How to develop Agent checks

StackState Self-hosted v4.6.x

This page describes StackState version 4.6.

Go to the documentation for the latest StackState release.

This document covers how to create your first check with Agent v2 Check API. Following topics are covered in this document: the agent directory structure, configuring your check, writing your first check, sending topology, metrics, events, and service checks as well as how to add external python dependencies and putting it all together.

Installing Agent v2 StackPack

To install this StackPack navigate to StackState’s StackPacks page using left menu pane and locate the “StackState Agent V2” section. Click on the Agent V2 icon and this opens the installation page. Click on "Install" button and follow installation instructions provided by the StackPack.

Directory Structure

Before starting your first check it is worth understanding the checks directory structure. There are two places that you will need to add files for your check. The first is the checks.d folder, which lives in your Agent root.

For all Linux systems, you should find it at:


For Windows Server >= 2008 you should find it at:

C:\Program Files (x86)\StackState\StackState Agent\checks.d\


C:\Program Files\StackState\StackState Agent\checks.d\

The configuration folder is conf.d which also lives in your Agent root.

For Linux, you should find it at:


For Windows, you should find it at:

C:\ProgramData\StackState\StackState Agent\conf.d\


C:\Documents and Settings\All Users\Application Data\StackState\StackState Agent\conf.d\

You can also add additional checks to a single directory, and point to it in StackState.yaml:

additional_checksd: /path/to/custom/checks.d/

For the remainder of this document these paths will be referred to as checks.d and conf.d.

Check Configuration

Each check has a configuration directory and file that will be placed in the conf.d directory. Configuration is written using YAML. The folder name should match the name of the check (e.g.: and example.d containing the conf.yaml configuration file). We will be using the StackState "Skeleton" / bare essentials check and configuration as a starting point.

The configuration file for the "Skeleton" check has the following structure:


  - url: "some_url"
    # min_collection_interval: 30 # use in place of collection_interval for Agent v2.14.x or earlier 
    collection_interval: 30 # the collection interval in seconds. This check will run try to run every 30 seconds

YAML files must use spaces instead of tabs.


The init_config section allows you to have an arbitrary number of global configuration options that will be available on every run of the check in self.init_config.


The instances section is a list of instances that this check will be run against. Your check(...) method is run once per instance each collection interval. This means that every check will support multiple instances out of the box. A check instance is an object that should contain all configuration items needed to monitor a specific instance. An instance is passed into the execution of the check method in the instance parameter. collection_interval can be added to define how often the check should be run. If the value is set to 30, it means that this check will be scheduled for collection every 30 seconds. If the check runtime exceeds the collection_interval, the Agent will warn in the agent log with Check <name> did not finish execution with the defined collection_interval time '<time>', skipping execution.... The default is 40, if no collection_interval is specified. The collection_interval setting has superseded the min_collection_interval setting that was used before. The agent will still accept the min_collection_interval setting and interpret it as if the collection_interval was specified.

To synchronize multiple instances in StackState you have to create a multi-tenant StackPack.

➡️ Learn more about developing StackPacks..

Setting up your check configuration

Create a directory named {your_check_name}.d inside the conf.d directory. The YAML configuration below can be used as a starting point for your check configuration. Save it to a file called conf.yaml in the conf.d/{your_check_name}.d/ directory.

  - url: "some_url"

First Check

Now you can start defining your first check. The following "Skeleton" check can be used as a good starting point:

from stackstate_checks.base import AgentCheck, ConfigurationError, AgentIntegrationInstance

class ExampleCheck(AgentCheck):

    # This method should also be overriden to uniquely identify your integration instance. The AgentIntegrationInstance is synchronized by the StackState Agent V2 StackPack. All topology elements produced by this check can be found by filtering on the `integration-type:{example}` and `integration-url:{instance_url}` tags in StackState for this example.
    def get_instance_key(self, instance):
        if 'url' not in instance:
            raise ConfigurationError('Missing url in topology instance configuration.')
        instance_url = instance['url']
        return AgentIntegrationInstance("example", instance_url)

    # check is the entry point of your agent check, `instance` is a dictionary containing the instance that was read from conf.yaml  
    def check(self, instance):
        self.log.debug("starting check for instance: %s" % instance)
        self.log.debug("successfully ran check for instance: %s" % instance)

Setting up your check

You can now take the "Skeleton" Check snippet given above and save inside a file named: {your_check_name}.py in the checks.d directory.

