Klarna assessment
- Create an user interface to perform 3 tasks of sum fibonacci, factorial, and ackermann
- Have a web service
- Collect relative metrics and logs
- have a dashboard for monitoring
Load Balancer: Nginx
Web Server: flask
Language: Python
Log collection: filebeat
Uptime monitoring: elastic-heartbeat
Metrics datastore: Elasticsearch
Visualization: Kibana/Grafana
Docker for orchestration
To bring up all the services at once, run the command below for the fist time and only run docker-compose up after
docker-compose up --build
To only startup the flask web:
bash start.sh
Web:
- Main page: http://localhost:8080/
- Fibonacci endpoint: http://localhost:8080/tasks/fibonacci
- ackermann: http://localhost:8080/tasks/ackermann
- factorial: http://localhost:8080/tasks/factorial
Visualization:
grafana login: admin/123456
grafana dashboard: http://localhost:3000/d/9o2OM3R7z/klarna-demo?orgId=1
kibana: http://localhost:5601/app/uptime
logs: http://localhost:5601/goto/f3f46861569d64331f9331aa99398e12
Level 1: - site uptime - measure from nginx endpoint to ensure the overall uptime
Level 2:
- site availability rate 
.
Not be able to done it here. formula ( total requests - "5xx errors")/ (total requests)
Level 3: - site latency - measure from nginx and use it as an indicator of overall site performnace
others: - response - top hit routes - count of each http status - service performance
Application - application logs - Application performance management tool integration
Architecture - queue system for logs and metrics - central logging endpoint
Depends - depends on the use case, we will use different solution
Dont have alerting yet. We can simply use grafana alerting to create monitor on top of those graphs. But grafana alerting is not highly recommend (very limited)
Pagerduty to distribute the alert to the support owner by schedule
Implement anomaly detection mechanism to auto generate alert when the bad behaviour detected