Harbor Administration - Access Metrics - 《Harbor v2.10 Documentation》

Harbor Exporter Metrics
Harbor Core Metrics
Registry Metrics
Harbor Jobservice metrics
Scraping Metrics with Prometheus
- From a Kubernetes cluster

Harbor exposes some key metrics needed for operators and administrators to monitor how your Harbor instance is running in real time. Observability is a key feature for operating a service in production and using this data you can identify abnormal statuses and make informed decisions to fix issues when an error occurs. Harbor exposes metrics using the Prometheus data model so you can easily start scraping your Harbor instance’s metrics using Prometheus.

In Harbor v2.2 and later you are able to enable metrics in your Harbor configuration file. Harbor metrics are available at <harbor_instance>:<metrics_port>/<metrics_path> based on your configured values.

Harbor metrics show data related to

Runtime information from the GO library
Performance metrics about all API requests in core
Number of requests in flight in core
Metrics provided by the docker distribution itself
Some data related to business logic which already exist in the Harbor database

Metrics are exposed by several Harbor components: exporter, core, jobservice, and registry. In addition to runtime and performance data, these components also expose Harbor specific metrics. The following sections list the available Harbor metrics.

Harbor Exporter Metrics

The exporter component metrics relate to your Harbor instance configuration and collects some data from the Harbor database. Metrics are available at <harbor_instance>:<metrics_port>/<metrics_path>.

Metrics exposed by the Harbor Exporter
Name	Description	Labels (Values)	Metric type
`harbor_project_total`	Total number of public and private projects	public (`true`,`false`)	gauge
`harbor_project_repo_total`	Total number of repositories in a project	public (`true`,`false`), project_name	gauge
`harbor_project_member_total`	Total number of members in a project	project_name	gauge
`harbor_project_quota_usage_byte`	Total used resources of a project	project_name	gauge
`harbor_project_quota_byte`	Quota set in a project	project_name	gauge
`harbor_artifact_pulled`	Number of images pulled in a project	project_name	gauge
`harbor_project_artifact_total`	Total number of artifacts type in a project	artifact_type , project_name, public (`true`,`false`)	gauge
`harbor_health`	Current status of Harbor		gauge
`harbor_system_info`	Information about your Harbor instance	auth_mode (`db_auth`, `ldap_auth`, `uaa_auth`, `http_auth`, `oidc_auth`), harbor_version, self_registration(`true`,`false`)	gauge
`harbor_up`	Running status of Harbor components	component (`chartmuseum`, `core`, `database`, `jobservice`, `portal`, `redis`, `registry`, `registryctl`, `trivy`)	gauge
`harbor_task_queue_size`	The total number of tasks per type in the queue	instance, job, type	gauge
`harbor_task_queue_latency`	How long ago the next job to be processed was enqueued per type	instance, job, type	gauge
`harbor_task_scheduled_total`	Number of scheduled tasks	instance, job	gauge
`harbor_task_concurrency`	Total number of concurrent tasks per type on a pool	instance, job, pool, type	gauge

Harbor Core Metrics

The following are metrics pulled from the Harbor core pod and are available at <harbor_instance>:<metrics_port>/<metrics_path>?comp=core.

Metrics exposed by Harbor Core
Name	Description	Labels (Values)	Metric type
`harbor_core_http_inflight_requests`	The total number of requests	operation (values from `operationId` in Harbor API. Some legacy endpoints do not have an `operationId`, so the label value is `unknown`)	gauge
`harbor_core_http_request_duration_seconds`	The time duration of the requests	method (`GET`, `POST`, `HEAD`, `PATCH`, `PUT`), operation (values from `operationId` in Harbor API. Some legacy endpoints do not have an `operationId`, so the label value is `unknown`), quantile	summary
`harbor_core_http_request_total`	The total number of requests	method (`GET`, `POST`, `HEAD`, `PATCH`, `PUT`), operation (values from `operationId` in Harbor API. Some legacy endpoints do not have an `operationId`, so the label value is `unknown`)	counter

Registry Metrics

The following are metrics pulled from the Docker distribution and are available at <harbor_instance>:<metrics_port>/<metrics_path>?comp=registry.

