How I made a powerful cache system using Go

Sylvain Combraque

Creator of Souin HTTP cache

Træfik helper/ambassador

Cache-handler maintainer

Open-source contributor

@darkweak

@darkweak_dev

Some context

Traefik

Emile Vauge

Written in GO

Easy to use

French Quality

Where is the cache ?

github.com/darkweak/Souin

Leave a ⭐️ on Github

Varnish vs Souin

[disclaimer trolling]

Demo

How does Go can help us to achieve performance

The power of

go func() {
	_, _ = new(http.Client).Do(req)
}()

With channels

go func(rs http.ResponseWriter, rq *http.Request) {
	if rc != nil && <-coalesceable {
		rc.Temporize(req, rs, nextMiddleware)
	} else {
		errorBackendCh <- nextMiddleware(rs, rq)
		return
	}
	errorBackendCh <- nil
}(res, req)

select {
case <-req.Context().Done():
	switch req.Context().Err() {
	case ctx.DeadlineExceeded:
		cw := res.(*CustomWriter)
		rfc.MissCache(cw.Header().Set, req, "DEADLINE-EXCEEDED")
		cw.WriteHeader(http.StatusGatewayTimeout)
		_, _ = cw.Rw.Write(serverTimeoutMessage)
		return ctx.DeadlineExceeded
	case ctx.Canceled:
		return nil
	default:
		return nil
	}
case v := <-errorBackendCh:
	if v == nil {
		_, _ = res.(souinWriterInterface).Send()
	}
	return v
}

Interfaces to the rescue

type Badger struct {}

func (provider *Badger) ListKeys() []string {}
func (provider *Badger) Prefix(key string, req *http.Request) *http.Response {}
func (provider *Badger) Get(key string) []byte {}
func (provider *Badger) Set(key string, value []byte) error {}
func (provider *Badger) Delete(key string) {}
func (provider *Badger) DeleteMany(key string) {}
func (provider *Badger) Init() error {}
func (provider *Badger) Name() string {}
func (provider *Badger) Reset() error {}

type Olric struct {}

func (provider *Olric) ListKeys() []string {}
func (provider *Olric) Prefix(key string, req *http.Request) *http.Response {}
func (provider *Olric) Get(key string) []byte {}
func (provider *Olric) Set(key string, value []byte) error {}
func (provider *Olric) Delete(key string) {}
func (provider *Olric) DeleteMany(key string) {}
func (provider *Olric) Init() error {}
func (provider *Olric) Name() string {}
func (provider *Olric) Reset() error {}

type Storer interface {
	ListKeys() []string
	Prefix(key string, req *http.Request) *http.Response
	Get(key string) []byte
	Set(key string, value []byte) error
	Delete(key string)
	DeleteMany(key string)
	Init() error
	Name() string
	Reset() error
}

func NewStorages(c AbstractConfigurationInterface) map[string]Storer {
	providers := make(map[string]AbstractProviderInterface)
	olric, _ := OlricConnectionFactory(c)
	providers["olric"] = olric
	badger, _ := BadgerConnectionFactory(configuration)
	providers["badger"] = badger

	for _, p := range providers {
		_ = p.Init()
	}
	return providers
}

Context

ctx := context.Background()
valCtx := ctx.Value("MY_CONTEXT_KEY")
if valCtx == nil {
	return "not found"
}
val, ok := valCtx.(string)
// not found

ctx := context.Background()
ctx = context.WithValue(ctx, "MY_CONTEXT_KEY", "a string")

ctx := context.Background()
valCtx := ctx.Value("MY_CONTEXT_KEY")
if valCtx == nil {
	return "not found"
}
val, ok := valCtx.(string)
// val: "a string"
// ok: true

const (
	CacheName           ctxKey = "souin_ctx.CACHE_NAME"
	RequestCacheControl ctxKey = "souin_ctx.REQUEST_CACHE_CONTROL"
)

func (cc *cacheContext) SetContext(req *http.Request) *http.Request {
	co, _ := cacheobject.ParseRequestCacheControl(req.Header.Get("Cache-Control"))
	return req.WithContext(
		context.WithValue(
			context.WithValue(
				req.Context(),
				CacheName,
				cc.cacheName
			),
			RequestCacheControl,
			co
		)
	)
}

Cool features in Souin

default_cache:
  ttl: 10s
reverse_proxy_url: 'http://reverse-proxy'

