Elastic Agent and Synthetic (Real Browser) Integration

[paulb-elastic]

This builds upon the initial Phase 1 implementation.

As an Elastic user
I want to be able to add an real browser Synthetic monitor to an Elastic Agent via an integration
So that I can centrally deploy the monitor without having to install and configure a separate Synthetic or Heartbeat Agent binary and YML file

ACs:

• User can add a Synthetic (real browser) Integration (in the same way they can other integrations)
• User can add a monitor, with the configuration of:
  • Name, frequency and other base fields
  • Monitor Script - which can be defined in one of two ways:
    • Inline (the actual script, as is currently put into the heartbeat.yml file)
    • Suite ZIP location (as implemented in Remote Zip URL support for heartbeat synthetics #271)
      • Which requires manadatory location, optional folder, optional username/password
• User can define which Elastic Agent, the integration will run on (via the integration policy -> agent policy -> agent assignment)
• The real browser Synthetic monitor will run at the given frequency, and results will appear in Uptime
• The Synthetic monitor will operate via the Docker implementation (note that this adds an environment variable to ensure Heartbeat will only run Synthetics from Docker)

Consider the AC referenced below

[andrewvc]

@drewpost @paulb-elastic @dominiqueclarke I'd like to expand the last AC around the experience setting up the docker image. We should do the following IMHO:

1. Attempt to detect if the agent policy is backed by offline agents (e.g. ones capable of running synthetics)
2. Show an error if that is the case before the policy is saved clearly explaining the situation
3. Create a tutorial in correctly setting up synthetics with fleet and link to it in the UI. (It should be part of https://www.elastic.co/guide/en/observability/master/synthetics-quickstart.html)

[paulb-elastic]

I do like the idea of this - giving a more useful pointer of a misconfiguration before it’s set up (at which point the user is left debugging why they have no data).

I guess the only thing would be if there is a use case where a user might want to set up a load of monitors before assigning the policy to an agent, in which case this would prevent that. I’m not convinced that is a valid use case, just asking the question (/cc @drewpost)

[paulb-elastic]

Linking to elastic/beats#26398 which is the requirement for an Elastic Agent + Synthetics dependencies image

[dominiqueclarke]

This has been resolved.