This glossary covers the terms used in Managed Service for Prometheus. Understanding these terms helps you configure data collection, write queries, and set up alerts.
At a high level, the Prometheus monitoring model works like this: a Prometheus agent discovers targets through service discovery, scrapes metrics from exporters, stores data as timelines of samples in remote storage, and uses PromQL to query data and define alert rules.
Data model
| Term | Description |
|---|
| metric | A set of tagged data points exposed by a collection target. Metrics reflect the operational or business status of monitored objects. Managed Service for Prometheus uses the OpenMetrics standard format to describe metrics. Each metric is identified by its name and a set of tags. |
| tag | A key-value pair attached to a metric that differentiates a specific dimension, such as method="GET" or status="200". Tags are similar to labels in the open source Prometheus ecosystem. The combination of a metric name and its tags uniquely identifies a timeline. |
| timeline | A unique time series identified by a metric name and a specific set of tags. For example, http_requests_total{method="GET", status="200"} is one timeline, while http_requests_total{method="POST", status="500"} is another. Each timeline contains a sequence of samples. |
| sample | A single data point on a timeline, consisting of a float64 value and a millisecond-accurate timestamp. |
Data collection
| Term | Description |
|---|
| exporter | An application that runs alongside a monitored object and converts its monitoring data into the OpenMetrics format so that the Prometheus agent can scrape it. More than 100 official and third-party exporters are available. For more information, see Exporters. |
| target | An endpoint that the Prometheus agent scrapes for metric data. A target either exposes its own metrics directly, or the Prometheus agent exposes the metrics of the monitored objects on behalf of the target. Targets are organized into jobs. |
| job | A configuration that defines scrape behavior for a group of targets, including the scrape interval and access control settings. |
| Prometheus agent | A component deployed in a Kubernetes cluster on the user side or the cloud service side. The Prometheus agent automatically discovers targets through service discovery, scrapes metric data, and writes the data to remote storage. |
| service discovery | A mechanism that allows the Prometheus agent to find targets automatically without static configuration. Supported methods include Kubernetes service discovery, Consul, and Eureka. You can use ServiceMonitor and PodMonitor to expose collection targets. |
Data storage and querying
| Term | Description |
|---|
| remote storage | A time series storage component developed by Alibaba Cloud. Remote storage supports the Prometheus Remote Write protocol and is fully hosted as a cloud service. The Prometheus agent writes scraped data to remote storage for long-term retention and querying. |
| PromQL | The query language for Managed Service for Prometheus. PromQL supports instant queries and range queries, and provides built-in functions and operators. Use PromQL to aggregate, filter, predict, and combine metric data. PromQL is also used to define alert rules and pre-aggregation rules. |
| pre-aggregation | A mechanism based on recording rules that pre-computes PromQL expressions and saves the results as new metrics. Pre-aggregation reduces query latency for frequently used or computationally expensive queries. |
Alerting and monitoring
| Term | Description |
|---|
| alert rule | A PromQL-based condition that triggers an alert when the expression evaluates to true. Alert rules follow the Prometheus Alerting Rule format. |
| cloud service monitoring | A built-in integration that connects Managed Service for Prometheus with monitoring data from other Alibaba Cloud services. To monitor a specific Alibaba Cloud service, create a Prometheus instance for that service. |
Service management
| Term | Description |
|---|
| Prometheus Service | A fully managed monitoring service provided by Alibaba Cloud. Managed Service for Prometheus is compatible with the open source Prometheus ecosystem, provides out-of-the-box dashboards for monitoring a wide variety of components, and supports multiple types of Prometheus instances. |
| Prometheus instance | A logical unit in Managed Service for Prometheus that manages data collection, storage, and analysis. You can create multiple Prometheus instances to separate monitoring for different environments or services. |