Celery AutoScale

Celery Autoscale is a service powered by AWS Lambda that runs every minute and collect total pending task from broker (Redis or RabbitMQ) put the metric(pendingTask) on Cloudwatch. These celery inspect support redis and rabbitMQ broker.

This metric can be used for following purpose:

• To trigger scale up and scale down activity for celery worker autoscaling group.
• To trigger alarm mails if pendingtask count goes too high.

Supported broker

• Redis
  • It connects to redis and poll queue stats. So, you have to allow aws lambda to make redis connection.
• RabbitMQ
  • It uses rabbitmq_management plugin http api to query stats. So, You have to enable rabbitmq_management plugin with this rabbitmq help.

Update celery info in serverless

  CeleryCloudWatch:
    handler: handler.celery_task_status_watch_handler
    description: Collects celery peding task count to celery cloudwatch
    events:
      - schedule:
          rate: rate(1 minute)
          enabled: true
          input:
            service: test
            broker: redis://localhost:6379/0
            queues: ["celery"]

CeleryCloudWatch Function need 3 inputs

Installation

• serverless

• python3.6

  We are using serverless for lambda setup on AWS. To deploy run this command

  ./deploy.sh <stage> [aws profile] stage can be dev or staging

Additional

You can use serverless-plugin-aws-alerts plugin to create cloud watch alerts on pendingTask metrics.

Let's say I want an email alert whenever I see more than 1000 pending task in 5 minutes:

  # serverless.yml

    plugins:
      - serverless-plugin-aws-alerts
    
    custom:
      alerts:
        stages:
          - producton
        topics:
          alarm: 
            topic: ${self:service}-${opt:stage}-alerts-alarm
            notifications:
              - protocol: email
                endpoint: [email protected]
        definitions:
          tooManyTasksAlarm:
            description: 'Pending Task Overflow'
            namespace: 'Celery/Queue'
            metric: pendingTask
            threshold: 1000
            statistic: Sum
            period: 600
            evaluationPeriods: 1
            comparisonOperator: GreaterThanOrEqualToThreshold
        alarms:
          - tooManyTasksAlarm

License

GNU General Public License v3.0 (see LICENSE)