Chapter 8 checkpoint

Chapter 8 checkpoint: services, communication and consensus

Every question from this chapter’s lessons. Your first answer to each is the one that counts; afterward you can practice again as often as you like. Passing needs 75% right on the first attempt.

0 of 17 answered0 right on the first attempt

Question 1 of 17

In Protocol Buffers, how many bytes does field number 1 holding the integer 150 take?

Your answer

Answers within 1% count, in any unit you choose.

Question 2 of 17

With HTTP/1.1 pipelining, a client sends three requests on one connection, and the first is slow. Why do the other two responses wait?

Responses must be sent in the order the requests were receivedTCP packet loss on the connectionHeader compression must finish firstHTTP/1.1 cannot send a second request before the first response

Question 3 of 17

Match each part of an Istio service mesh to its role.

Data planeControl plane (istiod)Permissive modeStrict mode

Question 4 of 17

After moving to HTTP/2, many concurrent requests share one connection. On a lossy network, all of them stall together whenever a packet is lost. Why?

Protobuf messages cannot be split across packetsHTTP/2 allows only one stream at a timeTCP head-of-line blocking: a lost packet delays every stream on the connectionHTTP/2 still sends responses in request order

Question 5 of 17

An orchestrator has 100 worker threads, one per call in flight. Its recommendations dependency slows from 20 ms to 4 seconds per call. How many calls per second to it are enough to occupy all 100 threads?

Your answercalls/s

Answers within 2% count.

Question 6 of 17

What moves an open circuit breaker to half-open, and what decides where it goes next?

The failure rate drops below the threshold while it is openA configured wait elapses; the failure rate of a few trial calls then decides open or closedAn operator must reset itThe first successful call reopens traffic fully

Question 7 of 17

A failing service normally receives 8,000 requests per second. Every call now fails, and each client retries up to 3 times. What load does the failing service see?

Your answerRPS

Answers within 2% count.

Question 8 of 17

Five layers of services each call the next, and the last one queries a database. Each layer tries every call three times when the call fails. If the database is failing, how many times its normal load can it receive?

Your answer×

Answers within 1% count.

Question 9 of 17

A slow recommendations service keeps taking all of the orchestrator’s threads, and checkout stops working. Which pattern keeps checkout running even while recommendations is slow?

More retries on recommendations callsSwitching to HTTP/2A bulkhead: a separate thread pool for calls to recommendationsOne bigger shared thread pool

Question 10 of 17

What is the core problem with two-phase commit, in Gray and Lamport’s words?

It needs hand-written compensating transactionsIt blocks if the coordinator fails, leaving participants holding their locksIt is not atomicIt takes no locks, so data can be inconsistent

Question 11 of 17

A checkout saga creates an order, reserves stock, then fails to charge the card. Put what happens in order.

Create the order as PENDING
Set the order to CANCELED
Reserve the stock
Charging the card fails
Release the stock reservation

Question 12 of 17

Match each kind of saga step to its definition.

CompensablePivotRetryable

Question 13 of 17

Other transactions can see an order while its saga is still running. Which countermeasure warns them that the order is mid-saga?

Commutative updatesNothing is needed; sagas are isolatedHolding a database row lock until the saga finishesA semantic lock, such as setting the order’s status to PENDING

Question 14 of 17

A 5-server Raft cluster splits into 2 servers (including the leader) and 3 servers. What happens to writes on each side?

The 2-server side commits nothing; the 3-server side elects a leader and keeps committingThe side with the old leader keeps committingNeither side can do anything until the partition healsBoth sides keep committing and merge later

Question 15 of 17

How many servers does a Raft cluster need to keep working with any 2 of them down?

Your answerservers

Answers within 1% count.

Question 16 of 17

Why does Raft choose each server’s election timeout at random from a range such as 150–300 ms?

So one server usually times out first and wins, making split votes rareTo spread heartbeat trafficTo stop malicious servers predicting electionsTo keep the servers’ clocks synchronized

Question 17 of 17

How does Raft make sure a newly elected leader already holds every committed entry?

The server with the longest timeout winsThe new leader commits a no-op entryThe new leader copies missing entries from followers after winningVoters refuse candidates whose logs are less up to date than their own