Metrics exposed by Harbor Core
Name	Description	Labels (Values)	Metric type
`registry_http_in_flight_requests`	The in-flight HTTP requests	handler	gauge
`registry_http_request_duration_seconds_bucket` `registry_http_request_duration_seconds_sum` `registry_http_request_duration_seconds_count`	The HTTP request latencies in seconds	handler, method (`GET`, `POST`, `HEAD`, `PATCH`, `PUT`), le	histogram
`registry_http_request_size_bytes_bucket registry_http_request_size_bytes_sum registry_http_request_size_bytes_count`	The HTTP request sizes in bytes	handler, le	histogram
`registry_http_requests_total`	Total number of HTTP requests made	code,handler,method,le	counter
`registry_http_response_size_bytes_bucket` `registry_http_response_size_bytes_sum` `registry_http_response_size_bytes_count`	The HTTP response sizes in bytes	handler,le	histogram
`registry_storage_action_seconds_bucket` `registry_storage_action_seconds_sum` `registry_storage_action_seconds_count`	The number of seconds that the storage action takes	action,driver,le	histogram
`registry_storage_cache_total`	The number of cache request received	type	gauge

Harbor Jobservice metrics

The following are metrics pulled from the Harbor Jobservice and are available at <harbor_instance>:<metrics_port>/<metrics_path>?comp=jobservice.

Metrics exposed by Harbor Jobservice
Name	Description	Labels (Values)	Metric type
`harbor_jobservice_info`	The information of Jobservice	instance, job, node, pool, workers	gauge
`harbor_jobservice_task_total`	The number of processed tasks per job type	instance, job, status, type	counter
`harbor_jobservice_task_process_time_seconds`	The duration of the task processing time	instance, job, quantile, status, type	summary

Scraping Metrics with Prometheus

To begin accessing your Harbor instance’s metrics with Prometheus,

Enable exposing metrics in your harbor.yml configuration file and set the port and path for metrics to be exposed on. Also see more about reconfiguring your Harbor instance.
Set up a Prometheus server, see the Prometheus documentation for more information on installing.

Configure your Prometheus config file to scrape Harbor metrics exposed at your configured port and path. Below is an example scrape config, see the Prometheus documentation for all available scrape configuration options.

  scrape_configs:
    - job_name: 'harbor-exporter'
      scrape_interval: 20s
      static_configs:
        # Scrape metrics from the Harbor exporter component
        - targets: ['<harbor_instance>:<metrics_port>']
    - job_name: 'harbor-core'
      scrape_interval: 20s
      params:
        # Scrape metrics from the Harbor core component
        comp: ['core']
      static_configs:
        - targets: ['<harbor_instance>:<metrics_port>']
    - job_name: 'harbor-registry'
      scrape_interval: 20s
      params:
        # Scrape metrics from the Harbor registry component
        comp: ['registry']
      static_configs:
        - targets: ['<harbor_instance>:<metrics_port>']
    - job_name: 'harbor-jobservice'
      scrape_interval: 20s
      params:
        # Scrape metrics from the Harbor jobservice component
        comp: ['jobservice']
      static_configs:
        - targets: ['<harbor_instance>:<metrics_port>']

Once you have configured your Prometheus server to collect your Harbor metrics, you can use Grafana to visualize your data. An example Grafana dashboard is available in the Harbor repo to help you get started visualizing Harbor metrics.

From a Kubernetes cluster

You can also use Prometheus to collect metrics from a Harbor instance deployed in your Kubernetes cluster. You should already have installed Prometheus and set up to pull metrics from your cluster.

Create a ServiceMonitor in Prometheus for Harbor.

apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: harbor
  labels:
    app: harbor
spec:
  selector:
    matchLabels:
      app: harbor
  endpoints:
  - port: metrics

Enable Harbor to expose metrics by updating your harbor-helm values.yaml file and set metrics.enabled to true. You can also edit the port and path the metrics are exposed on by updating the available harbor-helm chart configuration options for metrics.

Prometheus should now show your Harbor instance’s metrics.