Monitoring 101
summary
data dog claims it handles trillions of data points daily and uses Postgres with WAL‑E backups to provision read replicas quickly; alerting focuses on replication lag, base backup age and standby presence, not raw CPU. Aberdeen survey numbers are cited: ~$1k/hr small, $215k/hr medium, $700k/hr large. Practical fixes include lowering work_mem and watching temp files to force disk writes, using EXPLAIN ANALYZE / auto_explain to find bad queries, tagging metrics, anomaly detection and composite alerts to reduce noise.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 2:00Why you need monitoring
- 4:40Quality of metrics
- 10:50Types of metrics
- 12:35Work vs resource metrics
- 14:10Scaling Postgres
- 18:23Monitoring Postgres
- 20:34Work metrics
- 22:11Performance gains
- 24:27Slow queries
- 27:43Explanation Analyze
related talks
24mDevOps. Monitoring Databases. On-Premise and Cloud environments
59mMonitoring A-Z
41mKnow your data: The stats behind your alerts
49mProfessional PostgreSQL monitoring made easy
41mMonitoring Big Data Systems Done "The Simple Way"
12mUnlocking the Secrets of Client Data
35mDriving down the cost of observability
23mMastering Monitoring & Alerting: Key to Seamless Performance