Prometheus collects and stores its log-metrics as time series data, i.e. metrics information is stored with the timestamp at which it was recorded, alongside optional key-value pairs called labels.
/PCA/exam-mock.md
/PCA/exam-mock-answer.md
- Demo Prometheus UI: https://demo.promlabs.com/
- Demo Proemtheus Metrics: https://demo.promlabs.com/metrics
/getting-started/setup
/getting-started/instrumentator
/getting-started/recording-rules
/getting-started/alert-rules
/getting-started/alertmanager
/getting-started/blackbox-exporter
/getting-started/pushgateway
/getting-started/service-discovery
/getting-started/promql
/kubernetes/setup
/kubernetes/exporter
/kubernetes/instrumentator
/kubernetes/grafana
/aws/deploy_prometheus_ec2_server
/aws/scrape_ec2_instance
- https://blog.cloudflare.com/how-cloudflare-runs-prometheus-at-scale/
- https://blog.cloudflare.com/monitoring-our-monitoring/
- https://www.robustperception.io/blog/
- https://satyanash.net/software/2021/01/04/understanding-prometheus-range-vectors.html
- https://sre.google/sre-book/monitoring-distributed-systems/
- https://www.robustperception.io/whats-the-difference-between-group_interval-group_wait-and-repeat_interval/
- https://www.atatus.com/blog/promql-introduction/
- https://promlabs.com/promql-cheat-sheet/
- https://jaanhio.me/blog/understanding-alertmanager/
- https://www.blameless.com/the-comprehensive-guide-on-slis-slos-and-error-budgets