Cloudera DataFlow for the Public Cloud (CDF-PC) is an entire self-service streaming knowledge seize and motion platform based mostly on Apache NiFi. It permits builders to interactively design knowledge flows in a drag and drop designer, which could be deployed as constantly operating, auto-scaling movement deployments or event-driven serverless capabilities. CDF-PC comes with a monitoring dashboard out of the field for knowledge movement well being and efficiency monitoring. Key efficiency indicators (KPIs) and related alerts assist prospects monitor what issues for his or her use instances.
Many organizations have invested in central monitoring and observability instruments corresponding to Prometheus and Grafana and are in search of methods to combine key knowledge movement metrics into their present structure.
On this weblog we’ll dive into how CDF-PC’s help for NiFi reporting duties can be utilized to observe key metrics in Prometheus and Grafana.
Goal structure: connecting the items
The important thing to insightful Grafana dashboards is getting access to related utility metrics. In our case, these are NiFi metrics of our movement deployment. We due to this fact want to have the ability to expose NiFi metrics for Prometheus so it might scrape them earlier than we are able to construct dashboards in Grafana. CDF-PC’s help for Prometheus reporting duties and inbound connections permits Prometheus to scrape metrics in actual time. As soon as the metrics are in Prometheus, querying it and constructing dashboards on high of it in Grafana is easy. So let’s take a more in-depth take a look at how we get from having metrics in our movement deployment to a totally featured Grafana dashboard by implementing the goal structure proven in Determine 1 beneath.
Configuring a CDF deployment to be scraped by Prometheus
Beginning with CDF-PC 2.6.1, now you can programmatically create NiFi reporting duties to make related metrics accessible to numerous third get together monitoring methods. The Prometheus reporting job that we’ll use for this instance creates an HTTP(S) metrics endpoint that may be scraped by Prometheus brokers or servers. To make use of this reporting job in a CDF-PC deployment, we’ve to finish the next steps:
- Make sure that the HTTP(s) metrics endpoint is reachable from Prometheus by configuring an inbound connections endpoint when creating the deployment.
- Create and configure the Prometheus reporting job utilizing the CDP CLI after profitable deployment creation.
Making a deployment with an inbound connections endpoint
When making a deployment, CDF-PC offers you the choice to permit NiFi to obtain knowledge by configuring an inbound connections endpoint. When the choice is checked, CDF-PC will counsel an endpoint hostname that you would be able to customise as wanted.
The inbound connections endpoint offers exterior functions the power to ship knowledge to a deployment, or in our case, connect with a deployment to scrape its metrics. Along with the endpoint hostname we even have to supply at the least one port that we need to expose. In our case we’re utilizing Port 9090 and are exposing it with the TCP protocol.
After you might have created your deployment with an inbound connection endpoint, navigate to the NiFi configuration tab within the deployment supervisor the place you will notice all related info to attach exterior functions to the deployment. Now that the deployment has been created, we are able to transfer on to the subsequent step—creating the reporting job.
Creating and configuring the NiFi Prometheus reporting job
We are able to now use the CDP CLI to create and configure the Prometheus reporting job. Obtain and configure the CDP CLI. Just be sure you are operating at the least model 0.9.101 by operating cdp –model.
The command we’re going to make use of is the cdp dfworkload create-reporting-task command. It requires the deployment CRN, setting CRN and a JSON definition of the reporting job that we need to create. Copy the deployment CRN from the deployment supervisor, get the setting CRN for the related CDP setting, and begin developing the command.
cdp dfworkload create-reporting-task --deployment-crn crn:cdp:df:us-west-1:9d74eee4-1cad-45d7-b645-7ccf9edbb73d:deployment:eb2717f3-1bdf-4150-bd33-5b15d715bc7d/5cdc4d43-2991-4d4c-99fc-c400cd15853d --environment-crn crn:cdp:environments:us-west-1:9d74eee4-1cad-45d7-b645-7ccf9edbb73d:setting:bf58748f-7ef4-477a-9c63-448b51e5c98f
The lacking piece is offering the details about which reporting job we need to create. All supported reporting duties could be handed in utilizing their configuration JSON file. Right here’s the JSON configuration for our Prometheus reporting job. It consists of a Prometheus-specific properties part adopted by generic reporting job configuration properties corresponding to whether or not the reporting job must be began, how steadily it ought to run, and the way it must be scheduled.
