Module 06 · Memory management · Drill

Memory management drill

Scenario prompts that check whether memory management is usable under pressure. Answer out loud, then compare.

4 prompts25 minNew

0 of 4 graded

Prompt 1
A Kubernetes pod running a JVM service with -Xmx4g and a memory limit of 4.5 GB is killed every few hours with exit code 137, although heap usage graphs never exceed 3.2 GB. Explain where the memory goes, how you would confirm it, and how you would size the limit.
Say your answer out loud first. Thirty seconds is enough.
Prompt 2
A search service memory-maps a 60 GB index on 64 GB machines. After a deploy that added a feature, p99 latency rises from 15 ms to 400 ms, CPU drops, and disk read IOPS climbs. Explain the mechanism, what metrics prove it, and three ways to fix it.
Say your answer out loud first. Thirty seconds is enough.
Prompt 3
An in-memory key-value store with a 200 GB dataset on a 256 GB machine shows high CPU with a low instructions-per-cycle rate, and perf stat reports a very high dTLB miss rate. Explain why, and what you would change at the OS and data-structure level.
Say your answer out loud first. Thirty seconds is enough.
Prompt 4
A Go service's resident memory grows from 1 GB to 6 GB over a day while the live heap reported by the runtime stays near 800 MB, and a C++ sidecar on the same node shows similar slow growth with no leaks. Diagnose both, and explain how garbage collection settings and allocator choices interact with container limits.
Say your answer out loud first. Thirty seconds is enough.