Monitoring Architecture

In above architecture diagram, there are two parts:

Application Host(s) - This virtual machine has running dockerized application along with node-explorer (to expose system metrics), promtail (to ship/expose logs metrics) and cadvisor (to expose container metrics). The application along with virtual machine host should be monitored by monitoring host, Grafana dashboards should show all metrics and loki explorer should show all application logs.
Monitoring Host - This virtual machine has monitoring prometheus server installed which gets metrics from application host(s) exposed by node-explorer, promtail and cadvistor etc and sends alerts to alertmanager based on prometheus rules.

Note: Monitoring tools can be installed in same application host machine, however, if you have many application host machines, it better to have a dedicated machine for monitoring all hosts.

Overview

This tool automatically generates deployment package required for monitoring host which includes configuration related to prometheus, blackbox, node-explorer, windows-exporer, cadvisor, loki and grafana.

Services for Application Host(s)

Please make sure below services are running on application host. (The configuration and deployment of services related to application host is not covered in this document, however you can find template files at path application.)

Application
- outputs log to file or console
- exposes at port 5000 (port can be anything)
node exporter
- an agent that gathers os metrics(cpu/memory/disk etc)
- 9100/metrics
Promtail
- an agent that ships the logs from the local system directly to the Loki cluster
- 9080/metrics: It can also expose endpoint related to custom metrics such as log count
cadvisor (if application is deployed in docker)
- Developed & maintained by Google which collects real-time monitoring data (resource usage, performance characteristics etc) of containers
- 8080/metrics

Services for monitoring host

prometheus
- metrics server
- 9090/metrics
blackbox exporter
- network monitoring agent
- it can monitor several protcols like tcp, ping, http, and so on
- 9115/metrics
Alertmanager
- send notification to somewhere like slack, mail, pagerduty and so on
- 9093/metrics
Loki
- log server
- it offers log recoding and searching
- 3100/metrics
Grafana
- viewer server
- it can send query to prometheus, loki, and so on
- 3000/metrics
cadvisor (used for docker containers running in monitoring host)
- Developed & maintained by Google which collects real-time monitoring data (resource usage, performance characteristics etc) of containers
- 8080/metrics
node exporter (used for monitoring host itself)
- an agent that gathers os metrics(cpu/memory/disk etc)
- 9100/metrics

Generate monitoring host deployment package automatically

To generate the package automatically, please follow below steps:

Go to .env file and fill it with required information related to monitoring host itself and information related to all the application hosts.
Run the bash script ./scripts/generate_skeleton.sh
The package will be generated at folder _package/monitoring. (Please refer sample package)
Now, you can copy monitoring folder to remote monitoring host machine.
The next step is to install monitoring/host/resource-monitoring.service

	cp monitoring/host/resource-monitoring.service /etc/systemd/system/resource-monitoring.service
	systemctl enable resource-monitoring.service

OR you can run docker compose directly

	cd monitoring/docker
	docker-compose up -d

Once all docker containers are running, please validate below urls:

	#prometheus
	curl -k http://localhost:9090/metrics
	#alertmanager
	curl -k http://localhost:9093/metrics
	#blackbox exporter
	curl -k http://localhost:9115/metrics
	#grafana
	curl -k http://localhost:3000/metrics
	#loki
	curl -k http://localhost:3100/metrics
	#cadvisor (related to monitoring host itself)
	curl -k http://localhost:8080/metrics
	#nodeexporter (related to monitoring host itself)
	curl -k http://localhost:9100/metrics