Monitoring & Logging

ELK pools every service's logs into one searchable place, and a correlation ID lets you pull the complete story of a single request out of billions of log lines with one filter.

Know the small set of flags that size the heap, pick and tune the collector, cap Metaspace, capture a heap dump on failure, log GC, and print the effective configuration.

Use jstat and jcmd for live inspection, MAT and async-profiler for deep analysis, and Micrometer-Prometheus-Grafana plus APM for continuous alerting on old-gen percentage, GC pause p99, GC time ratio, and monotonic heap growth.

Prometheus pulls metrics on a schedule into a time-series DB queried by PromQL, Grafana visualizes it, and good alerts fire on sustained user-facing symptoms, not momentary blips.

Latency, traffic, errors, and saturation are the four Golden Signals: serve-time, demand, failure rate, and how-full, and together they tell you nearly everything about a user-facing service.

Metrics tell you something is wrong, traces tell you where, and logs tell you why.

One time-series per unique label combination means an unbounded label like account ID or trace ID explodes series into the millions and crashes Prometheus: keep those identifiers in logs and traces, never in metric labels.

eBPF runs safe sandboxed programs in the Linux kernel to reveal syscalls, network drops, scheduler and disk waits with zero app changes: reach for it when a latency spike or restart has no application-level explanation.

Read CPU, memory, and disk utilization, compare each to a configurable threshold, and alert plus exit non-zero the moment any one crosses it.