🎯 Exam Caveats & Quick-Reference Cheatsheet

Last-minute review — exam traps, decision trees, and must-memorise numbers

⚠️ The Most Dangerous Exam Traps
📋 Must-Memorise Numbers
⚡ Decision Tree — Choosing the Right Service
🃏 Flash Card — One-Line Definitions
🔑 Feature Lock-In Summary
✅ Final Exam Checklist

⚠️ The Most Dangerous Exam Traps

These are the distractors and edge cases that most frequently appear in AZ-305 scenario questions. Know these cold.

Trap 1 — “Ordering” does NOT mean Service Bus alone

❌ “Use a Service Bus queue for guaranteed ordering”
✅ Guaranteed FIFO requires Service Bus Queue + Sessions

A plain Service Bus queue does not guarantee FIFO. Sessions are the enabling feature. If the exam omits Sessions in an “ordering required” scenario, Service Bus without Sessions is a wrong answer.

Trap 2 — Dead-letter queue is NOT a Storage Queue feature

❌ “Use Storage Queues — they have built-in dead-lettering”
✅ Storage Queues have no built-in DLQ

Dead-lettering must be implemented manually in application code (e.g., move to a separate queue after N failed attempts). The Azure Functions runtime does this for you via the {queuename}-poison queue, but that is an SDK behaviour, not a platform feature.

Trap 3 — Service Bus Topics require Standard or Premium

❌ “Configure a Service Bus Basic namespace with topics”
✅ Topics and Subscriptions require Standard or Premium

Basic SKU only supports queues. If the scenario mentions pub/sub or fan-out, the answer cannot be Service Bus Basic.

Trap 4 — Event Grid is for events, not large messages

❌ “Use Event Grid to transfer 5 MB order payloads to consumers”
✅ Event Grid has a 1 MB max event size — it sends notifications, not large data payloads

For large payloads, use Claim-Check pattern: store the payload in Blob storage, send a reference event via Event Grid, consumer fetches payload directly.

Trap 5 — Event Hubs is the only service with replay

❌ “Use Service Bus to replay messages after a processing bug”
✅ Only Event Hubs supports consumer offset reset and event replay

Service Bus messages are deleted once settled. Event Hubs retains the log for 1–90 days and consumers can rewind their offset.

Trap 6 — Geo-DR replicates metadata, not messages

❌ “Service Bus Geo-DR ensures no messages are lost during regional failure”
✅ Service Bus and Event Hubs Geo-DR replicate metadata only (entity definitions)

Messages in flight on the primary are not replicated. For zero message loss during regional failover, implement application-level active replication (dual-write to two namespaces).

Trap 7 — Partition count is immutable in Event Hubs

❌ “After launch we can increase partitions if throughput grows”
✅ Event Hubs partition count is fixed at creation

Plan partitions upfront based on expected parallelism. Increasing partitions requires creating a new Event Hub.

Trap 8 — SLA 99.99% without Premium SKU is only Event Grid

Service Bus: 99.99% requires Premium + Availability Zones
Event Hubs: 99.99% requires Dedicated cluster
Storage Queues: 99.99% read-only via RA-GRS (not write)
Event Grid: 99.99% on all tiers — no Premium required

Trap 9 — Private Endpoints for Storage Queues don’t require premium

❌ “Private endpoints for Storage Queues require a Premium storage tier”
✅ Private endpoints on Azure Storage (including queues) are available on all storage tiers

Service Bus, Event Hubs, and Event Grid namespace require a Premium/Dedicated tier for private endpoints. Storage is different.

Trap 10 — Event Hubs Capture is NOT available on Basic

❌ “Use Event Hubs Basic with Capture to archive the stream”
✅ Capture requires Standard, Premium, or Dedicated

Basic Event Hubs supports only 1 consumer group and 1-day retention with no Capture feature.

Trap 11 — Transactions in Service Bus are namespace-scoped

❌ “Use Service Bus transactions to atomically write to two namespaces”
✅ Transactions work only within a single namespace

Cross-namespace atomic operations require saga patterns or compensating transactions at the application level.

Trap 12 — Storage Queue message TTL can be set to infinite (`-1`)

A TTL of -1 on a Storage Queue message means it never expires. Most candidates assume 7 days is the hard maximum, but infinite TTL is valid.

