ziwon / terraform-fluentd-aggregator

Terraform managed Fluentd aggregator on AWS Fargate

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

terraform-fluentd-aggregator

This repo was created for the purpose of storing logs collected from multiple swarm clusters on Elastic Search.

You may utilize my fluent-bit-docker-metadata to ship logs to the aggregator from edge swarm cluster.

Fluentd Configuration

As shown in the following configuration file, the collected logs are dynamically indexed to ElasticSearch according to the log properties using record_transformer filter and elasticsearch dynamic plugin. e.g.) shinhan-aiagent-2020.08.15

<filter docker.**>
  @type record_transformer
  <record>
    service_name ${tag_parts[1]}
  </record>
</filter>

<match docker.**>
  @type elasticsearch_dynamic

  host "#{ENV['ES_HOST'] || 'elasticsearch'}"
  port "#{ENV['ES_PORT'] || '9200'}"
  user "#{ENV['ES_USER'] || 'elastic'}"
  password "#{ENV['ES_PASSWORD'] || 'changeme'}"
  scheme "#{ENV['ES_SCHEME'] || 'http'}"
  ssl_verify "#{ENV['ES_SSL_VERIFY'] || 'false'}"
  ssl_version TLSv1_2

  logstash_format true
  logstash_dateformat %Y.%m.%d
  logstash_prefix ${record['biz_client']}-${record['service_name']}

  ...

Local Swarm Cluster

On your mac, you can test the following pipeline by deploying a swarm cluster of four docker machines. Note that ES in your swarm dose run as standalone for testing.

fluent bit -> fluentd -> elastic search -> kibana

To deploy a swarm cluster:

$ make node-up

Build fluent-aggregator in docker-machine:

$ eval $(docker-machine env node-m1)
$ make docker-build

To start services:

$ make stack-start

You can access kibana with http://192.168.128.101 since the host IP keeps always the same value even if the docker machine is deployed multiple times.

To change a certification with a customized domain:

$ DOMAIN=awesome.domain make node-gen-cert

Environment Variables

We create and manage per-project isolated development environments according to the 12factor using direnv

Chanege to local env:

$ make envrc-local

Change to prod env:

$ make envrc-prod

You may need to set up the following values in production with AWS Cloud and Elastic Cloud to deploy using Terraform:

export PROJECT_NAME=ziwon
export APP_NAME=fluentd-aggregator

export AWS_ACCOUNT_ID=57xxxxxxxxxx
export AWS_REGION=ap-northeast-2
export AWS_PROFILE=default
export VPC_ID=vpc-xxxxxxxxx

export ECR_IMAGE_URL=${AWS_ACCOUNT_ID}.dkr.ecr.${AWS_REGION}.amazonaws.com/${APP_NAME}
export TAG=latest

export ES_HOST=elastic.ap-northeast-1.aws.found.io
export ES_PORT=9243
export ES_USER=elastic
export ES_PASSWORD=changeme
export ES_SCHEME=https
export ES_SSL_VERIFY=true

Makefile Targets

$ make
Usage: make [command] [args]
        Makefile:envrc-local             Change environment to local development       (e.g. make envrc-local)
        Makefile:envrc-prod              Change environment to prod development        (e.g. make envrc-prod)
        Makefile:help                    Show this help message                        (e.g. make help)
        docker.mk:docker-build           Build docker image                            (e.g. make docker-build)
        docker.mk:docker-commit          Commit current container using killed tag     (e.g. make docker-commit)
        docker.mk:docker-history         Show the history of an image                  (e.g. make docker-history)
        docker.mk:docker-push            Push an image to Amazon ECR registry          (e.g. make docker-push)
        docker.mk:docker-run             Run a command in a new container              (e.g. make docker-run)
        swarm.mk:node-add-cert           Add certification into the local machine      (e.g. make node-add-cert)
        swarm.mk:node-cleanup            Clean up the docker volume                    (e.g. make node-cleanup)
        swarm.mk:node-down               Terminate swarm nodes                         (e.g. make node-down)
        swarm.mk:node-gen-cert           Generate SSL certification                    (e.g. make node-gen-cert)
        swarm.mk:node-ip                 Show the address of node                      (e.g. make node-ip)
        swarm.mk:node-list               Show node list                                (e.g. make node-list)
        swarm.mk:node-up                 Bootstrap swarm nodes                         (e.g. make node-up)
        swarm.mk:stack-exec              Get executed the given command into container (e.g. make stack-exec sh)
        swarm.mk:stack-logs              Show logs from the service                    (e.g. make stack-logs elasticsearch)
        swarm.mk:stack-ps                List stack process                            (e.g. make stack-ps)
        swarm.mk:stack-reload            Reload the stack from swarm                   (e.g. make stack-reload)
        swarm.mk:stack-service           List stack services                           (e.g. make stack-service)
        swarm.mk:stack-start             Start the stack onto swarm                    (e.g. make stack-start)
        swarm.mk:stack-stop              Remove the stack from swarm                   (e.g. make stack-stop)
        swarm.mk:stack-viz               Visualize the swarm stack                     (e.g. make stack-viz)

Terraform

All the terraform files are in infra directory.

