Understanding Distributed Systems

Question 1

Pros of tests?

Answer 1

• working of current feature
• checking if current changes break existing features
• well written tests are great documentation
• by using TDD you force clean code on codebase

Question 2

How TDD helps with writing clean code?

Answer 2

• firstly you focus on writing requirements, so tests are more functional and code has more business methods than code methods.
• To easily write tests, classes needs to be shor, well divided and often written to interfaces.
• code coverage 100% facilitates refactor

Question 3

Parts of TDD?

Answer 3

Write failing test -> test passes -> refactor

Question 4

What is SUT?

Answer 4

System under tests (scope of the test)

Question 5

Types of tests

Answer 5

• unit
• integration: top-down, bottom-up, hybrid, contract tests
• end-to-end: user journey
• Nonfunctional tests like security tests, load tests, performance tests

Question 6

What unit tests should check?

Answer 6

Public API and state of SUT, which is a class.

Question 7

What is user-journey test?

Answer 7

It tests the whole scenario - like process of selecting an item and buying it. Exactly like user.
Pros:
- better testing most valuable functionality of app
- despite it’s big tests, quicker to write and maintain DJ test instead multiply EE tests for each step.

Question 8

What contract tests check?

Answer 8

It defines request and response of the dependency. It is used in integration tests to simulate it AND in the dependency test suite. Thanks to this, we are sure that responses are closest to real one.

Question 9

What is testing double?

Answer 9

This is prod code dependency which in tests needs to be somehow simulated.

Question 10

Type of test doubles

Answer 10

• fake
• stub
• mock

Question 11

What is fake in the testing scope?

Answer 11

lightweight implementation of a dependency, like in-memory db instead of real db.

Question 12

What is stub in the testing scope?

Answer 12

test double which just returns things - no way to check how system interacts with it during test.

Question 13

What is mock in the testing scope?

Answer 13

smarter stub - we can check how system interacts with it and still simulate return value.

Question 14

Fake vs stub vs mock?

Answer 14

Usually fake > mock > stub (especially when fakes are written by an author of real dependency)

Question 15

What is continuous integration?

Answer 15

Process of merging changes in code done by developers to master branch with additional automatic checks like “static analysis”, unit tests run on server

Question 16

What is continuous delivery?

Answer 16

If pipeline isn’t fully automated, but automation is done as much as possible in current organization, it’s Continuous delivery. Ideally a system where there is manual check on staging env, and then its approval for prod deployment is reachable for all organizations.

Question 17

What is continuous deployment?

Answer 17

Extreme version of “continuous” variations. Once PR is approved, everything is automated - new version hits the prod.
Quite rare

Question 18

What are parts of usual CI/CD?

Answer 18

1. Review
2. Build
3. Pre-prod rollout
4. Prod rollout

Question 19

How to do proper review in CI/CD?

Answer 19

Most important part.
to do:
- checking tests
- checking feature
- backward-compatibility
- logs

Question 20

What preprod rollout does?

Answer 20

• check if app correctly startup on cluster (no NullPoitner or config exception)
• smoke tests
• the same health check system like on prod

Question 21

What is canary deployment?

Answer 21

progressive rollout of deployment to subsets of users. Role of thumb are 3 batches. Next bigger than previous.

After each batch comprehensive checks.

Question 22

When rollback (app and db) are used?

Answer 22

After rollout, there should be many checks (not only health check):
- performance of service
- time of response

If something isn’t correct, the engineer should retry deployment, rollback and wait on hotfix.

Question 23

What is rollback?

Answer 23

It is bringing back previous version of app. It should be a mindless process, because decisions about rollback usually are made in stress moments on prod.

That’s why backward compatibility is so important.

Question 24

What to do when there is big change which simply cannot be backward compatible?

Answer 24

• Divide big task on smaller tasks - thanks to this, usually will be a couple of backward-compatible tasks and one incompatible.
• Leave incompatible task as the last one.
• try to make old mechanism and new mechanism works at the same time.

Example:
task: change of producer-consumer schema:
division:
1. Consumer can handle both schema (Can be rollaback)
2. producer can produce new schema messages (Can rollback)
3. Leaving old schema (Cannot rollback, because of new messages in system)

Question 25

What are types of http caching?

Answer 25

- client side caching - reverse-proxy

Question 26

What features are used by client-side http caching?

Answer 26

- cache-control (controlling of resource age) - e-tag (version-control)

Question 27

What is cache-control?

Answer 27

Is the mechanism used for client-side caching based on resource age. After first request, response has header Cache-Control. It can have many parameters, but most important are: - public - max-age - defines time after first request where resource isn’t stale Before next request for this resource, browser calculates age. If resource is stale, request goes to server.

Question 28

What is e-tag?

Answer 28

Name of header, where value is id of resources version. Next requests (if they are sent to server) contains If-None-Match header with mentions verision-id. If id didn’t change server responses with code 304 “Not modified” (no resoruce returned), and browser resets age count of this resource

Question 29

Timeline of client-side cache-control with etag

Answer 29

1. First request for resource 2. Second request: if (max-age is achived) { sending actual request for resource else { using cached resource } sending request uses etag and based od versioning results response contains 304 code or new resource.

Question 30

What are good practices in client-side caching?

Answer 30

Keeping resources immutable. Each change means new url, so whole etag mechanism isn’t needed. Better performance, because resources is cached forever (until new url will appear). It has some drawback because each new url needs to be stored. (while in etag, there is no need for keeping old version of resources.

Question 31

What is reverse-proxy?

Answer 31

It’s a server which sits between the internet and web servers. It hides the web server and mainly servers as protector of web server (load balancing).

Question 32

What are uses of reverse-proxy?

Answer 32

- authentication - rate-limiting - load balancing - caching static content - SSL termination

Question 33

What difference between forward proxy and reversed proxy?

Answer 33

A forward proxy sits between a group of client machines and the internet, hiding the clients. A reverse proxy sits between the internet and web servers, hiding the web servers.

Question 34

When use forward-proxy?

Answer 34

used for client - privacy - bypassing firewalls - content filtering.