Metrics provide insight into what is going on in the cluster. They are an invaluable resource for monitoring and debugging.
Alluxio has a configurable metrics system based on the .
The metrics system exposes metrics in Prometheus exposition format.
Alluxio’s metrics are partitioned into different instances corresponding to Alluxio components. The following instances are currently supported:
Coordinator: The Alluxio coordinator process
Worker: The Alluxio worker process
FUSE process: The Alluxio FUSE process
Usage
Send an HTTP request to /metrics/ of the target Alluxio processes and get a snapshot of all metrics.
# Get the metrics from Alluxio processes
$ curl <COORDINATOR_HOSTNAME>:<COORDINATOR_WEB_PORT>/metrics/
$ curl <WORKER_HOSTNAME>:<WORKER_WEB_PORT>/metrics/
$ curl <FUSE_HOSTNAME>:<FUSE_WEB_PORT>/metrics/
For example, for the local processes:
# Get the local coordinator metrics with its default web port 19999
$ curl 127.0.0.1:19999/metrics/
# Get the local worker metrics with its default web port 30000
$ curl 127.0.0.1:30000/metrics/
# Get the local fuse metrics with its default web port 49999
$ curl 127.0.0.1:49999/metrics/
Integration
Kubernetes Operator
The operator supports building a cluster with bundled Prometheus and Grafana. By default, both Prometheus and Grafana are enabled.
Grafana can be disabled by setting its enabled field to be false. Prometheus cannot be disabled.
The configuration and the Grafana template are already included. Just set the following switch in the AlluxioCluster configuration:
Grafana is a metrics analytics and visualization software used for visualizing time series data.
You can use Grafana to better visualize the various metrics that Alluxio collects.
The software allows users to more easily see changes in memory, storage, and completed operations in Alluxio.
Grafana supports visualizing data from Prometheus. The following steps can help you to build your Alluxio monitoring based on Grafana and Prometheus easily.
If your Grafana dashboard appears like the screenshot below, you have built your monitoring successfully.
By default, only the Cluster row is unfolded, to show the abstract of the current status.
The Process row shows the resource consumption and JVM-related metrics, which can be filtered by either services or instances at the top.
The other rows show the details of certain components and can be filtered by instances.
Accessing Grafana via Node Hostname
The Grafana will expose its service on the 8080 port on its host. Use kubectl to get the hostname:
kubectl -n alx-ns get pod $(kubectl -n alx-ns get pod -l app.kubernetes.io/component=grafana --no-headers -o custom-columns=:metadata.name) -o jsonpath='{.spec.nodeName}'
Assume the hostname is foo.kubernetes.org, then you can access the Grafana service on:
http://foo.kubernetes.org:8080/
Accessing Grafana via Port Forwarding
If you are unable to access Grafana directly through the node's hostname due to network issues,
you can use port forwarding to map Grafana's port to your local machine, allowing you to access it locally.
Run the following kubectl port-forward command to set up port forwarding:
kubectl -n alx-ns port-forward $(kubectl -n alx-ns get pod -l app.kubernetes.io/component=grafana -o jsonpath="{.items[0].metadata.name}") 3000:3000
You can then access the Grafana service locally at:
http://localhost:3000
Prometheus
If launching an independent Prometheus service, refer to the following prometheus.yml to scrape the metrics. Note that job_name should not be changed to be compatible with built-in Grafana dashboard.
If launching an independent Prometheus service in Kubernetes, refer to the following snippet to additionally extract pod annotations from the Kubernetes environment.
Again, note that job_name should not be changed to be compatible with built-in Grafana dashboard.
The following metadata for the Prometheus pod is required:
labels:
app.kubernetes.io/instance: alluxio # used to distinguish different alluxio cluster
app.kubernetes.io/component: worker # values from operator deployment are coordinator, worker, fuse, and csi-fuse. depends on the pod
annotations:
prometheus.io/scrape: "true"
# values should match with the port of the component. By default, it's 19999 for coordinator, 30000 for worker, and 49999 for fuse
prometheus.io/port: "30000"
prometheus.io/path: "/metrics/"
Datadog
Alluxio emits metrics in the Prometheus format, which allows Datadog to directly integrate with Alluxio.
Update Datadog configuration by adding multiple prometheus_url entries under the instances section in the configuration file
Example configuration snippet for polling multiple sources:
By following these steps, Datadog can seamlessly collect and monitor Alluxio metrics, providing insights and visibility into your Alluxio cluster's performance and health.
Refer to for more details of the exported metrics.
Download the Grafana template JSON file for Alluxio:
Import the template JSON file to create a dashboard. See this for importing a dashboard.
Add the Prometheus data source to Grafana with a custom name, ex. prometheus-alluxio. Refer to the for help on importing a dashboard.