YAML configuration

func GetConfiguration() *Configuration {
	data := readFile("./configuration/configuration.yml")
	var config Configuration
	if err := yaml.Unmarshal(data, config); err != nil {
		log.Fatal(err)
	}
	return &config
}

Configuration as plugin

No configuration required

# souin/docker-compose.yml
version: '3.4'

x-networks: &networks
  networks:
    - your_network

services:
  souin:
    image: darkweak/souin:latest
    ports:
      - 80:80
      - 443:443
    environment:
      GOPATH: /app
    volumes:
      - ./configuration.yml:/configuration/configuration.yml
    <<: *networks

networks:
  your_network:
    external: true

# anywhere/docker-compose.yml
version: '3.4'

x-network:
  &network
  networks:
    - your_network

services:
  traefik:
    image: traefik:latest
    command: --providers.docker
    ports:
      - 81:80
      - 444:443
      - 8080:8080
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock
    <<: *network

networks:
  your_network:
    external: true

Container first

Middlewares

APP

Headers middleware

Middlewares

func (s *DummyCaddyPlugin) ServeHTTP(rw http.ResponseWriter, r *http.Request, next caddyhttp.Handler) error {
	req := s.Retriever.GetContext().SetBaseContext(r)
	
	# do some stuff....

	customWriter := newInternalCustomWriter(rw, bufPool)
	getterCtx := getterContext{customWriter, req, next}
	ctx := context.WithValue(req.Context(), getterContextCtxKey, getterCtx)
	req = req.WithContext(ctx)
	req.Header.Set("Date", time.Now().UTC().Format(time.RFC1123))
	next.ServeHTTP(customWriter, r)

	return nil
}

Request coalescing system support

var result singleflight.Result
select {
case <-timeout:
	http.Error(rw, "Gateway Timeout", http.StatusGatewayTimeout)
	return
case result = <-ch:
}

if result.Err != nil {
	http.Error(rw, result.Err.Error(), http.StatusInternalServerError)
	return
}

API management

Configuration

api:
  basepath: /souin-api
  prometheus:
    basepath: /anything-for-prometheus-metrics
  souin:
    basepath: /anything-for-souin

Cache management API

Default base path: /souin

    basePathAPIS := c.GetAPI().BasePath
    if basePathAPIS == "" {
        basePathAPIS = "/souin-api"
    }
    for _, endpoint := range api.Initialize(provider, c) {
        if endpoint.IsEnabled() {
            http.HandleFunc(
              fmt.Sprintf(
                "%s%s", 
                basePathAPIS, 
                endpoint.GetBasePath()
              ), 
              endpoint.HandleRequest
            )
			http.HandleFunc(
              fmt.Sprintf(
                "%s%s/", 
                basePathAPIS, 
                endpoint.GetBasePath()
              ), 
              endpoint.HandleRequest
            )
        }
    }

How to code this?

Choose your storage

nutsdb

badger

redis

olric

etcd

Surrogate-Keys

First group

Second group

Third group

key 1

key 3

key 564

key 90

key ABC

key 3

key 789

key 1

key 789

Cache-Groups

CDN tag invalidation

Compatible with many softwares

Used in production by ETH.LIMO

It was funny parts to dev

Respect the RFC

such a pain

When you think it's simple, RFC tells you it is not

1. Introduction

HTTP is typically used for distributed information systems, where performance can be improved by the use of response caches. This document defines aspects of HTTP/1.1 related to caching and reusing response messages. An HTTP cache is a local store of response messages and the subsystem that controls storage, retrieval, and deletion of messages in it. A cache stores cacheable responses in order to reduce the response time and network bandwidth consumption on future, equivalent requests. Any client or server MAY employ a cache, though a cache cannot be used by a server that is acting as a tunnel. A shared cache is a cache that stores responses to be reused by more than one user; shared caches are usually (but not always) deployed as a part of an intermediary. A private cache, in contrast, is dedicated to a single user; often, they are deployed as a component of a user agent. The goal of caching in HTTP/1.1 is to significantly improve performance by reusing a prior response message to satisfy a current request. A stored response is considered "fresh", as defined in Section 4.2, if the response can be reused without "validation" (checking with the origin server to see if the cached response remains valid for this request). A fresh response can therefore reduce both latency and network overhead each time it is reused. When a cached response is not fresh, it might still be reusable if it can be freshened by validation (Section 4.3) or if the origin is unavailable (Section 4.2.4).

