Prometheus Metrics

Prometheus integration provides a mechanism for ingesting Prometheus metrics.

Version compatibility

Tested Prometheus version: 2.22.1

Prometheus feature enhancements and metric name changes between versions can affect dashboards. To find the latest information on Prometheus versions, see https://github.com/prometheus/prometheus/releases.

Prerequisites

• A Kubernetes environment that contains a Prometheus server.
• A Kubernetes environment with at least 1GB of RAM available for the Prom-Agent.

Recommended Prometheus Remote-Write Configuration tuning:

See: https://prometheus.io/docs/practices/remote_write/

Prometheus handles most of the above configurations dynamically based on the responsiveness of OpsRamp Agent, so it is best to leave them at their default values. If that does not work, the values can be altered based on the above descriptions.

Constraints

The number of active series per metric per client is 50000. You can avoid the limit by configuring Prometheus to filter metrics. For example, use the following configuration to limit apiserver_request_duration_seconds_bucket, and etcd_request_duration_seconds_bucket metrics:

remoteWrite:
  - url: http://prom-agent-service.prom-agent:20460/push
    writeRelabelConfigs:
    - action: drop
      regex: apiserver_request_duration_seconds_bucket
      sourceLabels:
      - __name__
    - action: drop
      regex: etcd_request_duration_seconds_bucket
      sourceLabels:
      - __name__

Add a Prometheus integration and agent

Adding an integration

1. From All Clients, select a client.
2. Go to Setup > Account.
3. Select the Integrations and Apps tab.
4. The Installed Integrations page, where all the installed applications are displayed. Note: If there are no installed applications, it will navigate to the Available Integrations and Apps page.
5. Click + ADD on the Installed Integrations page. The Available Integrations and Apps page displays all the available applications.
6. Search for Prometheus metrics using the search option available.
   Note: Alternatively, you can use the All Categories option to search.
7. Click +ADD on the Prometheus metrics tile.
8. Enter a Name for your Prometheus integration. The name must contain only lower case alphanumeric characters or dash (-) and start and end with an alphanumeric character. For example, my-name or 123-abc.
9. Select the type from the dropdown.
10. Click Next.

Configuring Prometheus and the agent

11. Follow the Prom-Agent Deployment Configuration, Prometheus remote-write Configuration, and Prometheus Label Configuration instructions, copying from the code snippets on the right, as noted in the instructions.
12. (optional) After adding a Prometheus integration and an agent, click Test Connection to verify connectivity. You should receive a Connected successfully message. If not, troubleshoot the possible fault conditions described, below.
13. Click Finish.

Test Connection troubleshooting

Problem with network connection. Unable to test

The cloud is unable to communicate with the agent in the Prometheus environment. Check the agent deployment.

Possible connectivity issue. Verify that the Prometheus remote-write settings are correctly configured.

Prometheus is not able to push metrics to the agent. Check the Prometheus remote-write configuration parameter. If the configuration is correct but not reflected in Prometheus, delete the Prometheus Kube POD so restarts with the new configuration.

Connect OpsRamp Prometheus Agents behind a Proxy

To connect OpsRamp Prometheus agent using a proxy, set the following environment variables:

CONN_MODE=proxy
PROXY_SERVER=<ProxyServerIP>
PROXY_PORT=<ProxyPort>

If the proxy server needs authentication, set the following credentials:

PROXY_USER=<User>
PROXY_PASSWORD=<Password>

Set the Prometheus configuration parameters as shown in the following examples:

externalLabels:  
  OpsRampIntegrationName: PrometheusIntegration1

remoteWrite:  
- url: http://prom-agent-service.prom-agent:20460/push

Use the OpsRampIntegrationName label to associate the metrics with the integration.

Configure high-availability (HA) Prometheus

Assign unique externalLabels to each Prometheus replica. Otherwise, data clashes can occur between replicas.

You can also try the combinations described in Multiple prom-agent setup instructions to point the HA Prometheus instance to OpsRamp.

Configuration to filter metrics pushed to remote-write

Prometheus can be configured to filter the metrics to be pushed to a remote endpoint like the agent. See the Prometheus configuration remote_write property documentation for more information.

Consider the following examples.

Drop samples example

The following writeRelabelConfigs configuration drops samples with a metric name that starts with go:

remoteWrite:
- url: http://prom-agent-service.prom-agent:20460/push
  writeRelabelConfigs:
  - sourceLabels: [__name__]
    regex: 'go_.*'
    action: drop

Keep samples example

The following writeRelabelConfigs configuration keeps samples with a metric name that starts with go:

remoteWrite:
- url: http://prom-agent-service.prom-agent:20460/push
  writeRelabelConfigs:
  - sourceLabels: [__name__]
    regex: 'go_.*'
    action: keep

Replace samples example

You can also change a label using the replace action before pushing it to a remote endpoint. The following example replaces samples with a metric name that starts with go with the name sample_go_label:

remoteWrite:
- url: http://prom-agent-service.prom-agent:20460/push
  writeRelabelConfigs:
  - sourceLabels: [__name__]
    regex: 'go_.*'
    action: replace
    targetLabels: sample_go_label

Multiple prom-agent setup instructions

Multiple-Prometheus setups have three valid use cases.

Single integration deployed across multiple clusters

Create a single Prometheus integration and deploy the same provided agent YAML configuration across multiple Kubernetes clusters:

To differentiate between the clusters and identify the source of the metrics, add another external label.

Different integration for each cluster

Create a Prometheus integration for each Kubernetes cluster that has a Prometheus server:

Use the OpsRampIntegrationName label to associate the metrics with the integration.

Agent deployed in one cluster with services exposed

Create a Prometheus integration, deploy the provided agent YAML configuration in one of the clusters, and expose its service using NodePort on the network:

Add another external label to identify the source of the metrics and to differentiate between clusters. Make sure you allocate more CPU and memory resources to the agent so it can handle metrics from multiple clusters.

To expose the Prometheus agent service to the network:

1. Create a YAML file with the following content and deploy it:

       apiVersion: v1
       kind: Service
       metadata:
         name: prom-agent-expose-service
         namespace: prom-agent
       spec:
         type: NodePort
         ports:
           - nodePort: 30460   #The NodePort that you want to give to OpsRamp Prometheus Agent.
             port: 20460       #The Default OpsRamp Prometheus Agent port. Change this if you have modified the port for OpsRamp Prometheus Agent.
             protocol: TCP
             targetPort: 20460 #The Default OpsRamp Prometheus Agent port. Change this if you have modified the port for OpsRamp Prometheus Agent.   selector:
           app: prom-k8s-agent
       

2. After deployment, configure the Prometheus Remote Write URL based on the node IP of the Prometheus agent:

       remoteWrite:
       - url: http://{node IP}:{node Port}/push #example: http://172.26.32.230:30460/push
       

After exposing the service on the network, you can point any number of Prometheus servers in the cluster to the agent. To be able to get the communication status, or health, of an individual Prometheus server, add the following environment variable in the agent deployment YAML:

- name: DISABLE_REMOTE_ADDRESS
  value: "TRUE"

Viewing and monitoring health status

The agent is self-monitoring and pushes its health status to the cloud as the promagent_health_status metric. Health is reported periodically at default five-minute intervals. You can view health status metric data on the resource details page on the Infrastructure and My Dashboards pages. To view in the dashboard, mouse over Dashboards and choose Dashboard 2.0.

The promagent_health_status metric includes id, integration_uuid, and prometheus_address labels.

You can define an alert on the agent health status metric by navigating to Setup > Account > Alert Definitions. Create an Alert Definition Policy on the promagent_health_status metric and define the policy using agent health status metric labels.