Load values from the instance config

Values can be loaded from the instance config object in the following ways:

# gets the value of the `url` property
url = instance['url']

# gets the value of the `default_timeout` global config from the init_config section of conf.yaml or defaults to 5
default_timeout = self.init_config.get('default_timeout', 5)

# gets the value of the `timeout` property or defaults `default_timeout` and casts it to a float data type. This can be done to override any global configuration on an instance basis.
timeout = float(instance.get('timeout', default_timeout))

StackState Topology Snapshots

Components and relations can be sent as part of a snapshot. A snapshot represents the total state of some external topology. By putting components and relations in a snapshot, StackState will persist all the topology elements present in the snapshot, and remove everything else for the topology instance. Creating snapshots is facilitated by two functions, these are already in place in the StackState "Skeleton" check:

  • self.start_snapshot() - used to start a snapshot. Internally, the AgentCheck interface uses the get_instance_key function to uniquely identify this topology instance.

  • self.stop_snapshot() - used to stop a snapshot. This should be done at the end of the check, after all data has been submitted.

Disable snapshots

You can disable snapshots by setting with_snapshots=False when you define a TopologyInstance. This replaces the start/stop_snapshot() invocations described above.

Send Topology

Topology data can be submitted using the self.component() and self.relation() functions in the AgentCheck interface. The example below shows how to submit two components with a relation between them:

self.component("this-host-unique-identifier", "Host", {
    "name": "this-host",
    "domain": "Webshop",
    "layer": "Machines",
    "identifiers": ["urn:host:/this-host-fqdn"],
    "labels": ["host:this-host", "region:eu-west-1"],
    "environment": "Production"

self.component("some-application-unique-identifier", "Application", {
    "name": "some-application",
    "domain": "Webshop",
    "layer": "Applications",
    "identifiers": ["urn:application:/some-application"],
    "tags": ["application:some-application", "region:eu-west-1", "hosted_on:this-host"],
    "environment": "Production",
    "version": "0.2.0"

self.relation("some-application-unique-identifier", "this-host-unique-identifier", "IS_HOSTED_ON", {})

This creates two components in StackState. One for the host named this-host and one for an application named some-application. The domain value is used in the horizontal grouping of the components in StackState and layer is used for vertical grouping. The labels, tags and environment add some metadata to the component and can also be used for filtering in StackState. An IS_HOSTED_ON relation is created between some-application and this-host. The labels and tags fields can also be used on relations to add some metadata. The component types (Host, Application) and relation type (IS_HOSTED_ON) will be automatically created in StackState and can later be used in the synchronization to create mappings for the different types.

The identifiers and the external identifier, for example some-application-unique-identifier will be used as the StackState Id. The external identifer should be unique within this integration.

Merge Identifiers

The identifiers are used within StackState to merge components across different integrations. Components with a matching identifier will be merged within StackState.

Given the following example:

self.component("this-host-unique-identifier-integration-a", "Host", {
    "name": "this-host",
    "identifiers": ["urn:host:/this-host-fqdn"],

self.component("this-host-unique-identifier-integration-b", "Host", {
    "name": "this-host",
    "identifiers": ["urn:host:/this-host-fqdn"],

These two components will be merged into a single component called this-host containing data from both integrations.

➡️ Learn more about the Agent Check Topology API

Send Metrics

The StackState Agent Check interface supports various types of metrics.

Metric data can be submitted using i.e. the self.gauge() function, or the self.count() function in the AgentCheck interface. All metrics data is stored in the StackSate Metrics data source that can be mapped to a metric telemetry stream for a component/relation in StackState:

The example below submits a gauge metric system.cpu.usage for our previously submitted this-host component:

self.gauge("system.cpu.usage", 24.5, tags=["hostname:this-host"])

Note: It is important to have a tag or combination of tags that you can use to uniquely identify this metric and map it to the corresponding component within StackState.

➡️ Learn more about the Agent Check Metric API

Send Events

Events can be submitted using the self.event() function in the AgentCheck interface. Events data is stored in the StackState Generic Events data source that can be mapped to a telemetry stream on a component in StackState:

The example below submits an event to StackState when a call to the instance that is monitored exceeds some configured timeout:

    "timestamp": int(time.time()),
    "source_type_name": "HTTP_TIMEOUT",
    "msg_title": "URL timeout",
    "msg_text": "Http request to %s timed out after %s seconds." % (instance_url, timeout),
    "aggregation_key": "instance-request-%s" % instance_url

➡️ Learn more about the Agent Check Event API

Health Synchronization

StackState can ingest check states from external monitoring systems. Before getting started, you can read up on the core concepts of health Synchronization.

