¿Es realmente imposible prescindir de kafokas y conejos cuando recibes 10,000 eventos por segundo?

Una vez organicé un seminario web sobre cómo recibir 10,000 eventos por segundo. Mostró esta imagen, la audiencia vio una capa de color lila, y comenzaba: "Chicos, ¿por qué necesitamos todos estos kafkas y conejos, realmente no podemos prescindir de ellos?" Respondimos: "¡Por qué, por qué, pasar la seguridad social!" 

Muy gracioso, pero déjame explicarte de todos modos.

---

.

,

, , . Kafka, Rabbit . , . 

: 10 , — , — 50 . , . 2 , 10 . , . — .

— -, . 

, . . : , . http- — 5, 10 — , . 

. « », : « , , ». «» , .

— — , . 

, , . , , , e-mail . , , , .

— — -, .

, . ? , . , . - , — (, ). , .

---

, .

-. . .

resource "yandex_compute_instance_group" "events-api-ig" {
  name               = "events-api-ig"
  service_account_id = yandex_iam_service_account.instances.id
      
      

        
        
        
      

    
        
        
        
      
      

        
        
        
      

    
    

. CPU, , ..

instance_template {
    platform_id = "standard-v2"
    resources {
      memory = 2
      cores  = 2
    }
    boot_disk {
      mode = "READ_WRITE"
      initialize_params {
        image_id = data.yandex_compute_image.container-optimized-image.id
        size = 10
      }
      
      

        
        
        
      

    
        
        
        
      
      

        
        
        
      

    
    

, .

}
    network_interface {
      network_id = yandex_vpc_network.internal.id
      subnet_ids = [yandex_vpc_subnet.internal-a.id, yandex_vpc_subnet.internal-b.id, yandex_vpc_subnet.internal-c.id]
      nat = true
    }
      
      

        
        
        
      

    
        
        
        
      
      

        
        
        
      

    
    

— scale_policy.

fixed scale A, B, C.

scale_policy {
    fixed_scale {
      size = 3
    }
  }

  allocation_policy {
    zones = ["ru-central1-a", "ru-central1-b", "ru-central1-c"]
  }
      
      

        
        
        
      

    
        
        
        
      
      

        
        
        
      

    
    

auto_scale — .

scale_policy {
auto_scale {
    initial_size = 3
    measurment_duration = 60
    cpu_utilization_target = 60
    min_zone_size = 1
    max_size = 6
    warmup_duration = 60
    stabilization_duration = 180
}
      
      

        
        
        
      

    
        
        
        
      
      

        
        
        
      

    
    

, , — cpu utilization target. , . .

, .

- load-. , 84.201.147.84 80, - — 8080. 

, Yandex.Tank . 20 5 . 

---

, .

(A, B C), , . .

, . , .

. , , ( ) 23 , 12,8 . - . , .

— .

, , CPU . : overload.yandex.net/256194.

---

golang. .

package main

import (
    "encoding/json"
    "flag"
    "io"
    "io/ioutil"
    "log"
    "net/http"
    "strings"

)
      
      

        
        
        
      

    
        
        
        
      
      

        
        
        
      

    
    

github.com/Shopify/sarama — .

github.com/prometheus/client_golang/prometheus, API Metrics.

github.com/streadway/amqp rabbitmq.

, .

var (
    // Config options
    addr     = flag.String("addr", ":8080", "TCP address to listen to")
    kafka    = flag.String("kafka", "127.0.0.1:9092", "Kafka endpoints")
    enableKafka    = flag.Bool("enable-kafka", false, "Enable Kafka or not")
amqp    = flag.String("amqp", "amqp://guest:guest@127.0.0.1:5672/", "AMQP URI")
enableAmqp    = flag.Bool("enable-amqp", false, "Enable AMQP or not")
sqsUri    = flag.String("sqs-uri", "", "SQS URI")
sqsId    = flag.String("sqs-id", "", SQS Access id")
sqsSecret    = flag.String("sqs-secret", "", "SQS Secret key")
enableSqs    = flag.Bool("enable-sqs", false, "Enable SQS or not")

    

    // Declaring prometheus metrics
    apiDurations = prometheus.NewSummary(
        prometheus.SummaryOpts{
            Name:       "api_durations_seconds",
            Help:       "API duration seconds",
            Objectives: map[float64]float64{0.5: 0.05, 0.9: 0.01, 0.99: 0.001},
        },
    )
      
      

        
        
        
      

    
        
        
        
      
      

        
        
        
      

    
    

— ().

— .

.

— .

— amqp .

— sqs ..

. prometheus .

main , Load.

sqs, ..

http :

/status okey, load-, .

/post/kafka, . /post/amqp /post/sqs.

— , . , .

- . , SSD ( , ), zookeeper. 200 ! — « », . 120 .

, — CPU, , - . , .

: topic, , . topic 50 partitions. .

, topic load 3 . Partition 1 Kafka 1, — 2, — 3. . , , , . .

50 , 50 1 — 50 . .

zookeeper. 3 . , , 2 . 2 2 . , : «, », 2 .

Terraform

terraform-, kafka.tf .

3 : resource “yandex compute instance” “zookeeper count = 3”.

“zookeeper_deploy”, . , , . . ansible .

, , .

RabbitMQ 

, , . , ! , .

— exchanges , delayed messages, deadletter . . , . — . , , . — .

. , , .

— RabbitMQ. — , , .. — kafka. .

: .