Run Metricbeat on Cloud Foundryedit

You can use Metricbeat on Cloud Foundry to retrieve and ship metrics.

Create Cloud Foundry credentialsedit

To connect to loggregator and receive the logs, Metricbeat requires credentials created with UAA. The uaac command creates the required credentials for connecting to loggregator.

uaac client add metricbeat --name metricbeat --secret changeme --authorized_grant_types client_credentials,refresh_token --authorities doppler.firehose,cloud_controller.admin_read_only

Use a unique secret: The uaac command shown here is an example. Remember to replace changeme with your secret, and update the metricbeat.yml file to use your chosen secret.

Download Cloud Foundry deploy manifestsedit

You deploy Metricbeat as an application with no route.

Cloud Foundry requires that 3 files exist inside of a directory to allow Metricbeat to be pushed. The commands below provide the basic steps for getting it up and running.

curl -L -O https://artifacts.elastic.co/downloads/beats/metricbeat/metricbeat-7.12.0-linux-x86_64.tar.gz
tar xzvf metricbeat-7.12.0-linux-x86_64.tar.gz
cd metricbeat-7.12.0-linux-x86_64
curl -L -O https://raw.githubusercontent.com/elastic/beats/7.12/deploy/cloudfoundry/metricbeat/metricbeat.yml
curl -L -O https://raw.githubusercontent.com/elastic/beats/7.12/deploy/cloudfoundry/metricbeat/manifest.yml

You need to modify the metricbeat.yml file to set the api_address, client_id and client_secret.

Load Kibana dashboardsedit

Metricbeat comes packaged with various pre-built Kibana dashboards that you can use to visualize data in Kibana.

If these dashboards are not already loaded into Kibana, you must run the Metricbeat setup command. To learn how, see Load Kibana dashboards.

If you are using a different output other than Elasticsearch, such as Logstash, you need to Load the index template manually and Load Kibana dashboards.

Deploy Metricbeatedit

To deploy Metricbeat to Cloud Foundry, run:

cf push

To check the status, run:

$ cf apps

name       requested state   instances   memory   disk   urls
metricbeat   started           1/1         512M     1G

Metric events should start flowing to Elasticsearch. The events are annotated with metadata added by the add_cloudfoundry_metadata processor.

Scale Metricbeatedit

A single instance of Metricbeat can ship more than a hundred thousand events per minute. If your Cloud Foundry deployment is producing more events than Metricbeat can collect and ship, the Firehose will start dropping events, and it will mark Metricbeat as a slow consumer. If the problems persist, Metricbeat may be disconnected from the Firehose. In such cases, you will need to scale Metricbeat to avoid losing events.

The main settings you need to take into account are:

  • The shard_id specified in the cloudfoundry module. The Firehose will divide the events amongst all the Metricbeat instances with the same value for this setting. All instances with the same shard_id should have the same configuration.
  • Number of Metricbeat instances. When Metricbeat is deployed as a Cloud Foundry application, it can be scaled up and down like any other application, with cf scale or by specifying the number of instances in the manifest.
  • Output configuration. In some cases, you can fine-tune the output configuration to improve the events throughput. Some outputs support multiple workers. The number of workers can be changed to take better advantage of the available resources.

Some basic recommendations to adjust these settings when Metricbeat is not able to collect all events:

  • If Metricbeat is hitting its CPU limits, you will need to increase the number of Metricbeat instances deployed with the same shard_id.
  • If Metricbeat has some spare CPU, there may be some backpressure from the output. Try to increase the number of workers in the output. If this doesn’t help, the bottleneck may be in the network or in the service receiving the events sent by Metricbeat.
  • If you need to modify the memory limit of Metricbeat, remember that CPU shares assigned to Cloud Foundry applications depend on the configured memory limit. You may need to check the other recommendations after that.