aura.healthcare.local.aws.timescale:80/api/kibana/dashboards/export?dashboard=*id*" -H 'kbn-xsrf: true' replace id by yout dashboard's id in kibana - add the json script in a file in /kibana/files - add a task in /kibana/tasks/run_container.yml (just copy paste a load task and change the file in the body module) to upload your dashboard in kibana each time you rebuild the stack.

Problem

When TimescaleDB docker container is deployed, you have to modify manually following line's value : host all all all md5 by host all all all trust located in var/lib/postgresql/data/pg_hba.conf