1.1. Conformance and Error Handling

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119]. Conformance criteria and considerations regarding error handling are defined in Section 2.5 of [RFC7230].

1.2. Syntax Notation

This specification uses the Augmented Backus-Naur Form (ABNF) notation of [RFC5234] with a list extension, defined in Section 7 of [RFC7230], that allows for compact definition of comma-separated lists using a '#' operator (similar to how the '*' operator indicates Fielding, et al. Standards Track [Page 4]

RFC-7234

if isResponseTransparent {
	return transparent
}
if isResponseStale {
	return stale
}
if isResponseFresh {
	return fresh
}

expiresHeader := respHeaders.Get("Expires")
if expiresHeader != "" {
	expires, err := time.Parse(time.RFC1123, expiresHeader)
	if err != nil {
		lifetime = zeroDuration
	} else {
		lifetime = expires.Sub(date)
	}
}

RFC-9211

aka. Cache-Status RFC

Cache-Status

Souin; stored; fwd=uri-miss

Souin; hit; ttl=1234; key=GET-domain.com-/uri-another

Souin; fwd=uri-miss; key=GET-domain.com-/uri-another,
Caddy; hit; ttl=432; key=abcdef123

Cache-Status

res
    .header
    .Set(
        "Cache-Status", 
        fmt.Sprintf(
            "%s; fwd=uri-miss; key=%s; detail=UNCACHEABLE-STATUS-CODE",
            rq.Context().Value(context.CacheName),
            rfc.GetCacheKeyFromCtx(rq.Context()),
        )
    )

RFC-9213

aka. Targeted Cache-Control

CDN-Cache-Control

{name}-Cache-Control

Age: 1800
Cache-Control: max-age=600

Age: 1800
Cache-Control: max-age=600
CDN-Cache-Control: max-age=3600

Age: 1800
Cache-Control: max-age=600
CDN-Cache-Control: no-cache
Souin-Cache-Control: public; max-age=3600
LowLevelApp-Cache-Control: public; max-age=7200

https://www.youtube.com/watch?v=AaQxCzoT2X8

New version deployment

The workflow

Test

Build

Release

Testing

golangci-lint

go unit test

build plugins

E2E plugin tests

go unit test with services

jobs:
  lint-validation:
    name: Validate Go code linting
    runs-on: ubuntu-latest
    steps:
      - name: Checkout code
        uses: actions/checkout@v4
      - name: Install Go
        uses: actions/setup-go@v3
        with:
          go-version: ${{ env.GO_VERSION }}
      - name: golangci-lint
        uses: golangci/golangci-lint-action@v3
        with:
          args: --timeout=240s

golangci-lint

jobs:
  unit-test-golang:
    needs: lint-validation
    name: Unit tests
    runs-on: ubuntu-latest
    steps:
      - name: Checkout code
        uses: actions/checkout@v4
      - name: Install Go
        uses: actions/setup-go@v3
        with:
          go-version: ${{ env.GO_VERSION }}
      - name: Run unit static tests
        run: go test -v -race $(go list ./... | grep -v pkg/storage)

unit tests

jobs:
  unit-test-golang-with-services:
    needs: lint-validation
    name: Unit tests with external services
    runs-on: ubuntu-latest
    steps:
      - name: Checkout code
        uses: actions/checkout@v4
      - name: Install Go
        uses: actions/setup-go@v3
        with:
          go-version: ${{ env.GO_VERSION }}
      - name: Build and run the docker stack
        run: |
          docker network create your_network || true
          docker-compose -f docker-compose.yml.test up \
          -d --build --force-recreate --remove-orphans
      - name: Run pkg storage tests
        run: docker-compose -f docker-compose.yml.test exec -T souin \
             go test -v -race ./pkg/storage

unit tests with services

jobs:
  build-roadrunner-validator:
    name: Check that Souin build as middleware
    uses: ./.github/workflows/plugin_template.yml
    secrets: inherit
    with:
      CAPITALIZED_NAME: Roadrunner
      LOWER_NAME: roadrunner
      GO_VERSION: '1.21'

build plugins

jobs:
  plugin-test:
    name: Check that Souin build as ${{ inputs.CAPITALIZED_NAME }} middleware
    runs-on: ubuntu-latest
    env: 
      GO_VERSION: ${{ inputs.GO_VERSION }}
    # ...
 