To setup a health synchronization stream within a check, the following function needs to be defined:

from stackstate_checks.base import AgentCheck, ConfigurationError, HealthStreamUrn, HealthStream


class ExampleCheck(AgentCheck):


    def get_health_stream(self, instance):
        if 'url' not in instance:
            raise ConfigurationError('Missing url in topology instance configuration.')
        instance_url = instance['url']
        return HealthStream(HealthStreamUrn("example", instance_url))


This function specifies what health synchronization stream this agent check will produce to. Having this function defined will enable the api on the Agent Check.

Health Synchronization Snapshots

Like with topology, health data is presented to StackState using snapshots. This allows for removal of health information (check states) when they no longer exist in the source monitoring system. Snapshots are created by two functions:

  • - used to start a health snapshot.

  • - used to stop the snapshot, signaling that all submitted data is complete. This should be done at the end of the check, after all data has been submitted. If exceptions occur in the check, or for some other reason not all data can be produced, this function should not be called.

    Send Check State

Check states can be sent through the health synchronization API using the functions in the AgentCheck interface.

In the example below, a check state is created in StackState with the health value CRITICAL.

  • The check is attached to the component or relation matching the topology_element_identifier.

  • The check_state_id is used to distinguish check states within the current health stream.

Check states can be send through the health synchronization api, using the functions in the AgentCheck interface. The example below shows how to submit the data:

from stackstate_checks.base import Health

  name="Example check state",
  message="Optional clarifying message"

If after following the previous steps you health data does not show in StackState, there is a troubleshooting guide for the health synchronization.

Send in Stream Definitions and Health Checks

Stream Definitions and Health Checks for StackState can be sent in with Topology. This allows you to map telemetry streams with health checks onto components in your integration, leaving no work to be done in StackState. This example below sets up a metric stream called Host CPU Usage with a Maximum Average check in StackState on the this-host component.

# import
from stackstate_checks.base import AgentCheck, AgentIntegrationInstance, MetricStream, MetricHealthChecks# add this_host component with a metric stream and health check

this_host_cpu_usage = MetricStream("Host CPU Usage", "system.cpu.usage",
                                   conditions={"tags.hostname": "this-host", "tags.region": "eu-west-1"},

cpu_max_average_check = MetricHealthChecks.maximum_average(this_host_cpu_usage.identifier, "Max CPU Usage (Average)", 75, 90)

self.component("this-host-unique-identifier", "Host",
        "name": "this-host",
        "domain": "Webshop",
        "layer": "Machines",
        "identifiers": ["urn:host:/this-host-fqdn"],
        "labels": ["host:this_host", "region:eu-west-1"],
        "environment": "Production"

We create a MetricStream on the system.cpu.usage metric with some conditions specific to our component. We then create a maximum_average check on our metric stream using this_host_cpu_usage.identifier . The stream and check are then added to the streams and checks list in our this-host component.

➡️ Learn more about the Agent Check Telemetry API

Send Service Checks

Service Checks are used to submit the state of the agent integration instance. Service checks can be submitted using the self.service_check method in the AgentCheck interface. Service check data is also stored in the StackState Generic Events data source.

The example below submits a service check to StackState when it is verified that the check was configured correctly and it can communicate with the instance that is monitored:

# some logic here to test our connection and if successful:
self.service_check("example.can_connect", AgentCheck.OK, tags=["instance_url:%s" % instance_url])

The service check can produce the following states:

  • AgentCheck.OK

  • AgentCheck.WARNING

  • AgentCheck.CRITICAL

  • AgentCheck.UNKNOWN

➡️ Learn more about the Agent Check Service Check API

Add Python Dependencies

Sometimes your check may require some external dependencies. To solve this problem the StackState agent is shipped with python and pip embedded. When installing the dependencies needed by your custom check you should use the embedded pip to do so. This executable for pip can be found here:

For Linux, you should find it at:


For Windows, you should find it at:

C:\Program Files\StackState\StackState Agent\embedded\bin\pip.exe

Test your Check

Custom Agent checks need to be called by the agent. To test this, run:

For Linux:

sudo -u stackstate-agent -- stackstate-agent check <CHECK_NAME>

For Windows:

C:\Program Files\StackState\StackState Agent\embedded\agent.exe check <CHECK_NAME>

This executes your check once and displays the results.

Once you are happy with the result of your check, you can restart the StackState Agent service and your check will be scheduled alongside the other agent checks.

For Linux:

sudo service stackstate-agent restart

For Windows:

"C:\Program Files\StackState\StackState Agent\embedded\agent.exe" restart-service


To troubleshoot any issues in your custom check, you can run the following status command and find your check in the output:

For Linux:

sudo -u stackstate-agent -- stackstate-agent status

For Windows:

"C:\Program Files\StackState\StackState Agent\embedded\agent.exe" status

If your issue continues, contact Support with the help page that lists the paths it installs.

See also

Last updated