- Public Service
- 3 Products into one
- Metrics: Data related to AWS Products, other products or on-premises servers
- Some metrics are gathered natively, anything can be seen in product UI can be gathered natively by cloudwatch
- Logs: Any logs natively ingested (taken) by cloudwatch, e.g. Lambda logs, API logs will be logged in Cloudwatch.
- Events: Event logging of Cloudwatch Event changes, or scheduling.
- CloudWatch agent: For metrics gathering of service outside AWS, or generally not exposed to AWS (e.g. Process running in EC2 instance)
- Container for logs.
- Every AWS log namespace has
AWS/namespace scheme, (AWS/lambda/,AWS/ApiGateway etc..). And are reserved for AWS Services only.
- For any other logging. It can be anything you want.
- Anything to measure is a metric.
- Anything logged is datapoint.
- They have timestamp and value.
- Datapoint is logged to a metric
- Namevalue pair associated to datapoints.
- Useful to identify source, or differentiate data based on source.
- Key value pair.
- Some services sends some dimensions by default.
- One to one relationship with metric.
- Setting alarm based on metric.
- Three states:
INSUFFICIENT DATA,OK and ALARM
INSUFFICIENT DATA: Usually an initial stage, where alarm can't gather enough data to determine the stage.OK: A stage where metric doesn't create any problem compared to criteria defined in alarm. (E.g. Cost still below defined limit, CPU Usage is not upto defined limit)ALARM: A stage where metric creates a problem compared to criteria defined in alarm. (E.g. Cost going above defined limit, CPU Usage or storage going above defined limit)