Diagnose four slow workloads from their metrics alone
Task
Apply the whole contention objective: given four sets of symptoms, name the mechanism and the remedy for each. Three describe contention you do not own; one describes a limit you bought.
Steps
- Write the four cases into
lab/cases.md, one paragraph each: (a) guest CPU 30%, steal time 25%, latency doubled; (b) guest memory looks fine, host shows heavy swapping, disk latency enormous; (c) a volume benchmarks at 3000 IOPS for 40 minutes then settles permanently at 120; (d) a guest with 16 vCPUs performs worse than the same workload on a guest with 4. - For each, write the mechanism in
lab/diagnosis.csvwith columnscase,mechanism,evidence,remedy,prevention. - For case (c), state the specific storage feature responsible and the volume type that removes it.
- For case (d), explain why more vCPUs made it slower, in terms of what the hypervisor scheduler must find free simultaneously.
- Add a row for a fifth case of your own, taken from a real system you have seen behave badly, with the evidence you actually had.
Verify
awk -F, 'NR>1 && NF>=5 {n++} END {print n" case(s) fully diagnosed"}' lab/diagnosis.csv
grep -Eci 'burst credit' lab/diagnosis.csv
grep -Eci 'ballo' lab/diagnosis.csv
grep -Eci 'provisioned iops' lab/diagnosis.csv
Five cases, and each of the three named mechanisms must appear. Missing "burst credit" means case (c) was diagnosed as a generic storage problem.
This is an independent study companion for CompTIA Cloud+ CV0-004 and is not produced by or endorsed by CompTIA.