Preamble
CPython’s memory story is simpler in day-to-day ops than the JVM’s: one big process, reference counting plus cycle GC, fewer knobs in production dashboards. Java exposes heap sizing, GC algorithms, and pause behavior explicitly. I do not need to become a GC researcher; I do need to read ops charts without glazing over.
Heap sizing: -Xms and -Xmx
-Xmx caps maximum heap; -Xms sets initial heap. Wild mismatches cause resize churn early in life; aligning them is a common stability tweak—tune with measurement, not superstition.
GC logging
Modern JDKs ship unified GC logging (-Xlog:gc*, flags vary by version). I correlate pause times with p99 latency spikes the same way I correlate Python’s GIL contention with tail latency in threaded services—different mechanisms, same skeptical eye.
When to escalate
If pauses dominate SLOs after sane heap tuning, I pull in specialists or experiment with collector choices on representative workloads—not on laptop demos alone.
Conclusion
Enough literacy to reproduce issues locally and interpret dashboards closes most incidents. asyncio Deep Dive: Cooperative Concurrency Without Celery returns to asyncio as Python’s cooperative I/O story.