Event capacity planning quiz
Related document: Event capacity planning
1. How are multiple KEDA metrics used to determine the target?
- A) Sum every value
- B) HPA combines per-metric recommendations through a maximum and applies behavior/min/max
- C) Cron guarantees Ready Pods
- D) Metrics always define the upper ceiling
Show answer
Answer: B
Cron supplies a scheduled demand floor; maxReplicaCount and other constraints bound it. Verify actual readiness separately.
2. What matters when using a five-field Cron for a dated event?
- A) The year is implicit
- B) It deletes itself after one run
- C) It can recur next year and needs timezone/DST checks
- D) It always uses UTC only
Show answer
Answer: C
The example tests annual recurrence and boundaries. Remove or update the event configuration afterward.
3. Which configuration describes a static Karpenter NodePool?
- A) Higher weight alone creates static capacity
- B) Use replicas without weight and constrain limits.nodes
- C) Remove replicas at any time to switch to dynamic
- D) Scale operations are always blocked by every NodePool budget
Show answer
Answer: B
Distinguish static/dynamic modes. Static scale bypasses NodePool disruption budgets but respects PDBs.
4. What can happen if placeholder Deployment desired count remains unchanged after handoff?
- A) It always remains free of additional cost
- B) Recreated placeholders can cause additional nodes to be provisioned
- C) Karpenter automatically sets desired to zero
- D) The real application is always immediately Ready
Show answer
Answer: B
Validate retention/expiry, removal of placeholder demand and actual application readiness together.
5. Does adding 30% capacity guarantee survival of one AZ loss?
- A) Always
- B) Yes for On-Demand
- C) No; calculate placement and surviving throughput
- D) Yes whenever a PDB exists
Show answer
Answer: C
The 15/11-node example fails baseline peak after losing the larger AZ. Validate the input assumptions with measurements too.
6. What does a targeted Capacity Reservation mean?
- A) A security ACL restricted to one NodePool
- B) Instances explicitly target it and must satisfy matching criteria
- C) It automatically applies to every existing instance
- D) An On-Demand discount product
Show answer
Answer: B
Check IAM/sharing, AZ, type, platform, tenancy and actual consumption. It is distinct from an RI.
7. What happens when an immediate reservation is created at D-30 with only a future end_date?
- A) It waits free until the event
- B) end_date specifies a future start
- C) Unused active capacity can incur charges
- D) Occupied capacity is always charged twice
Show answer
Answer: C
Include the active interval in the cost model and avoid double-counting instances occupying reservations.
8. Which statement about CloudWatch scaler metricStat and minMetricValue is correct?
- A) metricStatType is the correct field
- B) minMetricValue is always activation threshold
- C) metricStat selects the statistic; minMetricValue is NoData fallback, overridden by ignoreNullValues=false
- D) Sum fits every rate gauge
Show answer
Answer: C
Distinguish deltas/counts from rates/cumulative counters and verify period, collection window and publishing delay.
9. How should a KEDA-managed workload be handled when metrics lag?
- A) Direct Deployment scaling always persists
- B) The HPA always has the Deployment name
- C) Review the ScaledObject floor within approved bounds and GitOps ownership
- D) Create an unbounded NodePool
Show answer
Answer: C
Direct HPA/Deployment changes can be overwritten at reconciliation. Record and restore original settings.
10. Which statement about preparation and cleanup is correct?
- A) Every image supports sh/echo and keeps its cache forever
- B) Deleting a NodePool after thirty minutes is always safe
- C) Validate app/cache constraints, baseline triggers and actual node/reservation cleanup
- D) Deleting a ScaledObject always restores original replicas
Show answer
Answer: C
Startup, GC, architecture and permissions vary. Cleanup must account for owning controllers and deletion lifecycle.