# Building a Custom Agent Check
---
# Agent Architecture
## Four components
### Collector
### Forwarder
### DogstatsD
### SupervisorD
--
# The Collector
## Runs every 15 seconds
## Runs each enabled integration check
## Also runs any custom agent checks
--
# The Forwarder
## Forwards collected metrics to Datadog
## Stores metrics collected in case of network disruption
--
# DogStatsD
## Python implementation of StatsD (Etsy)
## Implements tags
--
# SupervisorD
## Runs master process
## Forks all subprocesses as the user **dd-agent**
---
# Running Checks
## Every 15 seconds Collector collects
## Official integrations
## Custom Checks
--
# Custom Checks
## Checks have both a Python check and YAML configuration
--
# Custom Checks
## Checks in `/etc/dd-agent/checks.d`
## Configurations in `/etc/dd-agent/conf.d`
**On Linux, see http://docs.datadoghq.com/guides/agent_checks/#directory for other platforms.*
--
# Example Configuration
```
init_config:
min_collection_interval: 20
key1: val1
key2: val2
instances:
- username: jon_smith
password: 1234
- username: jane_smith
password: 5678
```
--
# Checks
## Inherit from **AgentCheck**
## Require a `check(self, instance)` method
--
# Simple Check
```
from checks import AgentCheck
class HelloCheck(AgentCheck):
def check(self, instance):
self.gauge('hello.world', 1)
```
---
# Datadog Metric Types
## Gauge
## Rate
## Count
--
# Gauge
* Record any value
* If not updated, uses previous value
--
# Rate
* Count per unit of time
--
# Count
* Quantity