📋 Must-Memorise Numbers

Message & Event Size Limits

Service	Max Size	Memory Hook
Storage Queues	64 KB	“Sixty-four is tiny”
Event Grid	1 MB	“Grid = grid paper = small”
Event Hubs	1 MB	“Same as Grid”
Service Bus Standard	256 KB	“Four times Storage Queue”
Service Bus Premium	100 MB	“One hundred — largest”

Retention Periods

Service	Retention
Storage Queues	≤ 7 days (or infinite with `-1`)
Service Bus	Configurable; default 14 days
Event Grid	24-hour retry window
Event Hubs Standard	1–7 days
Event Hubs Premium/Dedicated	1–90 days

SLA Values

Service / Tier	SLA
Service Bus Basic / Standard	99.9%
Service Bus Premium + AZ	99.99%
Storage Queues LRS/GRS	99.9%
Storage Queues RA-GRS (read)	99.99%
Event Grid (all tiers)	99.99%
Event Hubs Basic / Standard	99.9%
Event Hubs Premium	99.95%
Event Hubs Dedicated	99.99%

Consumer Limits

Service	Consumer Groups / Subscriptions
Service Bus	Up to 2,000 subscriptions per topic
Event Hubs Basic	1 consumer group
Event Hubs Standard	20 consumer groups
Event Hubs Premium	100 consumer groups

Partition Limits

Service	Max Partitions
Event Hubs Standard	32
Event Hubs Premium	100
Event Hubs Dedicated	2,000

⚡ Decision Tree — Choosing the Right Service

flowchart TD
    Start(["Start: What is the\nprimary requirement?"])

    Start --> A{"Stream of events\nat massive scale?"}
    A -->|Yes| B{"Need Kafka compat\nor event replay?"}
    B -->|Yes| EH["🟩 Event Hubs"]
    B -->|No| EH2["🟩 Event Hubs\n(still best fit)"]

    A -->|No| C{"React to Azure\nservice state changes?"}
    C -->|Yes| EG["🏁 Event Grid\n(System Topic)"]

    C -->|No| D{"Custom notifications /\nserverless triggers?"}
    D -->|Yes| EG2["🏁 Event Grid\n(Custom Topic)"]

    D -->|No| E{"Need ordering, DLQ,\ntransactions, or fan-out?"}
    E -->|Yes| SB["📨 Service Bus"]

    E -->|No| F{"Very cheap,\n> 80 GB, simple?"}
    F -->|Yes| SQ["📦 Storage Queues"]
    F -->|No| SB2["📨 Service Bus\n(still a fine choice)"]

🃏 Flash Card — One-Line Definitions

Service	One-Line Definition
Service Bus	Enterprise message broker: guaranteed delivery, ordering (Sessions), transactions, pub/sub
Storage Queues	Cheap, simple, massive-scale queue built on Storage; no DLQ, no ordering, no transactions
Event Grid	Serverless event router: react to state changes across Azure services; 99.99% SLA by default
Event Hubs	High-throughput event stream: ingest millions/sec, replay from offset, Kafka-compatible

🔑 Feature Lock-In Summary

If the exam says…	The answer is…
Guaranteed FIFO ordering	Service Bus + Sessions
Dead-letter failed messages	Service Bus
Replay events after failure	Event Hubs
Millions of events per second	Event Hubs
Archive stream to Data Lake	Event Hubs Capture
Kafka without code changes	Event Hubs
React to BlobCreated / VM deleted	Event Grid
Fan-out to different backends	Event Grid or Service Bus Topics
Queue > 80 GB	Storage Queues
Lowest cost queue, no special features	Storage Queues
Private endpoint, no premium cost	Storage Queues (Storage account)
99.99% SLA, no premium fee	Event Grid
99.99% SLA on Service Bus	Premium + Availability Zones
99.99% SLA on Event Hubs	Dedicated
Atomic write across queues	Service Bus Transactions (same namespace)
Large messages > 256 KB	Service Bus Premium
> 90-day event retention	❌ None natively — archive to Blob via Capture

✅ Final Exam Checklist

Before sitting the exam, verify you can answer these without hesitation:

← 05 - Feature Comparison

Back to Home →