Requirements

Name Version
terraform ~> 0.12.0
aws ~> 2.57.0
null ~> 2.0

Providers

Name Version
aws ~> 2.57.0
null ~> 2.0

Usage

Fluentd Aggregator runs as one ECS Fargate application, is deployed in a Blue/Green deployment, and routes the traffic load to the Network Load Balancer.

module "app" {
  source = "./modules/app"

  internal_load_balancer = false

  vpc_id                = var.vpc_id
  app                   = local.app_id
  region                = var.region
  environment           = var.environment
  service_replicas      = var.service_replicas
  task_definition       = var.task_definition
  pipeline_image_tag    = var.pipeline_image_tag
  primary_domain        = var.primary_domain
  dns_name              = "${var.project}-${var.app}.${var.primary_domain}"
  certificate_arn       = var.certificate_arn
  kms_arn               = local.kms_arn
  repository_name       = var.repository_name
  enable_cpu_high_alarm = true
  enable_cpu_low_alarm  = false

  tags = merge(
    map("Last Updated", "${module.global.build_time}"),
    var.tags
  )
}

Using Makefile

We assume that all the environment variables with your AWS Cloud are properly injected to .envrc.prod

Set up your aws profile:

$ make tf-aws-profile
AWS Access Key ID [****************J5JJ]:
AWS Secret Access Key [****************VCWH]:
Default region name [ap-northest-2]:
Default output format [json]:

Intialize your terraform:

$ make tf-init

Generate tfvars according to .envrc.prod:

$ make tf-vars

Generates and show an execution plan:

$ make tf-plan

Builds or chagnes infrastructure:

$ make tf-apply

Other Targets

$ make
Usage: make [command] [args]
        Makefile:help                    Show this help message                        (e.g. make help)
        Makefile:tf-apply                Builds or changes infrastructure              (e.g. make tf-apply {project} {env})
        Makefile:tf-aws-profile          Configure aws profile                         (e.g. make tf-aws-profile)
        Makefile:tf-clean                Clean the Terraform-generated files           (e.g. make tf-clean)
        Makefile:tf-destroy              Destroy Terraform-managed infrastructure      (e.g. make tf-destroy {project} {env})
        Makefile:tf-get-tools            Install tools for development                 (e.g. make tf-get-tools)
        Makefile:tf-import               Import existing infrastructure                (e.g. make tf-import {project} {env} {resource} {arn})
        Makefile:tf-init                 Initialize s3 backend                         (e.g. make tf-init {project} {env})
        Makefile:tf-lint                 Lint the Terraform files                      (e.g. make tf-lint {project} {env})
        Makefile:tf-plan                 Generate and show an execution plan           (e.g. make tf-plan {project} {env})
        Makefile:tf-pull                 Get remote state to local                     (e.g. make tf-pull)
        Makefile:tf-refresh              Refresh the state file with infrastructure    (e.g. make tf-refresh {project} {env})
        Makefile:tf-vars                 Create a tfvars file per environment          (e.g. VPC_ID=vpc-53a49e3b make tf-vars {project} {env})
        Makefile:tf-vpc-id               Show the VPC ID with the given group          (e.g. make tf-vpc-id {group})

Notifications

To enable the notification of AWS CodePipeline to the slack, you may need to change the following value to yours:

arn:aws:chatbot::57xxxxxxxxxx:chat-configuration/slack-channel/your-slack-noti

infra/modules/app/notification.tf

resource "aws_codestarnotifications_notification_rule" "deloy" {
  detail_type = "FULL"
  event_type_ids = [
    "codepipeline-pipeline-action-execution-failed",
    "codepipeline-pipeline-action-execution-succeeded",
    "codepipeline-pipeline-pipeline-execution-failed",
    "codepipeline-pipeline-manual-approval-failed",
    "codepipeline-pipeline-manual-approval-succeeded"
  ]

  name     = "deploy-${var.app}"
  resource = aws_codepipeline.main.arn

  # The terraform not support AWS Chatbot (Beta)
  target {
    address = "arn:aws:chatbot::57xxxxxxxxxx:chat-configuration/slack-channel/your-slack-noti"
    type    = "AWSChatbotSlack"
  }

  tags = var.tags
}

Inputs

(TBD)

Outputs

(TBD)

About

Terraform managed Fluentd aggregator on AWS Fargate


Languages

Language:HCL 63.7%Language:Shell 18.1%Language:Makefile 15.6%Language:Smarty 1.8%Language:Dockerfile 0.6%Language:Ruby 0.2%