{ "identify": "PrometheusReportingTask", "sort": "org.apache.nifi.reporting.prometheus.PrometheusReportingTask", "properties": { "prometheus-reporting-task-metrics-endpoint-port": "9090", "prometheus-reporting-task-metrics-strategy": "All Parts", "prometheus-reporting-task-instance-id": "${hostname(true)}", "prometheus-reporting-task-client-auth": "No Authentication", "prometheus-reporting-task-metrics-send-jvm": "false" }, "propertyDescriptors": {}, "scheduledState": "RUNNING", "schedulingPeriod": "60 sec", "schedulingStrategy": "TIMER_DRIVEN", "componentType": "REPORTING_TASK" }
| Configuration Property | Description |
| prometheus-reporting-task-metrics-endpoint-port | The port that this reporting job will use to show metrics. This port should match the port you specified earlier when configuring the inbound connection endpoint. |
| prometheus-reporting-task-metrics-strategy | Defines granularity on which to report metrics. Supported values are “All Parts,” “Root Course of Group,” and “All Course of Teams.” Use this to restrict metrics as wanted. |
| prometheus-reporting-task-instance-id | The ID that shall be despatched alongside the metrics. You need to use this property to determine your deployments in Prometheus. |
| prometheus-reporting-task-client-auth | Does the endpoint require authentication? Supported values are “No Authentication,” “Need Authentication,” or “Want Authentication”. |
| prometheus-reporting-task-metrics-send-jvm | Defines whether or not JVM metrics are additionally uncovered. Supported values are “true” and “false.” |
Desk 1: Prometheus configuration properties of the NiFi Prometheus reporting job.
You possibly can both move the JSON file as a parameter to the CLI command or reference a file. Let’s assume we’re saving the above JSON content material in a file referred to as prometheus_reporting_task.json.
Now we are able to assemble our ultimate CLI command that may create the specified reporting job:
cdp dfworkload create-reporting-task --deployment-crn crn:cdp:df:us-west-1:9d74eee4-1cad-45d7-b645-7ccf9edbb73d:deployment:eb2717f3-1bdf-4150-bd33-5b15d715bc7d/5cdc4d43-2991-4d4c-99fc-c400cd15853d --environment-crn crn:cdp:environments:us-west-1:9d74eee4-1cad-45d7-b645-7ccf9edbb73d:setting:bf58748f-7ef4-477a-9c63-448b51e5c98f --file-path prometheus_reporting_task.json
After executing the command, it's best to get a response again that comprises the reporting job CRN: { "crn": "crn:cdp:df:us-west-1:9d74eee4-1cad-45d7-b645-7ccf9edbb73d:reportingTask:eb2717f3-1bdf-4150-bd33-5b15d715bc7d/66a746af-018c-1000-0000-00005212b3ea" }
To verify that the reporting job was created, navigate to the NiFi configuration tab within the deployment supervisor and confirm that the reporting job part displays the reporting duties you created utilizing the CLI.
Now that our movement deployment and reporting job have been created, we are able to transfer on to the subsequent step and configure the Prometheus server to scrape this deployment.
Configuring Prometheus to observe a CDF deployment
To outline a brand new scraping goal for Prometheus, we have to edit the Prometheus configuration file. Open the prometheus.yaml file so as to add the CDF deployment as a goal.
Create a brand new job, e.g. with CDF deployment as its identify. Subsequent, copy the endpoint hostname of your CDF deployment from the deployment supervisor and add it as a brand new goal.
scrape_configs:
# The job identify is added as a label `job=` to any timeseries scraped from this config. - job_name: "CDF Deployment" # metrics_path defaults to '/metrics' # scheme defaults to 'http'. static_configs: - targets: ["wikipediaprometheus.inbound.dfx.q5crnxxe.xcu2-8y8x.dev.cldr.work:9090"]
Apply the configuration modifications and navigate to the Prometheus net console to substantiate that our CDF deployment is being scraped. Go to the Standing→Targets and confirm that your CDF Deployment is “Up.”
As soon as Prometheus has began scraping, you possibly can discover all NiFi metrics within the metrics explorer and begin constructing your Prometheus queries.
Pattern Grafana dashboard
Grafana is a well-liked alternative for visualizing Prometheus metrics, and it makes it straightforward to observe key NiFi metrics of our deployment.
Create a Prometheus connection to make all metrics and queries accessible in Grafana.
Now that Grafana is linked to Prometheus, you possibly can create a brand new dashboard and add visualizations.
Let’s say we need to create a graph that represents the info that this deployment has acquired from exterior sources. Choose “add visualization” in your dashboard and ensure your Prometheus connection is chosen as the info supply.
Choose the nifi_amount_bytes_received metric. Use the label filters to slim down the part within the movement. Through the use of component_name and “Hiya World Prometheus,” we’re monitoring the bytes acquired aggregated by your entire course of group and due to this fact the movement. Alternatively you possibly can monitor all elements if no filter is outlined or monitor particular person processors too.
With all NiFi metrics being accessible in Grafana, we are able to now construct a full dashboard monitoring all related metrics. Within the instance beneath we’re monitoring whole bytes acquired/despatched, the variety of movement information queued in all elements, the common lineage length and the present NiFi JVM heap utilization which assist us perceive how our flows are doing.
Conclusion
The NiFi Prometheus reporting job, along with CDF inbound connections makes it straightforward to observe key metrics in Prometheus and create Grafana dashboards. With the not too long ago added create-reporting-task CDF CLI command, prospects can now automate organising Prometheus monitoring for each new deployment as a part of their commonplace CI/CD pipeline.
Check out CDF-PC utilizing the general public 5 day trial and take a look at the Prometheus monitoring demo video beneath for a step-by-step tutorial.