Monitoring tells you something is wrong; observability lets you ask why without shipping new code. The three pillars - metrics, logs and traces - each answer a different question, and the magic is in connecting them.
Metrics: is it wrong, and how badly?
sum(rate(http_requests_total{status=~"5.."}[5m])) / sum(rate(http_requests_total[5m]))Traces: where in the request did it break?
from opentelemetry import trace
tracer = trace.get_tracer(__name__)
with tracer.start_as_current_span("charge_card") as span:
span.set_attribute("order.id", order_id)
charge(order_id)Logs: what exactly happened?
{app="web"} | json | trace_id="4bf92f3577b34da6" # jump from a trace straight to its logsThe payoff is correlation: an alert (metric) links to the slow trace, and the trace_id links to the exact log lines. You go from 'error rate is up' to the offending line in three clicks.
Buy or build whatever you like - just make sure a metric can lead you to a trace, and a trace to its logs.
Standardize on OpenTelemetry for instrumentation and keep a shared trace_id across pillars; that single convention is what turns three tools into one investigation.