Cloudwatch: Monitoring your AWS services with Metrics and Alarms
1 like2,444 views
O AWS CloudWatch é um serviço de monitoramento que coleta métricas de diversas instâncias e serviços da AWS, permitindo a visualização e tomada de decisões informadas sobre recursos. Ele oferece métricas internas e a possibilidade de definir métricas personalizadas, alarmes e ações automáticas com base em critérios específicos. O CloudWatch também suporta a captura de logs de aplicativos e interage com serviços de terceiros para aprimorar a funcionalidade de monitoramento.
Cloudwatch: Monitoring your AWS services with Metrics and Alarms
- 1. AWS Cloudwatch Felipe Almeida Rio Cloud Computing Meetup October / 2016 Monitoring your AWS Services with Metrics and Alarms
- 2. Structure ● Intro ● Usage ● Metrics - Builtin ● Metrics - Custom ● Alarms ● Events ● Logs ● Third-party services ● Keep in mind 2
- 3. Intro ● CloudWatch is the AWS metric collection service ● Metrics are individual measurements of some quantity of interest, at a given point in time: Examples: ● At 23:45:29 GMT, on 23 February 2016, CPU usage for instance XPTO was 56% 3
- 4. Intro ● Nearly all AWS services can be monitored using CloudWatch ● Metrics provide visibility into your applications ● They enable you to make informed decisions such as ○ Whether to downscale a machine that’s not being used too much (save $$) ○ Whether to scale up a machine that’s close to full capacity (avoid failures) ○ Decide what caused an application to fail (debug) 4
- 5. Usage ● You can use CloudWatch in multiple ways: ○ Using the AWS Console ○ Via other services that have monitoring functionality ○ Using the CloudWatch API through the AWS SDK 5
- 6. Usage - Via other services ● Via other services ○ Nearly all services have a monitoring or metrics tab 6EMR
- 7. Usage - Via other services 7Elastic Beanstalk
- 8. Usage - Via other services 8DynamoDB Table
- 9. Usage - Via other services ● When you open the monitoring tab for a service, what you’re looking at are actually CloudWatch Metrics: 9 Monitoring Tab for an Elastic Beanstalk Environment uses CloudWatch Metrics
- 10. Usage - Console ● Via cloudwatch itself (on the console) 10
- 11. Usage - Console ● You can build a Dashboard with metrics you choose 11
- 12. Usage - API ● CloudWatch (like most AWS Services) can also be used programatically ○ I.e. it has an API that you can use via the AWS SDK (library for Java, .NET, Python, etc) 12
- 13. Metrics - Builtin ● Services usually have a set of builtin metrics ○ These metrics are collected by default 13
- 14. Metrics - Builtin ● For example, the following metrics (not exhaustive list) are collected by default for EC2 Instances: ○ CPUUtilization (percent) ○ DiskReadBytes (bytes) ○ DiskWriteBytes (bytes) ○ NetworkIn (bytes) ○ NetworkOut (bytes) 14
- 15. Metrics - Custom ● You can also use CloudWatch for your own metrics. ● You can then do anything you can for regular (builtin) metrics, such as: ○ View them on the AWS console ○ Create Dashboards for them ○ Set up alerts 15
- 16. Metrics - Custom ● For some types of custom metrics, there are community-provided scripts to help you publish them to CloudWatch, such as: ○ Memory utilization for EC2 Instances ○ Swap Utilization for EC2 Instances ○ Disk Space Usage for EC2 Instances ● These metrics are controlled by the O.S. and therefore cannot be accessed automatically by AWS. 16
- 17. Alarms ● You can also create alarms for any individual metric on CloudWatch, which are triggered if the criteria you have defined are met ● For example, you can have AWS send you an e-mail if FreeStorageSpace metric for a ElasticSearch cluster you have becomes lower than 10GB. 17
- 18. Alarms ● In addition to notifying you when something happens, you can also configure AutoScaling Actions to be taken if some criteria are met: ○ For example, if the CPUUtilization metric reaches 90% for all machines in a AutoScaling Group, add extra instances to that environment 18
- 19. Events ● CloudWatch Events are a new addition in CloudWatch ● They enable you to execute custom actions in response to State Changes such as ○ An EC2 instance has been started ○ An EC2 instance has been terminated ● Custom actions include running a Lambda Function, publishing the event to Kinesis, etc. 19
- 20. Logs ● CloudWatch Logs refers to Agents you can install on your instances to have them send application logs to CloudWatch ● You can filter your logs and set up Alarms when certain criteria are met, such as: ○ There have been more than 10 Errors in Apache in the last minute ○ There have been 5 log messages matching “IllegalArgumentException” in the last hour 20
- 21. Related Services (Third-Party) ● There are tons of third-party services that build on top of or otherwise leverage CloudWatch metrics, such as: ○ SignalFX (extends and sends custom metrics to CloudWatch) ○ AppDynamics (sends all sorts of low-level and application-level metrics to CloudWatch and displays them in an unified manner; also provides event correlation) ○ NewRelic (sends data to CLoudWatch) ○ Grafana (uses CloudWatch as a Data source) ○ Logstash (you can use CloudWatch as an output) ○ Nagios (consumes CloudWatch metrics) 21
- 22. Keep in mind ● Metrics last 2 weeks by default ● Metrics are per-region ● Some services (e.g. Elastic BeanStalk) have extra metrics that need to be explicitly enabled before they can be used. ● CloudWatch can also be used to monitor AWS Costs (Billing), even though it’s not a service per se 22
- 23. Keep in mind ● In order to publish custom Amazon CloudWatch metrics, the instances in your environment need permission to use CloudWatch (see your instance profile for more information). ● CloudWatch is not free so you may want to track only relevant metrics ○ In addition, if you use CloudWatch via the AWS SDK, try to collect high-level metrics to keep cost down ■ E.g. track BeanStalk metrics rather than metrics for individual instances 23
- 24. Links ● Monitoring Scripts for Linux EC2 Instances ● All Builtin Metrics for all supported AWS services ● AWS CloudWatch Pricing 24