    steps:
      - name: Check if the configuration is loaded to define if Souin is loaded too
        uses: nick-invision/assert-action@v1
        with:
          expected: 'Souin configuration is now loaded.'
          actual: ${{ env.MIDDLEWARE_RESULT }}
          comparison: contains
      - name: Run ${{ inputs.CAPITALIZED_NAME }} E2E tests
        uses: anthonyvscode/newman-action@v1
        with:
          collection: "docs/e2e/Souin E2E.postman_collection.json"
          folder: ${{ inputs.CAPITALIZED_NAME }}
          reporters: cli
          delayRequest: 5000

E2E plugin tests

jobs:
  generate-souin-docker:
    name: Generate souin docker
    runs-on: ubuntu-latest
    steps:
      - name: Build & push Docker image containing only binary
        id: docker_build
        uses: docker/build-push-action@v4
        with:
          push: true
          file: ./Dockerfile-prod
          platforms: linux/arm64, # others...
          build-args: |
            "GO_VERSION=${{ env.GO_VERSION }}"
          tags: |
            darkweak/souin:latest
            darkweak/souin:${{ env.RELEASE_VERSION }}

Build and Release

jobs:
  generate-artifacts:
    name: Deploy to goreleaser
    runs-on: ubuntu-latest
    steps:
      - name: Run GoReleaser
        uses: goreleaser/goreleaser-action@v3
        with:
          version: latest
          args: release --clean
          workdir: ./plugins/souin
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          GO_VERSION: ${{ secrets.GITHUB_TOKEN }}

Release

jobs:
  generate-tyk-versions:
    name: Generate Tyk plugin binaries
    runs-on: ubuntu-latest
    env:
      LATEST_VERSION: v5.0
      PREVIOUS_VERSION: v4.3
      SECOND_TO_LAST_VERSION: v4.2
    steps:
      - name: Generate Tyk amd64 artifacts
        run: cd plugins/tyk && make vendor && docker compose up
      - name: Upload Tyk amd64 artifacts
        uses: actions/upload-artifact@v3
        with: 
          path: plugins/tyk/*.so

Release

My reaction when all jobs are green and the new version is available

The deployment is done

RESULTS
================================================================================
---- Global Information --------------------------------------------------------
> request count                                     101000 (OK=101000 KO=0     )
> min response time                                      0 (OK=0      KO=-     )
> max response time                                     56 (OK=56     KO=-     )
> mean response time                                     8 (OK=8      KO=-     )
> std deviation                                          4 (OK=4      KO=-     )
> response time 50th percentile                          8 (OK=8      KO=-     )
> response time 75th percentile                         10 (OK=10     KO=-     )
> response time 95th percentile                         15 (OK=15     KO=-     )
> response time 99th percentile                         21 (OK=21     KO=-     )
> mean requests/sec                                3884.615 (OK=3884.615 KO=-  )
---- Response Time Distribution ------------------------------------------------
> t < 800 ms                                        101000 (100%)
> 800 ms < t < 1200 ms                                   0 (  0%)
> t > 1200 ms                                            0 (  0%)
> failed                                                 0 (  0%)
================================================================================

Roadmap

👷‍♂️ Stream every responses
🚀 Better performances
🛠 Managed version? (ask for the beta access)

Any idea ?

Open an issue at https://github.com/darkweak/souin/issues/new

Incoming conference 📣

https://www.billetweb.fr/devfest-mons-2023

Thank you for your attention

How I made a powerful cache system using Go

By darkweak

How I made a powerful cache system using Go

How I made a powerful cache system using Go

Sylvain Combraque

Some context

Traefik

Where is the cache ?

github.com/darkweak/Souin

Varnish vs Souin

Demo

The power of

With channels

Interfaces to the rescue

Context

Cool features in Souin

YAML configuration

Configuration as plugin

No configuration required

Container first

Middlewares

Middlewares

Middlewares

Request coalescing system support

API management

Configuration

Cache management API

How to code this?

Choose your storage

Surrogate-Keys

Cache-Groups

CDN tag invalidation

CDN tag invalidation

Compatible with many softwares

Used in production by ETH.LIMO

It was funny parts to dev

Respect the RFC

RFC-9211

Cache-Status

Cache-Status

RFC-9213

CDN-Cache-Control

{name}-Cache-Control

New version deployment

The workflow

Testing

golangci-lint

unit tests

unit tests with services

build plugins

E2E plugin tests

Build and Release

Release

Release

The deployment is done

Roadmap

Incoming conference 📣

Thank you for your attention

How I made a powerful cache system using Go

More from darkweak